WO2022166232A1 - 岩石鉴定方法、系统、装置、终端及可读存储介质 - Google Patents

岩石鉴定方法、系统、装置、终端及可读存储介质 Download PDF

Info

Publication number
WO2022166232A1
WO2022166232A1 PCT/CN2021/121840 CN2021121840W WO2022166232A1 WO 2022166232 A1 WO2022166232 A1 WO 2022166232A1 CN 2021121840 W CN2021121840 W CN 2021121840W WO 2022166232 A1 WO2022166232 A1 WO 2022166232A1
Authority
WO
WIPO (PCT)
Prior art keywords
rock
mineral
image
component
slice
Prior art date
Application number
PCT/CN2021/121840
Other languages
English (en)
French (fr)
Inventor
余晓露
赵永强
倪春华
马中良
周生友
杨伟利
张俊
陶成
李贶
王强
郑伦举
Original Assignee
中国石油化工股份有限公司
中国石油化工股份有限公司石油勘探开发研究院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from CN202110181773.4A external-priority patent/CN112730326A/zh
Application filed by 中国石油化工股份有限公司, 中国石油化工股份有限公司石油勘探开发研究院 filed Critical 中国石油化工股份有限公司
Priority to US18/264,654 priority Critical patent/US20240054766A1/en
Priority to EP21924229.4A priority patent/EP4290470A1/en
Priority to JP2023548202A priority patent/JP2024508688A/ja
Publication of WO2022166232A1 publication Critical patent/WO2022166232A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/7715—Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/09—Supervised learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/90—Determination of colour characteristics
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/10—Image acquisition
    • G06V10/12—Details of acquisition arrangements; Constructional details thereof
    • G06V10/14—Optical characteristics of the device performing the acquisition or on the illumination arrangements
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10024—Color image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10056—Microscopic image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20024—Filtering details
    • G06T2207/20032—Median filtering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]

Definitions

  • the present application relates to the technical field of rock identification, and in particular, to a rock identification method, device, terminal and readable storage medium.
  • the traditional rock thin section identification method is still in use today.
  • researchers usually identify the rock thin section through microscopic observation of the rock thin section by manual identification.
  • researchers will also make more detailed observations of microscopic images by using equipment and computers to collect microscopic images.
  • the present application relates to a rock identification method, system, device, terminal and readable storage medium, which can improve the accuracy of rock identification.
  • the technical solution is as follows:
  • a rock identification method is provided, which is applied to a computer device of a rock identification system, wherein the rock identification system includes an image acquisition device and a computer device; the image acquisition device is connected to the computer device, and the method includes:
  • rock slice image sent by an image acquisition device, where the rock slice image is an image obtained by photographing a rock slice, the rock slice is a slice obtained by cutting a rock sample, and the rock slice image includes at least one component region;
  • the method further includes:
  • the segmentation model of rock slice images is adjusted based on segmentation differences.
  • the method further includes:
  • the method further includes:
  • the component species identification model is adjusted based on identification differences.
  • the non-single-crystalline component set image includes an amorphous component set image and a polycrystalline component set image
  • the non-single-crystalline component area includes an amorphous component area and a polycrystalline component set image.
  • the amorphous component area corresponds to the identification of the amorphous component
  • the polycrystalline component area corresponds to the identification result of the polycrystalline component.
  • the rock flakes are clastic rock flakes
  • the single crystal component region includes the quartz feldspar clastic region and the siliceous calcareous cement region, and the non-single crystal component region includes the non-terrigenous clastic region, the detrital clastic region, the heterobasic interstitial material region and the mud area of cementitious material;
  • the classification and recognition model includes a classification sub-model group and an identification sub-model group;
  • the classification sub-model group includes particle classification sub-model, debris classification sub-model, terrigenous debris classification sub-model, interstitial material classification sub-model and cement classification sub-model;
  • the identification sub-model group includes non-terrestrial debris identification sub-model, lithic debris identification sub-model, quartz feldspar clastic identification sub-model, siliceous calcareous cement identification sub-model, argillaceous cement identification sub-model and miscellaneous cement identification sub-model.
  • the segmented images of rock slices are input into the classification and recognition model, and the output of single-crystal component collection images and non-single-crystal component collection images, including:
  • the cement sub-image is input into the cement classification sub-model, and the siliceous calcareous cement sub-image and the argillaceous cement sub-image are outputted;
  • the shale cement sub-image is input into the shale cement identification sub-model, and the shale cement identification result is output, and the shale cement region is determined based on the shale cement identification result;
  • the single crystal component region is determined
  • the non-single-crystal composition region is determined based on the non-terrigenous clastic region, the lithic clastic region, the argillaceous cement region, and the hetero-base interstitial region.
  • the method further includes:
  • sample slice image is marked with the sample segmentation area and the sample component type characteristics
  • the rock slice image recognition model is adjusted.
  • the identification result of the rock slice includes the identification result of the largest particle size
  • rock thin section structure identification results including:
  • the maximum particle size corresponding to the component region is determined based on the geometric features, and a maximum particle size identification result is generated based on the maximum particle size.
  • the rock slice identification results include particle sorting results
  • rock thin section structure identification results including;
  • the sorting degree component area set includes at least two component areas
  • the particle sorting degree result is determined.
  • the identification result of the rock slice includes the identification result of particle roundness
  • rock thin section structure identification results including:
  • the roundness identification component region set includes at least two component regions
  • the roundness grading correspondence relationship Based on the geometric features corresponding to the component regions in the roundness component identification region set, determine the roundness grading correspondence relationship, where the roundness grading correspondence includes at least two roundness levels, and the roundness level is in the roundness level the number of component regions;
  • the particle roundness identification result is determined.
  • the identification result of the rock slice includes the result of the degree of particle eumorphism
  • rock thin section structure identification results including:
  • the euhedral degree component region set corresponding to the grain euhedral degree result based on the mineral characteristics, and the euhedral degree component region set includes at least two euhedral degree component regions;
  • the identification result of the rock slice includes the identification result of the particle contact mode
  • rock thin section structure identification results including:
  • the identification result of the particle contact mode is determined.
  • the rock slice name is determined based on the rock slice identification result, including:
  • the rock slice name is generated.
  • a rock identification system in another aspect, includes an image acquisition device and a computer device, and the image acquisition device is connected with the computer device;
  • Image acquisition equipment for generating rock slice images; sending rock slice images to computer equipment;
  • Computer equipment for receiving rock slice images sent by image acquisition equipment where the rock slice images are images obtained by photographing rock slices, the rock slices are slices obtained by cutting rock samples, and the rock slice images include at least one component region ; Generate geometric features, mineral features, and structural features corresponding to the rock thin section based on the rock thin section image, wherein the geometric features are used to indicate the division of the component regions of the rock thin section, and the mineral features are used to indicate the composition area in the rock thin section.
  • the corresponding mineral species distribution, the obtained mineral identification results, and the structural features are used to indicate the rock species of the rock sample; based on the geometric characteristics, mineral characteristics and structural characteristics, the identification results of rock slices are generated.
  • a rock identification device comprising:
  • the receiving module is used for receiving the rock slice image sent by the image acquisition device, the rock slice image is an image obtained by photographing the rock slice, the rock slice is the slice obtained by cutting the rock sample, and the rock slice image includes at least one component area ;
  • the generation module is used to generate geometrical features, mineral features and structural features corresponding to the rock thin section based on the rock thin section image, wherein the geometric feature is used to indicate the division of the component regions of the rock thin section, and the mineral feature is used to indicate the composition of the rock thin section.
  • the distribution of mineral species corresponding to the sub-region, the obtained mineral identification results, and the structural characteristics are used to indicate the rock species of the rock sample;
  • the generation module is used to generate identification results of rock slices based on geometrical, mineral and structural features.
  • a computer device in another aspect, includes a processor and a memory, the memory stores at least one instruction, at least one program, a code set or an instruction set, and the processor can load and execute at least one instruction, at least one program , a code set or an instruction set to implement the rock identification method provided in the above embodiments of the present application.
  • a computer-readable storage medium in which at least one instruction, at least one piece of program, code set or instruction set is stored, and the processor can load and execute at least one instruction, at least one piece of program, code set set or instruction set to implement the rock identification method provided in the above embodiments of the present application.
  • a computer program product or computer program comprising computer program instructions stored in a computer-readable storage medium.
  • the processor reads the computer instructions from the computer-readable storage medium, and executes the computer instructions, so that the computer device performs the rock identification method as provided in the embodiments of the present application.
  • the rock slice images are extracted from three dimensions based on geometric features, mineral characteristics and structural features.
  • the final result is an identification result including a textual description.
  • the image is extracted with multiple dimensions of features, and the rock slice is identified with reference to the features of multiple dimensions, which improves the performance of the rock slice. Accuracy of rock identification.
  • Fig. 1 shows a schematic diagram of a rock identification system provided by an exemplary embodiment of the present application
  • FIG. 2 shows a schematic structural diagram of an image acquisition device provided by an exemplary embodiment of the present application
  • Fig. 3 shows a flow chart of a rock identification method provided by an exemplary embodiment of the present application
  • FIG. 4 shows a schematic structural diagram of a rock slice image provided by an exemplary embodiment of the present application
  • FIG. 5 shows a flowchart of a method for determining geometric features, mineral features and structural features provided by an exemplary embodiment of the present application
  • Fig. 6 shows a schematic flowchart of a method for segmenting a rock slice image provided by an exemplary embodiment of the present application
  • FIG. 7 shows a structural block diagram of an image segmentation model provided by an exemplary embodiment of the present application.
  • FIG. 8 shows a schematic diagram of a segmented rock slice segmented image provided by an exemplary embodiment of the present application
  • FIG. 9 shows a flowchart of a method for identifying a mineral feature provided by an exemplary embodiment of the present application.
  • Fig. 10 shows a schematic diagram of extracting a segmented region from a segmented image of a rock slice provided by an exemplary embodiment of the present application
  • Figure 11 shows a schematic diagram of another rock identification system provided by an exemplary embodiment of the present application.
  • Fig. 12 shows a schematic device diagram of a mineral data acquisition device provided by an exemplary embodiment of the present application
  • FIG. 13 shows a schematic flowchart of a method for identifying mineral characteristics corresponding to rock slices provided by an exemplary embodiment of the present application
  • Fig. 14 shows a schematic diagram of a rock slice corresponding to clastic rock provided by an exemplary embodiment of the present application
  • Fig. 15 shows a schematic structural diagram of a rock identification model corresponding to clastic rock provided by an exemplary embodiment of the present application
  • Fig. 16 shows a schematic flowchart of a method for determining a mineral feature corresponding to a component region provided by an exemplary embodiment of the present application
  • FIG. 17 shows a schematic structural diagram of a mineral spectrum database provided by an exemplary embodiment of the present application.
  • FIG. 18 shows a schematic diagram of a process for determining geometric features and mineral features provided by an exemplary embodiment of the present application
  • Fig. 19 shows a schematic diagram of a method for generating identification results of rock slices provided by an exemplary embodiment of the present application
  • Fig. 20 shows a schematic process diagram of a rock identification method provided by an exemplary embodiment of the present application
  • Figure 21 shows a schematic diagram of the content of a thin slice identification report provided by an exemplary embodiment of the present application.
  • Fig. 22 shows a schematic process diagram of a rock identification method provided by an exemplary embodiment of the present application
  • Figure 23 shows a schematic process diagram of a rock identification method provided by an exemplary embodiment of the present application.
  • Fig. 24 shows a schematic diagram of a classification facies diagram of an intrusive rock provided by an exemplary embodiment of the present application
  • Fig. 25 shows a structural block diagram of a rock identification device provided by an exemplary embodiment of the present application
  • Fig. 26 shows a structural block diagram of another rock identification device provided by an exemplary embodiment of the present application.
  • Fig. 27 shows a schematic structural diagram of a computer device for performing a rock identification method provided by an exemplary embodiment of the present application.
  • Components refer to the individual components in a mixture.
  • constituents refer to the constituents of solid state materials. Since the solid material is a mixture, at least two separate components will be included in the mixture.
  • the solid-state material is also referred to as a solid-state multi-component hybrid material.
  • the mixture is realized as a metal wire, and the composition of the metal wire includes at least one of lead, cadmium, bismuth, and tin; in another example, the mixture is realized as a rock, and the composition of the rock includes but not Limited to at least one of terrigenous detrital components, non-terrestrial detrital components, cement components, and hetero-based components.
  • the composition refers to the composition contained in the rock slice, and in the rock slice image, there are different composition regions corresponding to different compositions.
  • Image processing technology is a technology for processing image information by computer.
  • Image processing technology mainly includes image digitization, image enhancement and restoration, image data encoding, image segmentation and image recognition.
  • image segmentation is the technology and process of dividing an image into several specific and unique regions and proposing objects of interest.
  • the method of image segmentation includes at least one of a threshold-based segmentation method, a region-based segmentation method, an edge-based segmentation method, and a specific theory-based segmentation method.
  • Image recognition refers to the technology of using computers to process, analyze and understand images to identify various patterns of targets and objects. It is a practical application of deep learning algorithms. At this stage, image recognition technology is generally divided into face recognition and commodity recognition.
  • Face recognition is mainly used in security inspection, identity verification and mobile payment; commodity recognition is mainly used in the process of commodity circulation, especially unmanned shelves, smart retail cabinets, etc. Unmanned retail field.
  • image recognition and image segmentation technologies will be applied to the field of rock identification.
  • Artificial intelligence is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. Artificial intelligence attempts to understand the essence of intelligence and produce an intelligent machine that can respond in a similar way to human intelligence. The purpose of artificial intelligence is to enable machines to have the functions of perception, reasoning and decision-making.
  • Artificial intelligence technology is a comprehensive discipline covering a wide range of fields. Artificial intelligence basic technologies include but are not limited to sensor technology, artificial intelligence chip technology, cloud computing technology, big data processing technology, and mechatronics technology.
  • the artificial intelligence technology applied in the embodiments of the present application is a machine learning technology, and the machine learning device is applied to a computer device.
  • Machine Learning is a multi-disciplinary interdisciplinary subject involving probability theory, statistics, algorithm complexity theory and other disciplines.
  • the discipline of machine learning is specially used to study how computers simulate or realize human learning behaviors, so that computers can acquire new knowledge, reorganize existing knowledge structures, and then improve their own performance.
  • Machine learning is usually combined with deep learning.
  • Machine learning and deep learning usually include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, teaching learning and other technologies.
  • the spectrum is a pattern in which the dispersed monochromatic light is arranged at one time according to the wavelength after the polychromatic light is split by the dispersion system.
  • the full name of the spectrum is the optical spectrum.
  • the spectrum is obtained by collecting the measurement points in the component area by the spectrum data collecting device. Taking the component as mineral component as an example, when the mineral is irradiated by electromagnetic radiation, the molecules inside it will produce a transition between quantized energy levels, which will lead to emission, absorption or scattering, and radiation phenomena, thereby generating a spectrum. .
  • Spectroscopy is implemented as a visual representation of spectral data. That is, while generating the spectrum, the spectral data acquisition device will also generate corresponding spectral data at the same time, and send such spectral data to other computer devices.
  • a database is a virtual warehouse that organizes, stores and manages data according to the data structure.
  • the computer device can retrieve the spectral data database to confirm and analyze the spectral data.
  • Test instruments used in the area scan include but are not limited to molecular spectroscopy such as infrared spectroscopy and Raman spectroscopy, X-ray Fluorescence spectroscopy (XRF), X-ray energy spectroscopy (Energy Dispersive X-ray Spectroscopy, EDS) etc., and in some embodiments of the present application, the area scanning process can be combined with a microscope or a scanning electron microscope, combining microscopic observation with micro-area analysis.
  • molecular spectroscopy such as infrared spectroscopy and Raman spectroscopy
  • XRF X-ray Fluorescence spectroscopy
  • EDS X-ray energy spectroscopy
  • the area scanning process can be combined with a microscope or a scanning electron microscope, combining microscopic observation with micro-area analysis.
  • the present application provides a rock identification system and a rock identification method in view of the complex identification process of the artificially identified rock identification method in the related art, the professional requirements are extremely high, but the accuracy rate is low, and with the help of the artificial intelligence-based image identification technology, In the process of rock identification, multi-dimensional features are generated, which improves the accuracy of rock identification.
  • FIG. 1 shows a schematic diagram of a rock identification system provided by an exemplary embodiment of the present application.
  • the apparatus includes an image acquisition device 101 and a computer device 102 , and the image acquisition device 101 is connected to the computer device 102 .
  • the computer equipment is the terminal equipment for executing the rock identification method.
  • the computer equipment has the functions of data transmission, data reception and data processing.
  • the image acquisition device is a device that provides images to the computer device during the process of the computer device performing the rock identification method.
  • the image acquisition device has the functions of image recording and image sending.
  • FIG. 2 shows a schematic structural diagram of an image acquisition device provided by an exemplary embodiment of the present application.
  • the image acquisition device includes a first light source 201, a first stage 202, a first objective lens 203, A polarizing device 204 , a Charge-Coupled Device (CCD) camera 205 , a first base 206 and a first support 207 . Please refer to FIG.
  • CCD Charge-Coupled Device
  • the first stage 202 is located in the center of the first bracket 207
  • the first objective lens 203 is located on the top of the first bracket
  • the first objective lens 203 has an analyzer
  • the bottom of the first objective lens 203 has a first installation objective lens Converter
  • the first installation objective lens converter is connected with the first objective lens 203 .
  • the CCD camera 205 is located on the top of the first objective lens 203
  • the polarizing device 204 includes a polarizer and an analyzer
  • the polarizer and the first light source 201 are located on the first base
  • the first base 206 is connected to the bottom of the first bracket 207 .
  • the first light source 201 When the image acquisition device 200 generates a rock slice image, the first light source 201 is activated and emits light, and the light path corresponding to the light passes through the polarizer and the analyzer, and finally projects onto the rock slice.
  • the rock slice is on top of the first stage 202 , that is, placed on the first stage 202 .
  • the CCD camera is in the first working state, that is, the CCD camera corresponds to the position of the objective lens, and captures a microscopic image corresponding to the rock thin section as a rock thin section image.
  • a communication connection is established between the image acquisition device and the computer device. Based on the communication connection, the image capture device can send the image to the computer device for storage. After the computer equipment receives the image sent by the image acquisition equipment, the image is processed to obtain the characteristics of the suitable rock identification process related to this scheme, and the rock identification is carried out based on the above-mentioned characteristics of the suitable rock identification process.
  • FIG. 3 shows a flow chart of a rock identification method provided by an exemplary embodiment of the present application, which is described by applying the method to the computer equipment of the rock identification system, and the method includes:
  • Step 301 receiving a rock slice image.
  • the computer device before receiving the rock slice image sent by the image acquisition device, the computer device sends an image acquisition instruction to the image acquisition device, and after receiving the image acquisition instruction, the image acquisition device performs image acquisition according to the image acquisition instruction, and The acquired images are sent to a computer device.
  • the computer equipment screens and processes the images sent by the image acquisition equipment, and finally obtains rock slice images.
  • a rock slice is a slice obtained by cutting a rock sample, that is, the components in the rock slice can represent and characterize the components of the rock sample.
  • the rock slice image also includes at least one component area, and each component area includes one component.
  • the component refers to the component corresponding to the rock. Please refer to FIG.
  • the rock slice image 400 since the rock slice image 400 is an image obtained by microphotography of the rock slice, the rock slice image includes a component area 401 .
  • the number of component areas is 18, and the 18 component areas are
  • the component region may represent the same component, or may represent components different from each other, or may represent at least two components that are different from each other, and there may be cases where the components in the component region are the same.
  • the embodiment of the present application does not limit the content of the component area.
  • the components included in the component region are realized as mineral components, and may also be realized as structural components.
  • the composition field when the content included in the composition field is a mineral composition, the composition field may indicate a type of mineral.
  • the composition area When the content included in the composition area is a structural component, the composition area may indicate a type of particle, or, a type of interstitial.
  • the number of rock slice images in the embodiment of the present application is 2, which are the single polarized light image and the orthogonal polarized light image corresponding to the rock slice respectively; or, the number of rock slice images in the embodiment of the present application is 1 , and the image is obtained by combining the single-polarization image and the cross-polarization image corresponding to the rock slice.
  • the embodiment of the present application does not limit the actual representation of the rock slice image, but the rock slice image needs to reflect the anisotropy exhibited by the rock slice under ordinary light and polarized light sources.
  • Step 302 generating geometrical features, mineral features and structural features corresponding to the rock thin section based on the rock thin section image.
  • the computer equipment when a rock thin section image is received, the computer equipment will perform feature extraction on the rock thin section image corresponding to the properties of the rock thin section, corresponding to the geometrical, mineral, and structural features of the rock thin section.
  • the geometric feature is used to indicate the composition-based regional division of the rock thin section
  • the mineral feature is used to indicate the mineral identification result obtained from the rock thin section
  • the structural feature is used to indicate the spatial structure of the rock sample.
  • the distribution image of the component area in the rock thin section image can represent the geometric feature
  • the specific mineral type of the component area in the rock thin section image can represent the mineral feature
  • the spatial structure property reflected in the rock thin section image can represent the Structure.
  • the embodiments of the present application do not limit the specific contents of geometric features, mineral features, and structural features.
  • Step 303 generating an identification result of the rock slice based on the geometrical features, mineral features and structural features.
  • the identification result of the rock slice is generated based on the above-mentioned features.
  • the identification results include a textual description of the characteristics of the rock sample.
  • the identification results include the naming of the rock sample, and a textual summary of the geometric, mineral, and structural features.
  • the identification result is expressed in the form of a list, or the identification result further includes the visual image content and is expressed in the form of a chart. The present application does not limit the generation method of the identification structure.
  • the rock slice image is extracted based on three dimensions of geometric features, mineral features and structural features, from the perspective of microscopic composition and macroscopic performance. , multiple feature dimensions determine the properties of the rock, and finally generate identification results including textual descriptions.
  • the image is extracted with multiple dimensions of features, and the rock slice is identified with reference to the features of multiple dimensions, which improves the performance of the rock slice. Accuracy of rock identification.
  • FIG. 5 shows a flowchart of a method for determining geometric features, mineral features and structural features provided by an exemplary embodiment of the present application. The method can replace step 302 shown in FIG. 3 and be implemented as steps 501 to 506 , the method includes:
  • Step 501 generating a primary rock structure feature based on the rock slice image.
  • the structural features include primary rock structural features and subdivided rock structural features.
  • the structural characteristics of the primary rock types are used to classify the rock samples corresponding to the thin slices, and are classified into the primary rock types.
  • the primary rock types are also sedimentary rocks, magmatic rocks and metamorphic rocks.
  • the rock slice image is input into the primary rock type selection model, and the primary rock type result is obtained as output, and the primary rock type result can indicate the primary rock type structural feature corresponding to the rock slice.
  • the primary rock structure selection model is a model constructed based on the primary rock image sample set. After the rock slice image is input into the primary rock type selection model, the primary rock type selection model will compare the rock slice image with the sample images saved in the sample set, and then determine the structural characteristics of the primary rock type.
  • Step 502 Determine the primary rock type corresponding to the rock slice based on the structural characteristics of the primary rock type.
  • the results of the primary selection of rock types include sedimentary rocks, magmatic rocks, and metamorphic rocks, that is, the primary selection of rock types are three types of rocks obtained by genesis.
  • the primary rock types can be obtained.
  • the primary rock type structural features are implemented as similarity to the corresponding primary rock type.
  • Step 503 Divide the rock slice image based on the position of the component image according to the primary rock type to obtain a rock slice segmented image, where the rock slice segmented image includes at least two segmented regions.
  • the computer device can further determine the content of the component area corresponding to the primary rock type, and the common shape of the component area.
  • the computer device segments the rock slice image based on the content of the component area and the common shape of the component area to obtain a rock slice segmented image, where the rock slice segment image includes at least two segmented areas.
  • the segmented area is used to indicate the segmentation of the rock slice image based on the component area it has.
  • Step 504 determining geometric features corresponding to the rock slices based on the segmented regions.
  • the geometric features corresponding to the rock slices can be determined based on the segmentation area.
  • the computer equipment regards the segmented area as a component area to identify geometric features.
  • the geometric feature can be used to characterize at least one of the area of the component region, the maximum diameter of the component region, and the shape of the component region.
  • Step 505 determining mineral features corresponding to the rock slices based on the segmented regions.
  • the mineral characteristics corresponding to the rock slices can be further determined based on the primary rock structure features.
  • the mineral feature is used to indicate the mineral species of the components in each segmented area.
  • the geometric features corresponding to the rock slices are first determined, and then the mineral characteristics corresponding to the rock slices are determined; in some other embodiments of the present application, the first determination is The mineral feature corresponding to the rock slice is secondly determined the geometric feature corresponding to the rock slice; in some other embodiments of the present application, the geometric feature corresponding to the rock slice and the mineral feature are simultaneously determined.
  • the present application does not limit the determination order of geometric features and mineral features.
  • Step 506 based on the geometric features and mineral features, determine the structural features of the subdivided rock types.
  • the subdivision rock structure feature can be further determined, and the subdivision rock structure feature is used to indicate the subdivision rock type corresponding to the rock sample, that is, in the genetic
  • the species are further determined.
  • the geometric features and mineral features are input into the subdivision rock type selection model, and the subdivision rock type result is obtained as output, and the subdivision rock type result indicates the subdivision rock type structure feature corresponding to the rock slice, and the subdivision rock type selection model It is a model constructed based on the geometry-mineral feature interaction sample set.
  • the geometry-mineral feature interaction sample set indicates the combination of geometric features and mineral features, and the corresponding relationship between the subdivision rock structure features.
  • the primary rock type structure feature of the corresponding rock sample indicates that the rock sample is sedimentary rock, and the rock sample is obtained according to the properties of the segmented regions in the thin slice of the rock sample and the type of the segmented region through the subdivision rock type selection model.
  • the corresponding subdivision rock structure features indicate that the rock samples are clastic rocks in sedimentary rocks.
  • both the primary rock type and the subdivision rock type are included in the identification result.
  • the method provided in the embodiments of the present application corresponding to the actual application content of structural features for rock identification, classifies the structural features into the structural features of primary rock types and the structural characteristics of subdivided rock types, and determines the primary rock type when determining the actual application content of rock types.
  • the geometric and structural features corresponding to the rock slice images are further determined, and then the structural features of the subdivided rock types are determined, so that in the process of rock identification, the determination process of geometric features and structural features is more targeted and further improved. accuracy of rock identification.
  • FIG. 6 shows a schematic flowchart of a method for segmenting a rock slice image provided by an exemplary embodiment of the present application, and the method can replace steps 503 to 504 in the embodiment shown in FIG. 5 .
  • the method includes:
  • step 601 a sample slice image is acquired, and the sample slice image is marked with a sample segmentation result.
  • the embodiments of the present application mainly illustrate the construction and application process of a rock thin section image segmentation model, and the rock thin section image segmentation model is a model for dividing component regions.
  • the machine learning process is implemented by means of sample training.
  • Step 601 is the process of acquiring a sample slice image, that is, constructing a sample set.
  • the sample slice image is an image selected from the content of the corresponding rock slice image.
  • the sample thin section image is a rock thin section image corresponding to a primary rock species of the rock thin section; in another example, the sample thin section image is a rock thin section image corresponding to a subdivided rock species of the rock thin section.
  • the present application does not limit the specific rock types corresponding to the sample slice images.
  • sample slice images are annotated with sample segmentation results.
  • the sample segmentation result is the method of dividing the determined sample slice image based on the segmentation area of the component area after the judgment is made by other methods.
  • Step 602 Input the sample slice image into the rock slice image segmentation model, and output the predicted segmentation result.
  • the rock slice image segmentation model is a Mask-RCNN network model based on machine learning.
  • the rock slice image segmentation model 700 includes a feature image generation network 701 , a region proposal network 702 and a result generation network 703 , and the three sub-networks are connected to each other in the rock slice image segmentation model.
  • the feature image generation network 701 includes a feature extraction sub-network 7011 and a feature enhancement sub-network 7012
  • the result generation network 703 includes a result region classification branch network 7031 and a border regression branch network 7032.
  • the preliminary segmentation feature image is input into the regional classification branch network, the properties of the preliminary segmentation feature image are classified through the regional classification branch network, and the preliminary segmentation feature image is corrected through the border regression branch network, that is, the final Output the segmentation result.
  • the predicted segmentation result is obtained by inputting the sample slice image.
  • the predicted segmentation results can be output in the form of sample rock slices superimposed with segmentation lines.
  • Step 603 Compare the difference between the predicted segmentation result and the sample segmentation result to obtain the segmentation difference.
  • This process is a process of comparing the predicted segmentation result with the sample segmentation result to determine the difference between the two.
  • the reason for the difference between the predicted segmentation results and the sample segmentation results is that the parameter adjustment of the rock slice image segmentation model has not been completed.
  • the segmentation difference is implemented as the length difference of the segmentation lines in the rock slice segmented image; in another example, the segmentation difference is implemented as the shape difference of the segmentation lines in the rock slice segmented image.
  • the present application does not limit the specific implementation form of dividing the difference.
  • Step 604 Adjust the rock slice image segmentation model based on the segmentation difference.
  • each "threshold” that appears in this application can be implemented as a threshold stored in a computer program, or can be implemented as a threshold manually input when a computer device performs corresponding steps.
  • the actual acquisition method of the threshold in this application Not limited.
  • the rock thin section image is segmented to extract the set features.
  • Step 605 in response to the completion of the training of the rock slice image segmentation model, input the rock slice image into the rock slice image segmentation model, and output the rock slice segment image.
  • the process corresponding to step 605 is the process of generating a rock slice segmented image corresponding to the rock slice, and the rock slice segmented image includes at least two segmented regions. Split situation.
  • the segmented area can directly represent the component area.
  • FIG. 8 in the rock slice segmented image 800 obtained after going through the rock slice image segmentation model, a plurality of component regions 802 are divided in the manner of dividing lines 801 . Taking FIG. 8 as an example, the area completely surrounded by the dividing line 801 is the component area 802 .
  • the segmentation result corresponding to the rock slice image can be generated.
  • the segmentation result includes at least one of the number of component regions corresponding to the rock thin section image, the distribution density of the component regions corresponding to the rock thin section image, and the size of the component regions corresponding to the rock thin section image.
  • the above-mentioned segmentation result may directly indicate geometric features corresponding to the rock slices, or the above-mentioned segmentation results may be processed to generate geometric features corresponding to the rock slices.
  • the rock thin section image is segmented through the rock thin section image segmentation model based on the Mask-RCNN network to obtain the rock thin section image.
  • the rock thin section image segmentation model based on the Mask-RCNN network is introduced, which provides better division guidance for the component regions in the rock slice images. The acquisition efficiency of geometric features is improved.
  • the embodiment shown in FIG. 6 is an embodiment in which a computer device acquires geometric features.
  • the geometric features also include other acquisition methods.
  • the present application does not limit the specific acquisition method of the geometric feature.
  • the geometric features indicate the division of the component regions, and each component region corresponds to different components, after the geometric features are determined, the mineral features are determined, that is, the determination of mineral distribution.
  • FIG. 9 shows a flowchart of a method for identifying a mineral feature provided by an exemplary embodiment of the present application, and the method can be implemented as step 505 in the embodiment shown in FIG. 5 instead. Taking the method applied to computer equipment as an example, the method includes:
  • Step 901 Determine a component image based on the segmented area, where the component image includes a component area.
  • the segmentation area 1001 is the segmentation result determined based on the rock slice segmentation image 1000 , and the segmentation area 1001 can be used to determine the component image 1010 , and the component image 1010 includes a component area. Since the rock slice image segmentation model has been trained, this component area corresponds to the component area included in the rock slice, and directly characterizes the components in the rock slice.
  • Step 902 Obtain a sample component image, where the sample component image is marked with a sample component type feature.
  • the component type feature corresponding to the component area is determined by the component type identification model, that is, the mineral composition corresponding to the component area is determined.
  • the construction of the component species identification model is similar to the construction of the rock slice image segmentation model. All samples are trained based on machine learning.
  • the sample component images are selected from the sample component image dataset, and each sample component image includes and only includes one component area, and the sample component area corresponds to the sample component type feature.
  • Step 903 Input the sample component image into the component type identification model, and output the predicted component type feature.
  • the predicted component type feature is the predicted component type feature obtained by identifying the sample component image by the component type identification model in the training process.
  • the formation neural network that can be selected by the component type identification model includes at least one of Mobilenet neural network, Resnet neural network, and VGG (Visual Geometry Group Network) neural network.
  • Step 904 compare the difference between the predicted component type feature and the sample component type feature to obtain the identified difference.
  • This process is the process of identifying the difference between the predicted component type feature and the sample component type feature.
  • the component type feature is represented in the form of a numerical value, and different numerical values represent different component types, and the probability that the component region belongs to the component type.
  • identifying the difference may indicate a difference between the predicted component type characteristic and the sample component type characteristic, or a ratio of the predicted component type characteristic to the sample component type characteristic.
  • the embodiment of the present application does not limit the actual output form of the component type feature.
  • Step 905 Adjust the component type identification model based on the identification difference.
  • This process is to adjust the component type identification model according to the component type identification difference until the difference is smaller than the difference threshold, or the training times of the component type identification model reaches the training times threshold.
  • Step 906 in response to the completion of the training of the component type identification model, input the component image into the component type identification model, and output the component type feature of the component image.
  • This process is the process of outputting component type features. It should be noted that the embodiment of the present application focuses on the identification of a single component in the component image, so when the component image inevitably includes two component regions, the component type identification model will The two component areas are identified uniformly, or the component type identification model will select an image area corresponding to any component area from the above two component areas for separate identification. Corresponding to the process of one component identification, the component type identification model will output only one component type feature.
  • Step 907 based on the component type feature corresponding to the component image, determine the mineral feature corresponding to the rock slice.
  • the computer device stores the correspondence between the component type features and the mineral species. After the component type features are output, the computer device can judge from the correspondence between the component type features and the mineral species. The mineral species corresponding to the component area.
  • the mineral characteristics corresponding to the rock slice can be obtained.
  • the method provided in the embodiment of the present application corresponds to a single component region, and performs component identification by inputting it into the component type identification model, and outputs the component type indicating the mineral characteristics of the component region. feature.
  • the identification of the detailed types of component particles is carried out by means of a machine learning-based recognition model, which improves the identification efficiency of components in rock slices.
  • the preliminary determination of mineral characteristics is performed based on the classification and identification model, and then the preliminary determination result of the classification and identification model is verified by the mineral data acquisition device.
  • mineral data acquisition equipment is also included in the rock identification system.
  • Fig. 11 shows a structural block diagram of a rock identification system provided by an exemplary embodiment of the present application.
  • the rock identification system includes an image acquisition device 1101, a computer device 1102, and a mineral data acquisition device 1103, wherein the image acquisition device 1101 and Mineral data acquisition devices 1103 are connected to computer devices 1102, respectively.
  • FIG. 11 shows a structural block diagram of a rock identification system provided by an exemplary embodiment of the present application.
  • the rock identification system includes an image acquisition device 1101, a computer device 1102, and a mineral data acquisition device 1103, wherein the image acquisition device 1101 and Mineral data acquisition devices 1103 are connected to computer devices 1102, respectively.
  • FIG. 11 shows a structural block diagram of a rock identification system provided by an exemplary embodiment of the present application.
  • the rock identification system includes an image acquisition device 110
  • the mineral data acquisition device includes a second light source 1201, a second stage 1202, a second objective lens 1203, The photoelectric signal converter 1204, the second base 1205 and the second bracket 1206; the second light source 1201 corresponds to a second optical path 1207, and the end point of the second optical path 1207 is projected to the second stage 1202;
  • the second objective lens 1203 is located on the top of the second bracket 1207, the bottom of the second objective lens 1203 has a second installation objective lens converter, and the second installation objective lens converter is connected with the second objective lens 1203;
  • the photoelectric signal converter 1205 is located in The top of the second objective lens 1203; when the mineral data acquisition device 1200 generates mineral data, the rock slice is located on the top of the second stage 1202, and the photoelectric signal converter 1204 is in the second working state to acquire spectral data.
  • the second working state is the state
  • the mineral data acquisition device and the image acquisition device can be implemented as the same device, that is, the mineral data acquisition device and the image acquisition device share the first stage, the first objective lens, the first base, and the first bracket. Both the CCD camera and the photoelectric signal converter are located on the first objective lens.
  • the integrated device of the mineral data acquisition device and the image acquisition device can acquire the rock thin section image and spectral data, and send the rock thin section image and spectral data to the computer equipment together. .
  • FIG. 13 shows a schematic flowchart of an identification method for determining mineral characteristics corresponding to rock slices provided by an exemplary embodiment of the present application.
  • the method can be alternatively implemented as step 505 in the embodiment corresponding to FIG. 5 .
  • the method includes:
  • Step 1301 Input the segmented image of the rock slice into the classification and recognition model, and output a single crystal component set image and a non-single crystal component set image.
  • the classification and recognition model is used to distinguish the component regions belonging to different crystalline types in the segmented image of rock slices.
  • the non-single crystalline components include polycrystalline components and amorphous components.
  • a crystalline component classification model is included in the classification recognition model to distinguish components into single crystal components and non-single crystalline components, and the classification recognition model further includes non-single crystalline component classification model, which divides non-single-crystalline components into poly-crystalline components and amorphous components, and sets up the identification model of single-crystalline components, the identification model of poly-crystalline components and the identification model of amorphous components correspondingly. In order to identify the specific mineral species corresponding to the component area.
  • the classification sub-model in the classification and identification model and the specific type of the identification sub-model can be set to adapt to the situation of different component regions included in different rock types.
  • the rock type of the rock slice is clastic rock.
  • the components indicated by the component area in the clastic rock include clastic components and interstitial components.
  • detritus is a component of sedimentary rock or sediment, which is the product of the mechanical elegance of the parent rock; interstitial matter is the connecting material between each detritus.
  • the detrital components were divided into terrigenous and non-terrigenous detrital components.
  • terrigenous detritus refers to the detrital material formed by physical weathering or mechanical damage of the parent rock in the terrigenous area, mainly including quartz, feldspar, debris, etc.
  • non-terrigenous detritus refers to the physical weathering or Detrital materials formed by mechanical damage, mainly including endogenous debris, volcanic debris, etc.
  • the interstitial component is divided into a cement component and a hetero-based component.
  • the hetero group indicates the fine mechanical admixture packed between the crumb particles.
  • the hetero group may be implemented as at least one of silt, clay material, finely divided, and carbonate stucco.
  • the cement indicates the site of chemical precipitation in the interstitial other than the hetero group.
  • the cement components are further divided into siliceous calcareous components and argillaceous components according to their constituent material components.
  • the terrigenous clastic components are divided into quartz feldspar and lithic components according to their clastic morphology. That is to say, for the rock type of clastic rock, the classification and identification model needs to have the characteristics of the quartz feldspar clastic region, the siliceous calcareous cement region, the non-terrestrial clastic region, the clastic clastic region, and the matrix interstitial material. The ability to identify areas and areas of argillaceous cement.
  • the rock slice of the clastic rock includes various components, please refer to FIG. 14 .
  • the rock slice 1400 of the clastic rock includes a quartz clastic region 1401, a siliceous calcareous cement region 1402, a non-terrigenous clastic region 1403, a lithic clastic region 1404, a matrix interstitial region 1405 and argillaceous cement Object area 1406.
  • the rock identification model 1500 includes a classification sub-model group and an identification sub-model group
  • the classification sub-model group includes a component classification sub-model 1511 and a clastic classification sub-model 1512, terrigenous debris classification sub-model 1513, interstitial material classification sub-model 1514 and cement classification sub-model 1515
  • the identification sub-model group includes non-terrestrial debris identification sub-model 1521, rock debris identification sub-model 1522, quartz Feldspar debris identification sub-model 1523 , siliceous calcareous cement identification sub-model 1524 , argillaceous cement identification sub-model 1525 and hetero-base interstitial material identification sub-model 1526 .
  • the component classification sub-model 1511 is connected with the debris classification sub-model 1512 and the interstitial material classification sub-model 1514, and the debris classification sub-model 1512 is respectively connected with the non-terrestrial debris identification sub-model 1521 and the terrigenous debris classification sub-model 1513 Connection, the terrigenous debris classification sub-model 1513 is connected with the rock debris identification sub-model 1522 and the quartz feldspar debris identification sub-model 1523 respectively, the interstitial material classification sub-model 1514 is respectively connected with the cement classification sub-model 1515 and the miscellaneous base fill The interstitial material identification sub-model 1526 is connected, and the cement classification sub-model 1515 is connected with the siliceous calcareous cement identification sub-model 1524 and the argillaceous cement identification sub-model 1525, respectively.
  • each classification sub-model when the rock slice segmentation image is input into the rock recognition model, each classification sub-model will generate the corresponding classification sub-image by dividing and recombining the rock slice segmentation image, so as to divide the rock slice image in the rock slice segmentation image. to determine the corresponding area.
  • Each identification sub-model corresponds to each sub-image to determine specific component types.
  • the quartz feldspar clastic area and the siliceous calcareous cement area belong to the single crystal component area;
  • the interstitial region belongs to the non-single crystal composition region.
  • the single-crystal component set image includes at least one single-crystal component region, and the single-crystal component identification result corresponding to the single-crystal component region, and the non-single-crystal component
  • the composition region includes at least one non-single-crystal composition region, and a non-single-crystal composition identification result corresponding to the non-single-crystal composition region.
  • classification and identification models in the embodiments of the present application are all neural network models based on deep learning.
  • Step 1302 Send a scan instruction to the mineral data acquisition device based on the single crystal component set image.
  • the scan data is spectral data collected by a mineral data collection device.
  • the mineral data acquisition device may also be implemented as, but not limited to, molecular spectroscopy such as infrared spectroscopy and Raman spectroscopy, X-ray fluorescence spectroscopy (X-ray Fluorescence, XRF), X-ray energy spectroscopy (Energy Dispersive X-ray Spectroscopy, EDS) at least one device.
  • the computer equipment sends a scan instruction to the mineral data acquisition equipment, and the scan instruction indicates to the mineral data acquisition equipment the position of the component area corresponding to the single crystal component set image.
  • the minerals corresponding to the single-crystal components are subject to alteration phenomenon (meaning that the minerals are affected by the outside world, the composition changes, and new minerals are generated.
  • process phenomena in one example, the erosion of feldspar into clay minerals), twinning phenomenon (referring to the regular growth of two or more crystals of the same kind, and the phenomenon in which the junction of crystals may interfere with data collection)
  • plagioclase includes polylamellar twinning phenomenon
  • inclusion phenomenon the phenomenon of a closed system composed of one phase or multiphase material in the mineral, and has a phase boundary with the host crystal mineral, in a
  • quartz has rutile inclusions
  • crack phenomenon referring to the phenomenon that minerals are fractured due to external stress, in one example, quartz has cracks), which affect the data acquisition results of the mineral data acquisition equipment, so, in the Before generating the scan command, first input the single crystal component set image into the area screening model, and obtain the area
  • Step 1303 Receive scan data fed back by the mineral data acquisition device based on the scan instruction.
  • the mineral data acquisition device After the mineral data acquisition device receives the scan command.
  • the mineral data acquisition device moves the objective lens based on the scanning instruction, or moves the stage, so that the area corresponding to the rock slice placed on the stage is opposite to the objective lens, and is irradiated by the light source with the photoelectric
  • the scan data may be implemented as spectral data.
  • Step 1304 verifying the identification result of the single crystal component based on the scanning data, to obtain a verification result.
  • the scan data can be determined from the chemical composition dimension, or the crystal structure dimension, for the type of components in each component region in the single crystal component set image, where each single crystal in the single crystal component set image.
  • Each of the crystalline component regions corresponds to scan data indicating its chemical composition, so the scan data can determine the single crystal identification result of each single crystal component in the single crystal component set.
  • Step 1305 based on the identification result of the non-single crystal component, the identification result of the single crystal component and the verification result, determine the mineral feature corresponding to the rock slice.
  • the verification result is the same as the identification result of the single crystal component, the verification result is determined to indicate that the verification is passed, and the identification result of the single crystal component is determined to remain unchanged;
  • the identification result of the single crystal component is determined based on the identification result generation rule.
  • the identification result generation rule is a determination rule generated based on the influence of the mineral data acquisition device on the identification result of the single crystal component. The reason for this rule is that the results of scanning data generated by the mineral data acquisition device are greatly influenced by the specific type of the mineral data acquisition device.
  • the mineral types corresponding to each component area can be summarized, and then the mineral characteristics corresponding to the rock slices can be determined.
  • the segmented image of the rock slice is preliminarily identified based on the crystal state through the classification and recognition model, and the After the area corresponding to the single crystal component area, the component identification result is verified twice, which improves the accuracy of the final generated mineral feature.
  • FIG. 16 shows a schematic flowchart of a method for determining a mineral feature corresponding to a component region provided by an exemplary embodiment of the present application.
  • the method can be implemented as step 505 in the embodiment corresponding to FIG. 5 instead. Taking the method applied to computer equipment as an example, the method includes:
  • Step 1601 Determine a component image based on the segmented area, where the component image includes a component area.
  • This process is the process of segmenting and determining the component area by segmenting the segmented area in the image according to the rock slice.
  • Step 1602 Send a spectral data acquisition instruction to the mineral data acquisition device based on the component image.
  • a spectral data acquisition instruction is sent to the mineral data acquisition device.
  • Step 1603 Receive spectral data sent by the mineral data acquisition device based on the spectral data acquisition instruction.
  • the mineral data acquisition device acquires the corresponding spectral data, and provides feedback to the computer device.
  • Step 1604 based on the spectral data, determine the primary mineral species corresponding to the component region in the mineral spectral data database.
  • the computer equipment stores a mineral spectral data database
  • the mineral spectral data database stores the corresponding relationship between the spectral data and the primary mineral species.
  • the primary mineral species are the species that are classified based on the representative properties of the minerals.
  • the computer device classifies the spectral data sample set in the spectral data database, and determines the rarity corresponding to the mineral according to the classified quantity of different minerals.
  • at least two subdivided mineral species are also included.
  • the process of obtaining the primary mineral species is as follows:
  • the primary category of the mineral species is a category based on the common degree of minerals. Includes common and uncommon mineral categories.
  • the method for determining the primary category of minerals is realized by comparing with the primary category frequency threshold.
  • the proportion of the spectral data corresponding to the mineral in the spectral data in the component area is greater than the first-class category frequency threshold, that is, when the frequency of occurrence of the spectral data corresponding to the mineral is greater than the first-class category frequency threshold, determine the mineral
  • the first-level category is a common mineral.
  • the mineral and the category are determined to be uncommon minerals.
  • common and uncommon mineral categories are determined based on the number of mineral samples in the spectral data database, that is, when the number of mineral samples in the spectral data database is less than the sample number threshold, the first-level mineral species is determined The category is the uncommon mineral category.
  • the first-level category of the mineral type is determined as the uncommon mineral category.
  • the primary mineral species corresponding to common mineral categories include mineral family species and common common mineral species, and the primary mineral species corresponding to uncommon mineral categories include inclusion mineral species, strongly sensitive mineral species and altered minerals type.
  • the common mineral species refers to the primary mineral species whose occurrence frequency is greater than the frequency threshold of a certain primary mineral species in the mineral spectral data database;
  • the inclusion mineral species refers to the minerals that appear as inclusions after the mineral grows.
  • the type of mineral that is selected is the most sensitive; the type of strongly sensitive minerals indicates that in the process of acquiring spectral data, due to chemical properties, the acquisition accuracy of the spectral data acquisition device will be affected, and the spectral data acquisition device will pay attention to the mineral species corresponding to its own minerals; Alteration mineral species refers to the mineral species corresponding to the minerals that produced the identification phenomenon in the mineral formation process.
  • first-class category frequency threshold and “mineral species frequency threshold” appearing in the above process may be the thresholds pre-stored in the computer equipment, or may be the received signals when the computer equipment performs the corresponding steps.
  • the present application does not limit the generation manner of each threshold.
  • the mineral spectral data database also has a plurality of sub-databases correspondingly.
  • the mineral spectral data database 1700 includes a common mineral category sub-database 1701 and an uncommon mineral category sub-database 1702, and the common mineral category sub-database 1701 also includes a common mineral category sub-database 1711 and a mineral group database 1712 , in the sub-database 1702 of uncommon mineral types, it also includes a database 1721 of inclusion mineral species, a database of strongly sensitive mineral species 1722 and a database of altered mineral species 1723.
  • the mineral spectral data database 1700 also includes an impurity category sub-database 1703 corresponding to the impurities.
  • the impurity type sub-database 1703 is an identification database corresponding to silicon carbide.
  • Step 1605 Determine the mineral type verification rule corresponding to the primary mineral type.
  • the mineral type verification rules include classification verification rules, direct verification rules, and point selection re-examination rules.
  • the verification rule is the point selection re-examination rule; in response to the primary mineral type being a common mineral type, or, the primary type corresponding to the primary mineral type is an uncommon mineral type, and the primary mineral type is not an inclusion mineral type , any one of strongly sensitive mineral species and altered mineral species, determine the mineral species verification rules as direct verification rules.
  • the point selection re-examination rule instructs the computer device to re-obtain the scan data from the spectral data acquisition device; the classification verification rule instructs the sub-database corresponding to the primary mineral type to be selected for secondary verification. ; Direct validation rule indicates the rule to use the result of the primary mineral species directly as the result of the verified mineral species.
  • the corresponding verification rule is the point selection retest rule, that is, due to the special properties of crystals in the component area , you need to re-select points in the component area.
  • the number of times of re-selection corresponds to the number of verification times pre-stored in the computer device.
  • the classification verification rule is a rule for further subdividing the mineral species based on the sub-database in the mineral spectral data database.
  • the primary selected mineral species is a mineral family species
  • the mineral species of the component region is subdivided based on the sub-database in the mineral family database, and a verification result is obtained. That is, the classification validation rules apply to minerals of the Mineral Group species.
  • the corresponding primary mineral type is a common mineral type, or, the first-level type corresponding to the mineral type is an uncommon mineral type, and the subdivided mineral type is not an inclusion mineral type, a strongly sensitive mineral
  • the direct verification rule is determined.
  • the direct verification rule is also the rule for determining the primary mineral category as the subdivided mineral category.
  • Step 1606 Determine the mineral species corresponding to the component region based on the mineral species verification rule and the primary mineral species.
  • This process is the process of determining the mineral species corresponding to the component area.
  • Step 1607 Determine the mineral feature corresponding to the rock slice based on the mineral species corresponding to the component image.
  • This process is the process of finally generating mineral features corresponding to rock slices based on the mineral species corresponding to each component area.
  • the corresponding database after acquiring the spectral data corresponding to the component area, the corresponding database performs preliminary determination of the mineral species in the component area, and correspondingly determines the verification rule. On this basis, the mineral species corresponding to the component area are finally determined.
  • the relevant verification rules determine the relevant verification rules, and further obtain its subdivided mineral species, which improves the The accuracy of the identification of mineral species further improves the efficiency and accuracy of the determination of mineral characteristics of rock slices.
  • the embodiment shown in FIG. 16 , the embodiment shown in FIG. 13 and the embodiment shown in FIG. 9 are used as the identification methods of three parallel mineral characteristics. In the process of rock identification, any The combination is superimposed to achieve the effect of identifying and verifying mineral characteristics, and further improving the accuracy of determining the mineral characteristics of rock slices. That is, the embodiment shown in FIG. 9 , the embodiment shown in FIG. 13 and the embodiment shown in FIG. 6 are three embodiments for obtaining mineral characteristics in the process of rock identification.
  • FIG. 18 shows a schematic diagram of a process for determining geometric features and mineral features provided by an exemplary embodiment of the present application. Please refer to FIG. 18 .
  • the method includes:
  • Step 1801 acquiring a sample slice image.
  • the rock slice image recognition model is a neural network model based on machine learning.
  • the neural network model can be obtained by combining the rock slice image segmentation model in the embodiment shown in FIG. 6 and the component type identification model in the embodiment shown in FIG. 9 , or the neural network model can be segmented according to the rock slice image
  • the model and the parameters of the component species identification model are constructed.
  • the sample corresponding to the training of the rock slice image recognition model is a sample slice image, and the sample slice image is marked with the sample segmentation result and the characteristics of the sample component type.
  • Step 1802 Input the sample slice image into the rock slice image segmentation and recognition model, and output the predicted segmentation area and predicted component type features.
  • This process is the process that the model predicts the sample thin-section image through the untrained rock thin-section image, and obtains the predicted segmentation area and the predicted component type characteristics.
  • Step 1803 Compare the predicted segmentation area with the sample segmentation area to obtain the sample segmentation difference, and compare the predicted component type feature with the sample component type feature to obtain the sample component difference.
  • the sample segmentation difference and the sample component difference are determined at the same time.
  • Step 1804 Adjust the rock slice image segmentation recognition model based on the sample segmentation differences and the sample component differences.
  • This process is the process of adjusting the rock slice image segmentation recognition model.
  • Step 1805 Input the rock thin section image into the rock thin section image segmentation recognition model, and output at least two divided regions corresponding to the rock thin section image and component type features corresponding to the divided regions.
  • This process is the process of inputting the rock slice image to be measured into the rock slice image segmentation and recognition model, and outputting the segmented area and the component type features corresponding to the area.
  • Step 1806 combining the segmented region and component type features, determine the geometric features and mineral features corresponding to the rock slices.
  • the segmented area corresponds to the geometric feature
  • the component type feature corresponds to the mineral feature
  • the method provided in the embodiment of the present application corresponds to the sample slice image as the input of the rock slice image segmentation and recognition model, and on the premise that the model training is completed, the geometric and mineral characteristics of the rock slice image are analyzed. Simultaneous extraction to further improve the efficiency of feature extraction during rock identification.
  • the embodiment shown in FIG. 18 is an embodiment in which a computer device simultaneously obtains geometric features and mineral features.
  • Fig. 19 shows a schematic diagram of a method for generating identification results of rock slices provided by an exemplary embodiment of the present application.
  • the method can be implemented as step 303 as shown in Fig. 3 instead, and the method is applied to computer equipment
  • the method includes:
  • Step 1901 Determine the type of at least one rock slice sub-identification result corresponding to the identification result based on the primary rock type feature and the subdivided rock type characteristic.
  • the sub-identification result of the rock slice is obtained based on the shape of the component area and the mineral properties of the component area, and has a characterizing function for the specific type of the rock slice.
  • the rock slice identification results include at least one of particle size interval identification results, maximum particle size identification results, particle sorting results, particle roundness identification results, mineral self-shape degree results, and particle contact mode identification results. kind.
  • Step 1902 based on the geometric features and mineral features, generate a rock slice sub-identification result.
  • the determination method corresponding to the above-mentioned particle size interval identification result, maximum particle size identification result, particle sorting result, particle roundness identification result, mineral self-shape degree result and particle contact method identification result is given as an example.
  • the identification result of the rock slice includes a particle size interval identification result
  • the particle size interval identification result indicates a particle size interval of a component region corresponding to a type of component in the rock slice. Then, first, determine the particle size interval component area set corresponding to the particle size interval identification result based on the mineral characteristics, and the particle size interval component area set includes at least two component areas, and secondly, based on the particle size interval component area set. Determine the particle size corresponding to the component area in the component area of the particle size interval based on the geometric features corresponding to the component area of the particle size interval, and finally, determine the particle size interval identification result based on the particle size corresponding to the component area.
  • the identification result of the rock slice includes the identification result of the largest particle size
  • the identification result of the largest particle size indicates the particle size of the component area with the largest particle size in the component area corresponding to a type of component in the rock slice .
  • the largest particle component area corresponding to the identification result of the largest particle size is determined based on the mineral feature
  • the largest particle size corresponding to the component area is determined based on the geometric feature
  • the largest particle size is generated based on the largest particle size Path recognition results.
  • the identification result of the rock slice includes a particle sorting result
  • the particle sorting result indicates the distribution of particle sizes corresponding to a type of particle in the rock slice.
  • a set of sortable component regions corresponding to the identification result of particle sortability is determined based on mineral characteristics, and the set of sortable component regions includes at least two component regions; secondly, based on the sortability
  • the geometric features corresponding to the component regions in the component region set determine the corresponding relationship of particle size classification. Quantity; finally, based on the particle size classification correspondence, the particle sorting results are determined.
  • the identification result of the rock slice includes the identification result of particle roundness
  • the particle roundness identification result indicates the roundness of the component region
  • the roundness corresponds to a roundness level.
  • a roundness identification component region set corresponding to the particle roundness identification result is determined based on the mineral characteristics, and the roundness identification component region set includes at least two component regions; secondly, based on the grinding
  • the roundness component identifies the geometric features corresponding to the component regions in the region set, and determines the roundness grading correspondence.
  • the roundness grading correspondence includes at least two roundness levels, and groups in the roundness levels. The number of sub-regions; finally, the particle roundness identification result is determined based on the roundness classification correspondence.
  • the rock section identification results include a mineral euhedral degree result, which indicates the degree to which the mineral in the component region has developed based on its own crystal habit.
  • the euhedral degree component region set corresponding to the mineral euhedral degree result is determined based on the mineral characteristics, and the euhedral degree component region set includes at least two euhedral degree component regions;
  • the geometric features corresponding to the component regions in the morphological component region set are used to determine the shape of the component region; finally, based on the shape of the component region, the result of the mineral eumorphism is determined.
  • the identification results of the particle contact patterns are included in the identification results of the rock slices.
  • the identification result of particle contact mode indicates the contact state of two adjacent component regions. In this case, firstly, at least two groups of component area pairs are determined based on mineral characteristics, and the component area pairs include two adjacent component areas; secondly, based on geometric characteristics, two groups of component area pairs are determined The intersection ratio of the sub-regions; finally, based on the intersection ratio of at least two component regions, the identification result of the particle contact mode is determined.
  • the component area corresponding to the rock slice sub-identification result in the rock slice image is determined based on the mineral type, and based on geometric features, from the group
  • the shape, shape and size of the sub-regions determine the results of the sub-identification of the rock slices.
  • Step 1903 Determine a naming rule corresponding to the name of the rock slice based on the characteristics of the subdivided rock types.
  • different rock types correspond to different naming rules, so it is necessary to determine the names of rock slices based on the characteristics of subdivided rock types.
  • Step 1904 based on the naming rule and the rock slice sub-identification result, generate a rock slice name.
  • the computer device may generate corresponding keywords according to the sub-evaluation results of the rock slices, and finally determine the names corresponding to the rock slices in combination with the characteristics of subdivided rock types.
  • a rock sample identification report is also generated.
  • the name of the rock slice and the identification results of each rock slice are included.
  • the computer device pre-stores at least two rock sample identification reports according to the characteristics of different subdivision rock types.
  • the selection of the rock slice identification sub-results corresponding to the rock type is performed, and the rock slice identification obtained by the selection is correspondingly performed.
  • the content of the finally generated rock slice identification result corresponds to the type of the rock slice, and the accuracy of the rock slice identification result is further improved.
  • the embodiment corresponding to FIG. 19 illustrates the process of obtaining rock thin section identification sub-results through geometric features, mineral characteristics and structural features, and then obtaining the final rock thin section identification result.
  • FIG. 20 shows a schematic process diagram of a rock identification method provided by an exemplary embodiment of the present application, and the process includes:
  • step 2001 a thin section image of clastic rock is collected.
  • the image of the clastic rock slice sample includes a single polarized light image and a crossed polarized light image.
  • images of the clastic rock thin section sample are continuously collected under the image acquisition device, and then spliced to finally obtain a single polarized light image and an orthogonal polarized light image corresponding to the clastic rock thin section sample.
  • Step 2002 Determine the primary rock type of the clastic rock slice image.
  • the rock classification model determines that the primary rock type of the clastic rock slice image is sedimentary rock.
  • Step 2003 determining the subdivision rock types of the clastic rock slices.
  • the subdivided rock type of the clastic rock slice is clastic rock and has a sand-like structure.
  • Step 2004 Input the clastic rock slice image into the clastic rock segmentation model to obtain a segmented clastic rock slice image.
  • This process is the process of segmenting the clastic rock slice image through the rock slice segmentation model to obtain the segmented image.
  • the computer equipment can determine the geometrical features of the clastic rock corresponding to the thin slice image of the clastic rock.
  • Step 2005 extracting the component region according to the segmented clastic rock slice image, and obtaining the component type feature through the component type identification model.
  • This process is the process of obtaining the mineral features corresponding to the clastic rock slice images.
  • Step 2006 Map the component type feature on the clastic rock thin section image to obtain an intermediate image of the clastic rock thin section sample.
  • This process is to label the clastic rock slice image based on the component type features and segmentation images, and obtain an intermediate image that can characterize the geometric and mineral characteristics of the clastic rock slice.
  • the component species identification model will directly output the intermediate image of the clastic rock slice sample.
  • Step 2007 on the intermediate image of the clastic rock slice sample, count the area of the corresponding component area.
  • the relative content of different components is determined by means of counting the area of the component area.
  • the proportion of siliceous cements is 2.0%
  • the proportion of calcareous cements is 10.5%
  • the proportion of interstitials is 1%
  • the proportion of cuttings is 13.84%.
  • feldspar debris accounted for 20.76%
  • quartz debris accounted for 51.9%.
  • Step 2008 Calculate the sub-region images corresponding to the quartz debris region, the feldspar debris region, the siliceous cement region, and the calcareous cement region, respectively, and perform clean region screening.
  • the components corresponding to the quartz clastic region, the feldspar clastic region, the siliceous cement region and the calcareous cement region are characterized by a single crystal state, it is necessary to screen clean regions that can be scanned.
  • this process is implemented through a machine learning based model.
  • Step 2009 Determine the plane scan coordinate map.
  • This process is the process of determining the position of the specific scanning area.
  • the scanning method is surface scanning.
  • step 2010 the face scan coordinates are calculated, an optimal path is planned, and a scan instruction is generated.
  • this process is a process in which the computer device determines the scan instruction. Since the areas that need to be scanned are not adjacent in actual situations, an optimal path needs to be planned.
  • Step 2011 controlling the mineral data acquisition device to collect scan data through the scan instruction.
  • This process is a process of sending the scan data to the mineral data acquisition device, so that the mineral data acquisition device generates corresponding scan data.
  • the scan data is spectral data.
  • Step 2012 generating mineral information based on the scan data.
  • This process is a process of generating, based on the scan data, the scan data sent by the mineral data acquisition device, and determining the mineral composition of the component area represented by the scan data.
  • Step 2013, compare the mineral information with the classification and identification results, and obtain the final result on each component area.
  • This process is the process of comparing the mineral composition with the scan data to obtain the final result.
  • the final verification result is determined according to the equipment type of the mineral data acquisition equipment.
  • Step 2014 based on the final result, replace the identification result of the corresponding component area on the intermediate image of the clastic rock thin section sample.
  • This process is the process of replacing the results of verification by surface scanning with the mineral identification results of the component regions obtained by the machine learning model on the intermediate images of the clastic rock thin-section samples. Since the spectral data is more accurate for the identification of mineral components, in this step, the verification results of the surface scan are used as the criterion to determine the mineral characteristics of the component area.
  • Step 2015 determining the final image of the updated clastic rock thin section sample.
  • the proportion of siliceous cement is 2.0%
  • the actual form corresponding to calcareous cement is calcite, accounting for 10.5%
  • the proportion of hetero-based interstitial material is 1%
  • Rock debris accounted for 13.84%
  • quartz debris accounted for 69.2%
  • feldspar debris accounted for 3.46%
  • the proportion of terrigenous debris was still 86.5%
  • the proportion of non-terrestrial debris was still 0% .
  • the proportion of particles is 86.5%
  • the proportion of interstitials is 13.5%
  • Step 2016 Determine the thin section identification result corresponding to the clastic rock thin section.
  • the flake sub-identification results include particle size interval identification results, particle sorting results, maximum particle size identification results, relative component content results, particle roundness identification results, and particle contact mode identification results.
  • the main particle size range of the particles is that fine sand (0.125-0.25mm) accounts for 10%, and medium sand (0.25 mm) accounts for 10%.
  • Step 2017 Determine the sheet name and sheet identification report based on the sheet sub-authentication result.
  • a clastic rock naming area 2110 is included, and the clastic rock naming area includes the clastic rock slice naming based on the slice sub-identification results.
  • the clastic rock is named "medium-coarse-grained lithic quartz sandstone".
  • the naming is based on the industry standard stored in the computer equipment, that is, the industry standard SY/T 5368-2016 "Rock Thin Section Identification”.
  • the thin section identification report 2100 also includes a thin section sample information area 2120 indicating where and how the clastic rock sample was obtained, a component feature area 2130 indicating the mineral composition of the clastic rock sample, and a section indicating the structural characteristics of the clastic rock sample. Structural feature area 2140 and rock slice image display area 2150.
  • the method provided in the embodiment of the present application corresponds to the case where the rock slice sample is clastic rock. And through the acquisition of spectral data, the recognition results obtained by image recognition are verified, and the method of finally generating the identification results is to combine the chemical features of the component regions represented in the spectral data and the image features of the component regions represented in the image.
  • the identification of clastic rock slices improves the efficiency and accuracy of the identification of clastic rock slices.
  • Fig. 20 shows the clastic rock type, through the determination of geometric characteristics, mineral characteristics and structural characteristics, and based on the three types of characteristics, to determine the sub-results of clastic rock slice identification, and finally get Process for the identification of clastic rock slices.
  • FIG. 22 shows a schematic process diagram of a rock identification method provided by an exemplary embodiment of the present application, and the process includes:
  • Step 2201 acquiring a sample image of a magmatic rock slice.
  • FIG. 23 shows a schematic diagram of a magmatic rock slice image provided by an example of the present application.
  • the magmatic rock slice 2300 includes at least one olivine component region 2301, common pyroxene component region 2302, basic plagioclase component region 2303, perilla pyroxene component region 2304 and magnetite component region 2305 .
  • the magmatic rock thin section image also includes a single-polarized light image and an orthogonal polarized light image corresponding to the magmatic rock thin section sample.
  • Step 2202 Determine the primary rock type corresponding to the magmatic rock slice image.
  • magmatic rock slice image it is determined to be a magmatic rock, and accordingly, the geometric features and mineral features are the features corresponding to the magmatic rock.
  • the subdivision structure of the magmatic rock structure may be further determined as a gabbro structure by means of model recognition. That is, after the primary rock type is determined, there is a case where the subdivided rock type is directly further determined. According to this, the geometrical and mineral features are the features corresponding to the magmatic rocks with the gabbro structure.
  • Step 2203 Input the magmatic rock slice sample image into the magmatic rock segmentation model to obtain a segmented magmatic rock slice image.
  • This process is the process of segmenting magmatic rock slice images to obtain segmentation results.
  • Step 2204 Extract component regions according to the segmented magmatic rock slice image, and obtain component type features through a component type identification model.
  • This process is the process of extracting the component area based on the segmentation result, and further identifying the mineral composition corresponding to the component area.
  • Step 2205 Perform clean area screening on the corresponding component areas in the segmented magmatic rock slice image to determine the surface scan area.
  • step 2205 is to pass the clean area screening to determine the surface scan area. process.
  • Step 2206 Calculate the coordinates of the face scan, plan an optimal path, and generate a scan command.
  • This process is the process of planning the optimal path to generate the scanning command sent to the mineral data acquisition device when the scanning area of the corresponding surface is not adjacent.
  • Step 2207 Control the mineral data acquisition device to collect scan data through the scan instruction.
  • This process is the process of scanning data acquisition by computer equipment controlling mineral data acquisition equipment.
  • Step 2208 generating mineral information based on the scan data.
  • This process is the process of receiving scanning data, that is, spectral data, from the mineral data acquisition equipment, and then determining the mineral species corresponding to the component area according to the spectral data.
  • Step 2209 compare the mineral information with the classification and identification results, and obtain the final result on each component.
  • This process is the process of finally determining the mineral species corresponding to each component area through verification.
  • Step 2210 determining the type of minerals contained in the magmatic rock slice.
  • the types of minerals contained in the magmatic rock slices can be determined.
  • the magmatic flakes contain mineral species including plagioclase, common pyroxene, perilla pyroxene, olivine, biotite, magnetite, and apatite.
  • Step 2211 Determine the subdivision rock type corresponding to the magmatic rock slice based on the contained mineral type.
  • the subdivided rock species corresponding to the magmatic rock slice is plagioclase, and the subgroup is labradorite, which belongs to basic plagioclase.
  • the number of subdivision structural features of the magmatic rock is at least two.
  • the subdivision structural feature corresponding to the magmatic rock further includes a reactive edge structural feature, which is used to indicate the type of reactive edge contained in the magmatic rock.
  • Step 2212 based on the rock subdivision type, determine the thin section identification result corresponding to the magmatic rock thin section.
  • the flake sub-identification results include particle size identification results, particle size shape results, mineral eumorphism results, particle interaction relationship results, and mineral content results.
  • the plagioclase content is 52.4%
  • the ordinary pyroxene content is 34.9%
  • the perilla pyroxene content is 3.2%
  • the olivine content is 7.6%
  • the mica content is 1.6%
  • the magnetite content is 0.4%
  • the apatite content is 0.1%.
  • the computer equipment can determine that the olivine has a perilla pyroxene border and a reaction edge structure, thereby indicating the magmatic rock. It is a basic intrusive rock or an ultrabasic intrusive rock.
  • the particle size of plagioclase is (1.6-2.1) x (1.8-3.5) mm
  • the particle size of ordinary pyroxene is 1.2-2.2 mm
  • the particle size of perilla pyroxene is 1.2-2.2 mm.
  • the particle size of olivine is 1.1-1.7mm
  • the particle size of olivine is 0.4-2.5mm
  • the particle size of biotite is 1.2-2.2mm
  • the particle size of magnetite is 0.05-0.2mm
  • the particle size of apatite is 0.05-0.1mm.
  • plagioclase is granular-short plate
  • common pyroxene is irregular granular and short column
  • perilla pyroxene is granular
  • olivine It is granular
  • biotite is plate
  • magnetite is granular
  • apatite is short column.
  • the plagioclase is hemimorphic-heteromorphic
  • the common pyroxene is hemimorphic-semi-hrigoid
  • the perilla pyroxene is hexomorphic
  • the olivine is hetmorphic
  • the biotite is hemimorphic. It is euhedral-semihedral, magnetite is euhedral, and apatite is semihedral-hedgomorphic.
  • the determination and generation process of the identification results of thin slices can be regarded as the verification process of structural characteristics.
  • the computer equipment can determine that there is a reaction edge structure in the magmatic rock slice, and the corresponding magmatic rock slice has a reaction edge structure.
  • the flake identifier sub-results include reactive edge structures that indicate the presence of reactive edges.
  • Step 2213 Determine the sheet name and sheet identification report based on the sheet sub-authentication result.
  • Element 1 In the magmatic rock slice image, the component area corresponding to the plagioclase indicates that the particle size of the plagioclase belongs to mesograin.
  • Element 2 In the magmatic rock slice image, the mineral features indicate that the main minerals are plagioclase and pyroxene.
  • Element 3 In the case where the main minerals are plagioclase and pyroxene, based on geometric characteristics, it is determined that both plagioclase and pyroxene are granular.
  • Element 4 In the case where the main minerals are plagioclase and pyroxene, based on geometric characteristics, it is determined that plagioclase and pyroxene have similar euhedral degrees and are both semihedral-heteromorphic.
  • the naming of the magmatic rock slice can be verified by the method in the following example. It should be noted that the following verification process is performed under the condition that the relevant national standard data and industry standard data are stored in the computer equipment.
  • Example method Based on the combination of mineral features and geometric features, the content is indicated by the area of the component area. At the same time, Q is defined as the sum of the content of quartz, tridymite and cristobalite, and A is alkaline feldspar (that is, positive feldspar).
  • the classification area corresponding to the magmatic rock slice can be determined as classification area 2401, and the classification area 2401 indicates that the rock name corresponding to the magmatic rock slice may include diorite, gabbro At least one of rock and plagioclase. In this case, it can be verified that the type of magmatic rock slice is gabbro.
  • the verification method also includes methods such as verifying by subdividing the glow-length structure indicated by the structural feature, which is not repeated in this embodiment of the present application.
  • olivine gabbro since the mineral characteristics indicate that the olivine content in the magmatic rock slice is 7.6%, which is greater than 5%, it is necessary to reflect the olivine characteristics when naming, and correspondingly, it can be verified to finally determine the magmatic rock.
  • the flakes are named "olivine gabbro".
  • the embodiment of the present application also provides a thin section identification report corresponding to the magmatic rock thin section.
  • the method provided in the embodiment of the present application corresponds to the case where the rock slice sample is magmatic rock.
  • the acquisition of spectral data verifies the recognition results obtained by image recognition, and finally generates the identification results.
  • the chemical characteristics of the component regions represented in the spectral data and the image features of the component regions represented in the image are combined to determine the magmatic rock. It can further identify the structural features and composition characteristics of magmatic rock thin sections, which improves the efficiency and accuracy of the identification of clastic rock thin sections.
  • FIG. 22 is an example of a magmatic rock identification method, the process of obtaining the magmatic rock slice identification sub-results through geometric features, mineral features and structural features, and then obtaining the final identification results.
  • Fig. 25 shows a structural block diagram of a rock identification device provided by an exemplary embodiment of the present application, and the device includes:
  • the receiving module 2501 is used to receive a rock slice image sent by an image acquisition device, the rock slice image is an image obtained by photographing a rock slice, the rock slice is a slice obtained by cutting a rock sample, and the rock slice image includes at least one component area;
  • the generating module 2502 is configured to generate geometric features, mineral features, and structural features corresponding to the rock thin sections based on the rock thin section images, wherein the geometric features are used to indicate the composition-based regional division of the rock thin sections, and the mineral features are used to indicate the rock thin section.
  • the mineral distribution of the rock samples, and the structural characteristics are used to indicate the spatial structure of the rock samples;
  • the identification results of rock slices are generated, and the identification results include textual descriptions of the characteristics of the rock samples.
  • the structural feature includes a primary rock structure feature
  • the primary rock structure feature is used to indicate a primary rock type corresponding to the rock sample, and the primary rock type includes sedimentary rocks, magmatic rocks, and metamorphic rocks rocks;
  • the generating module 2502 is further configured to generate a primary rock structure feature based on the rock slice image
  • the apparatus further includes a determination module 2503 for determining the primary rock species corresponding to the rock slices based on the structural features of the primary rock species;
  • the generating module 2502 is further configured to generate geometric features based on the rock slice images according to the primary rock species;
  • mineral features are generated based on rock slice images.
  • the structural feature further includes a subdivided rock structure feature, and the subdivided rock structure feature is used to indicate the subdivided rock species corresponding to the rock sample;
  • the determining module 2503 is further configured to determine the structural features of subdivided rocks based on the geometric features and mineral features.
  • the rock slice image includes at least two component images representing component regions
  • the apparatus further includes a division module 2504, configured to divide the rock slice image based on the position of the component image to obtain a rock slice segmented image.
  • the rock slice segmented image includes at least two segmented areas, and the segmented area is used to indicate the rock slice image. Segmentation based on component regions;
  • the determining module 2503 is further configured to determine geometric features corresponding to the rock slices based on the segmented area;
  • the mineral features corresponding to the rock slices are determined based on the segmented regions.
  • the apparatus further includes an input module 2505 for inputting the rock slice image into the rock slice image segmentation model, and outputting the rock slice image segmentation model.
  • the rock slice image segmentation model is a mask based on machine learning. -RCNN network model.
  • the determining module 2503 is further configured to determine a component image based on the segmented area, and the component image includes a component area;
  • the input module 2505 is also used to input the component image into the component type identification model, and output the component type feature of the component image.
  • the component type feature is used to characterize the mineral characteristics of the component area and identify the component type
  • the model is a neural network model based on machine learning;
  • the determining module 2503 is further configured to determine the mineral feature corresponding to the rock slice based on the component type feature corresponding to the component image.
  • the rock identification system further includes a mineral data acquisition device, and the mineral data acquisition device is connected to the computer device;
  • the input module 2505 is further configured to input the segmented image of the rock slice into the classification and recognition model, and output a single crystal component set image and a non-single crystal component set image, where the single crystal component set image includes at least one single crystal The state composition region, and the identification result of the single crystal state composition corresponding to the single crystal state composition region, the non-single crystal state composition region includes at least one non-single crystal state composition region, and the The identification result of the non-single crystal component corresponding to the region, and the classification and identification model is a neural network model based on deep learning;
  • the apparatus further includes a sending module 2506, configured to send a scanning instruction to the mineral data acquisition device based on the single crystal component set image;
  • the receiving module 2501 is further configured to receive scan data fed back by the mineral data acquisition device based on the scan instruction;
  • the device further includes a verification module 2507 for verifying the identification result of the single crystal component based on the scanning data to obtain the verification result;
  • the determining module 2503 is further configured to determine the mineral feature corresponding to the rock slice based on the identification result of the non-single crystal component, the identification result of the single crystal component and the verification result.
  • the input module 2505 is further configured to input the single crystal component set image into the regional screening model, and output the regional screening result, and the regional screening result is used to indicate the single crystal component set image used in the region screening model. for the part where the surface scan data is performed;
  • the generating module 2502 is further configured to generate a scan instruction based on the area screening result
  • the sending module 2506 is further configured to send a scanning instruction to the mineral data acquisition device.
  • the determining module 2503 is further configured to determine the scan identification result based on the scan data
  • the identification result generation rule includes based on the type of mineral collection equipment. the rule of.
  • the determining module 2503 is further configured to determine a component image based on the segmented area, and the component image includes a component area;
  • the sending module 2506 is further configured to send the spectral data acquisition instruction to the mineral data acquisition device based on the component image;
  • the receiving module 2501 is further configured to receive the spectral data sent by the mineral data acquisition device based on the spectral data acquisition instruction;
  • the determining module 2503 is further configured to determine the primary mineral species corresponding to the component region in the mineral spectral data database based on the spectral data;
  • the mineral features corresponding to the rock slices are determined.
  • the mineral spectral data database includes a mineral species sub-database
  • the determining module 2503 is further configured to determine the primary category of the primary mineral species corresponding to the component region in the mineral spectrum data database based on the spectral data.
  • the primary category of the mineral species is a category based on the common degree of the minerals.
  • the primary category includes common mineral categories and uncommon mineral categories.
  • the primary mineral categories corresponding to common mineral categories include mineral family categories and common common mineral categories.
  • the primary mineral categories corresponding to uncommon mineral categories include inclusion minerals, strong Sensitive mineral species and altered mineral species;
  • the primary mineral species corresponding to the component area is determined.
  • the mineral species verification rules include classification verification rules, direct verification rules, and point selection re-examination rules
  • the determining module 2503 is further configured to determine that the mineral species verification rule is a classification verification rule in response to the primary mineral species being a mineral family species;
  • the mineral species verification rule is a point selection retest rule
  • the mineral type verification rule is a direct verification rule.
  • the rock slice image includes at least two component images representing component regions
  • the input module 2505 is further configured to input the rock slice image into the rock slice image segmentation recognition model, and output to obtain at least two segmented regions corresponding to the rock slice image and component type features corresponding to the segmented regions, and the segmented regions are used to indicate the rock slices
  • the image is based on the segmentation of the component area
  • the component type feature is used to characterize the mineral characteristics of the component area in the segmented area
  • the rock slice image segmentation recognition model is a neural network model based on machine learning
  • the determining module 2503 is further configured to combine the segmented area and the component type features to determine the geometrical features and mineral features corresponding to the rock slices.
  • the input module 2505 is further configured to input the rock slice image into the primary rock type selection model, and output the primary rock type result, where the primary rock type result indicates the primary rock type corresponding to the rock slice Class structure features, the primary rock structure selection model is a model constructed based on the primary rock image sample set.
  • the identification result includes subdivision rock types
  • the input module 2505 is further configured to input geometric features and mineral characteristics into the subdivision rock type selection model, and the output obtains a subdivision rock type result.
  • the rock type results indicate the subdivision rock structure features corresponding to the rock slices.
  • the subdivision rock type selection model is a model constructed based on the geometric-mineral feature interaction sample set.
  • the geometry-mineral feature interaction sample set indicates the combination of geometric features and mineral features. Correspondence between subdivision rock structure features.
  • the identification result includes at least one of a rock slice sub-identification result and a rock slice name
  • the determination module 2503 is further configured to determine the type of at least one rock slice sub-identification result corresponding to the identification result based on the primary rock type feature and the subdivided rock type characteristic.
  • the rock slice identification result includes the particle size interval identification result, the largest particle size at least one of diameter identification results, particle sorting results, particle roundness identification results, mineral self-shape degree results, and particle contact mode identification results;
  • the generating module 2502 is further configured to generate a rock slice sub-identification result based on the geometric feature and the mineral feature;
  • the determining module 2503 is further configured to determine the rock slice name based on the rock slice identification result.
  • the device provided in this embodiment of the present application performs three-dimensional feature extraction based on geometric features, mineral features and structural features for the rock thin section images after acquiring the rock thin section images. From the perspective of microscopic composition and macroscopic performance , synthesizing multiple feature dimensions to determine the properties of rocks, and finally generating identification results including textual descriptions.
  • the process of rock identification after obtaining the microscopic visualization image corresponding to the rock slice, the image is extracted with multiple dimensions of features, and the rock slice is identified with reference to the features of multiple dimensions, which improves the performance of the rock slice. Accuracy of rock identification.
  • the rock identification device provided in the above embodiment is only illustrated by the division of the above functional modules. In practical applications, the above functions can be allocated to different functional modules according to needs. Divided into different functional modules to complete all or part of the functions described above.
  • FIG. 27 shows a schematic structural diagram of a computer device for performing a rock identification method provided by an exemplary embodiment of the present application, and the computer device includes:
  • the processor 2701 includes one or more processing cores, and the processor 2701 executes various functional applications and data processing by running software programs and modules.
  • the receiver 2702 and the transmitter 2703 may be implemented as a communication component, which may be a communication chip.
  • the communication component may implement a signal transmission function. That is, the transmitter 2703 can be used to transmit control signals to the image capturing device and the scanning device, and the receiver 2702 can be used to receive corresponding feedback instructions.
  • the memory 2704 is connected to the processor 2701 through the bus 2705.
  • the memory 2704 may be configured to store at least one instruction, and the processor 2701 may be configured to execute the at least one instruction to implement various steps in the above method embodiments.
  • Embodiments of the present application further provide a computer-readable storage medium, where at least one instruction, at least one piece of program, code set or instruction set is stored in the readable storage medium to be loaded and executed by a processor to implement the above rock identification method.
  • the application also provides a computer program product or computer program, the computer program product or computer program comprising computer instructions stored in a computer-readable storage medium.
  • the processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the rock identification method described in any of the above embodiments.
  • the computer-readable storage medium may include: Read Only Memory (ROM, Read Only Memory), Random Access Memory (RAM, Random Access Memory), Solid State Drive (SSD, Solid State Drives), or an optical disc.
  • the random access memory may include a resistive random access memory (ReRAM, Resistance Random Access Memory) and a dynamic random access memory (DRAM, Dynamic Random Access Memory).
  • ReRAM resistive random access memory
  • DRAM Dynamic Random Access Memory

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Multimedia (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Software Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Quality & Reliability (AREA)
  • Probability & Statistics with Applications (AREA)
  • Image Analysis (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)
  • Investigating, Analyzing Materials By Fluorescence Or Luminescence (AREA)
  • Analysing Materials By The Use Of Radiation (AREA)

Abstract

一种岩石鉴定方法、系统、装置、终端及可读存储介质，涉及岩石鉴定技术领域，该方法包括：接收图像采集设备发送的岩石薄片图像（301）；基于岩石薄片图像生成与岩石薄片对应的几何特征、矿物特征以及结构特征（302）；基于几何特征、矿物特征以及结构特征，生成岩石薄片的鉴定结果（303）。在获取岩石薄片图像后，对于岩石薄片图像进行基于几何特征、矿物特征以及结构特征的三个维度的特征提取，从多个特征维度对于岩石的性质进行确定，最终生成包括有文字化描述的鉴定结果。在获取岩石薄片对应的微观可视化图像后，对该图像进行多个维度的特征提取，并在参考多个维度的特征的情况下对于岩石薄片进行鉴定，提高了岩石鉴定的准确率。

Description

岩石鉴定方法、系统、装置、终端及可读存储介质
相关申请的交叉引用
本申请要求享有2021年02月08日提交的名称为“一种岩石薄片智能鉴定装置及方法”的中国专利申请CN 202110181773.4和2021年02月10日提交的名称为“一种碳酸盐岩薄片图像中生物的识别方法”的中国专利申请CN 202110182027.7的优先权，其全部内容通过引用并入本文中。
技术领域
本申请涉及岩石鉴定技术领域，特别涉及一种岩石鉴定方法、装置、终端及可读存储介质。
背景技术
应用光学原理，将岩石切割磨制成30微米厚的薄片，并用偏光显微镜对于岩石薄片进行观察以及研究，是地质行业科研生产中最基本、最简便、最廉价的方法之一，也是地质学家必须掌握的基本知识和基本技能之一。
传统岩石薄片鉴定方法沿用至今，在传统岩石薄片鉴定方法中，研究者通常通过人工识别的方式，通过对于岩石薄片的显微观察，进而对岩石薄片进行鉴定。同时，研究人员也会通过利用设备和计算机采集显微图像的方式，对于显微图像进行更为细致的观察。
然而，相关技术中，无论是对于岩石薄片进行直接观察，或是将岩石薄片图像输入计算机中进行观察的方式，均受到较大的主观因素的影响，存在无法避免的误差，致使岩石鉴定的准确率较低。
发明内容
本申请关于一种岩石鉴定方法、系统、装置、终端及可读存储介质，能够提高岩石鉴定的准确率。该技术方案如下：
一方面，提供了一种岩石鉴定方法，该方法应用于岩石鉴定系统的计算机设备中，所述岩石鉴定系统包括图像采集设备以及计算机设备；图像采集设备与计算机设备连接，该方法包括：
接收图像采集设备发送的岩石薄片图像，岩石薄片图像为对岩石薄片进行拍摄得到的图像，岩石薄片为对岩石样本进行切割得到的薄片，岩石薄片图像中包括至少一个组分区域；
基于岩石薄片图像生成与岩石薄片对应的几何特征、矿物特征以及结构特征，其中，几何特征用于指示岩石薄片的组分区域划分情况，矿物特征用于指示对岩石薄片中与组分区域对应的矿物种类分布情况，得到的矿物鉴定结果，结构特征用于指示岩石样本的岩石种类；
基于几何特征、矿物特征以及结构特征，生成岩石薄片的鉴定结果。
在一个可选的实施例中，该方法，还包括：
获取样本薄片图像，样本薄片图像标注有样本分割结果；
将样本薄片图像输入岩石薄片图像分割模型，输出得到预测分割结果；
将预测分割结果与样本分割结果进行差异比较，得到分割差异；
基于分割差异对岩石薄片图像分割模型进行调整。
在一个可选的实施例中，将预测分割结果与样本分割结果进行差异比较之后，还包括：
响应于分割差异指示预测分割结果与样本分割结果的差异阈值时，确定岩石薄片图像分割模型的训练完成；
或，
响应于岩石薄片图像分割模型在一个分割训练周期中的样本薄片图像输入次数达到分割次数阈值时，确定岩石薄片图像分割模型的训练完成。
在一个可选的实施例中，该方法，还包括：
获取样本组分图像，样本组分图像标注有样本组分类型特征；
将样本组分图像输入组分种类识别模型，输出得到预测组分类型特征；
将预测组分类型特征与样本组分类型特征进行差异比较，得到识别差异；
基于识别差异对组分种类识别模型进行调整。
在一个可选的实施例中，非单晶态组分集合图像包括非晶态组分集合图像和多晶态组分集合图像，非单晶态组分区域包括非晶态组分区域和多晶态组分区域，非晶态组分区域对应有非晶态组分识别，多晶态组分区域对应有多晶态组分识别结果。
在一个可选的实施例中，岩石薄片为碎屑岩岩石薄片；
单晶态组分区域包括石英长石碎屑区域与硅质钙质胶结物区域，非单晶态组分区域包括非陆源碎屑区域、岩屑碎屑区域、杂基填隙物区域和泥质胶结物区域；
分类识别模型中包括分类子模型组和识别子模型组；
分类子模型组中包括颗粒分类子模型、碎屑分类子模型、陆源碎屑分类子模型、填隙物分类子模型和胶结物分类子模型；
识别子模型组中包括非陆源碎屑识别子模型、岩屑碎屑识别子模型、石英长石碎屑识别子模型、硅质钙质胶结物识别子模型、泥质胶结物识别子模型和杂基填隙物识别子模型；
将岩石薄片分割图像输入分类识别模型中，输出得到单晶态组分集合图像和非单晶态组分集合图像，包括：
将碎屑岩岩石薄片输入颗粒分类子模型中，输出得到碎屑子图像和填隙物子图像；
将碎屑子图像输入碎屑分类子模型中，输出得到非陆源碎屑子图像和陆源碎屑子图像；
将陆源碎屑子图像输入陆源碎屑分类子模型中，输出得到岩屑碎屑子图像和石英长石碎屑子图像；
将填隙物子图像输入填隙物分类子模型中，输出得到杂基填隙物子图像和胶结物子图像；
将胶结物子图像输入胶结物分类子模型中，输出得到硅质钙质胶结物子图像和泥质胶结物子图像；
将非陆源碎屑子图像输入非陆源碎屑识别子模型中，输出得到非陆源碎屑识别结果，并基于非陆源碎屑识别结果确定非陆源碎屑区域；
将岩屑子图像输入岩屑碎屑识别子模型中，输出得到岩屑碎屑识别结果，并基于岩屑碎屑识别结果确定岩屑区域；
将石英长石子图像输入石英长石碎屑识别子模型中，输出得到石英长石碎屑识别结果，并基于石英长石碎屑识别结果确定石英长石碎屑区域；
将杂基填隙物子图像输入杂基填隙物识别子模型中，输出得到杂基填隙物识别结果，并基于杂基填隙物识别结果确定杂基填隙物区域；
将硅质钙质胶结物子图像输入硅质钙质胶结物识别子模型中，输出得到硅质钙质胶结物识别结果，并基于硅质钙质胶结物识别结果确定硅质钙质胶结物区域；
将泥质胶结物子图像输入泥质胶结物识别子模型中，输出得到泥质胶结物识别结果，并基于泥质胶结物识别结果确定泥质胶结物区域；
基于石英长石碎屑区域与硅质钙质胶结物区域，确定单晶态组分区域；
基于非陆源碎屑区域、岩屑碎屑区域、泥质胶结物区域以及杂基填隙物区域确定非单晶态组分区域。
在一个可选的实施例中，该方法，还包括：
获取样本薄片图像，样本薄片图像标注有样本分割区域以及样本组分类型特征；
将样本薄片图像输入岩石薄片图像识别模型中，输出得到预测分割区域以及预测组分类型特征；
将预测分割区域与样本分割区域进行比对，得到样本分割差异，并将预测组分类型特征与样本组分类型特征进行比对，得到样本组分差异；
基于样本分割差异以及样本组分差异，对岩石薄片图像识别模型进行调整。
在一个可选的实施例中，岩石薄片鉴定结果中包括最大颗粒粒径识别结果；
基于几何特征与矿物特征，生成岩石薄片结构鉴定结果，包括：
基于矿物特征确定与最大颗粒粒径识别结果对应的最大颗粒组分区域；
基于几何特征确定与组分区域对应的最大颗粒粒径，并基于最大颗粒粒径生成最大颗粒粒径识别结果。
在一个可选的实施例中，岩石薄片鉴定结果中包括颗粒分选度结果；
基于几何特征与矿物特征，生成岩石薄片结构鉴定结果，包括；
基于矿物特征确定与颗粒分选度识别结果对应的分选度组分区域集合，分选度组分区域集合中包括至少两个组分区域；
基于分选度组分区域集合中的组分区域对应的几何特征，确定颗粒粒径分级对应关系，颗粒粒径分级对应关系中包括至少两级颗粒粒径级别，以及处于颗粒粒径级别中的组分区域的数量；
基于颗粒粒径分级对应关系，确定颗粒分选度结果。
在一个可选的实施例中，岩石薄片鉴定结果中包括颗粒磨圆度识别结果；
基于几何特征与矿物特征，生成岩石薄片结构鉴定结果，包括：
基于矿物特征确定与颗粒磨圆度识别结果对应的磨圆度识别组分区域集合，磨圆度识别组分区域集合中包括至少两个组分区域；
基于磨圆度组分识别区域集合中的组分区域对应的几何特征，确定磨圆度分级对应关系，磨圆度分级对应关系中包括至少两个磨圆度级别，以及处于磨圆度级别中的组分区域的数量；
基于磨圆度分级对应关系，确定颗粒磨圆度识别结果。
在一个可选的实施例中，岩石薄片鉴定结果中包括颗粒自形程度结果；
基于几何特征与矿物特征，生成岩石薄片结构鉴定结果，包括：
基于矿物特征确定与颗粒自形程度结果对应的自形程度组分区域集合，自形程度组分区域集合中包括至少两个自形程度组分区域；
基于自形程度组分区域集合中的组分区域对应的几何特征，确定组分区域的形状；
基于组分区域的形状，确定颗粒自形程度结果。
在一个可选的实施例中，岩石薄片鉴定结果中包括颗粒接触方式识别结果；
基于几何特征与矿物特征，生成岩石薄片结构鉴定结果，包括：
基于矿物特征确定至少两组组分区域对，组分区域对中包括两个相邻的组分区域；
基于几何特征，确定组分区域对中的两个组分区域的交并比；
基于至少两个组分区域的交并比，确定颗粒接触方式识别结果。
在一个可选的实施例中，基于岩石薄片鉴定结果确定岩石薄片名称，包括：
基于细分岩类特征，确定与岩石薄片名称对应的命名规则；
基于命名规则以及岩石薄片结构鉴定结果，生成岩石薄片名称。
另一方面，提供了一种岩石鉴定系统，该岩石鉴定系统包括图像采集设备与计算机设备，图像采集设备与计算机设备连接；
图像采集设备，用于生成岩石薄片图像；向计算机设备发送岩石薄片图像；
计算机设备，用于接收图像采集设备发送的岩石薄片图像，岩石薄片图像为对岩石薄片进行拍摄得到的图像，岩石薄片为对岩石样本进行切割得到的薄片，岩石薄片图像中包括至少一个组分区域；基于岩石薄片图像生成与岩石薄片对应的几何特征、矿物特征以及结构特征，其中，几何特征用于指示岩石薄片的组分区域划分情况，矿物特征用于指示所述岩石薄片中与组分区域对应的矿物种类分布情况，得到的矿物鉴定结果，结构特征用于指示岩石样本的岩石种类；基于几何特征、矿物特征以及结构特征，生成岩石薄片的鉴定结果。
另一方面，提供了一种岩石鉴定装置，该装置包括：
接收模块，用于接收图像采集设备发送的岩石薄片图像，岩石薄片图像为对岩石薄片进行拍摄得到的图像，岩石薄片为对岩石样本进行切割得到的薄片，岩石薄片图像中包括至少一个组分区域；
生成模块，用于基于岩石薄片图像生成与岩石薄片对应的几何特征、矿物特征以及结构特征，其中， 几何特征用于指示岩石薄片的组分区域划分情况，矿物特征用于指示岩石薄片中与组分区域对应的的矿物种类分布情况，得到的矿物鉴定结果，结构特征用于指示岩石样本的岩石种类；
生成模块，用于基于几何特征、矿物特征以及结构特征，生成岩石薄片的鉴定结果。
另一方面，提供了一种计算机设备，计算机设备包括处理器和存储器，存储器中存储有至少一条指令、至少一段程序、代码集或指令集，处理器可加载并执行至少一条指令、至少一段程序、代码集或指令集，以实现上述本申请实施例中提供的岩石鉴定方法。
另一方面，提供了一种计算机可读存储介质，可读存储介质中存储有至少一条指令、至少一段程序、代码集或指令集，处理器可加载并执行至少一条指令、至少一段程序、代码集或指令集，以实现上述本申请实施例中提供的岩石鉴定方法。
另一方面，提供了一种计算机程序产品或计算机程序，该计算机程序产品或计算机程序包括计算机程序指令，该计算机程序指令存储于计算机可读存储介质中。处理器从计算机可读存储介质读取该计算机指令，并执行该计算机指令，使得该计算机设备执行如本申请实施例中提供的岩石鉴定方法。
本申请提供的技术方案带来的有益效果至少包括：
在获取岩石薄片图像后，对于岩石薄片图像进行基于几何特征、矿物特征以及结构特征的三个维度的特征提取，从微观组成和宏观表现的角度，综合多个特征维度对于岩石的性质进行确定，最终生成包括有文字化描述的鉴定结果。在对于岩石进行鉴定的过程中，在获取岩石薄片对应的微观可视化图像后，对该图像进行多个维度的特征提取，并在参考多个维度的特征的情况下对于岩石薄片进行鉴定，提高了岩石鉴定的准确率。
附图说明
为了更清楚地说明本申请实施例中的技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1示出了本申请一个示例性实施例提供的一种岩石鉴定系统的示意图；
图2示出了本申请一个示例性实施例提供的一种图像采集设备的结构示意图；
图3示出了本申请一个示例性实施例提供的一种岩石鉴定方法的流程图；
图4示出了本申请一个示例性实施例提供的一种岩石薄片图像的结构示意图；
图5示出了本申请一个示例性实施例提供的一种确定几何特征、矿物特征以及结构特征的方法流程图；
图6示出了本申请一个示例性实施例提供的一种对岩石薄片图像进行分割的方法的流程示意图；
图7示出了本申请一个示例性实施例提供的一种图像分割模型的结构框图；
图8示出了本申请一个示例性实施例提供的一种经过分割的岩石薄片分割图像的示意图；
图9示出了本申请一个示例性实施例提供的一种矿物特征的识别方法的流程图；
图10示出了本申请一个示例性实施例提供的一种岩石薄片分割图像中提取分割区域的示意图；
图11示出了本申请一个示例性实施例提供的另一种岩石鉴定系统的示意图；
图12示出了本申请一个示例性实施例提供的一种矿物数据采集设备的装置示意图；
图13示出了本申请一个示例性实施例提供的一种确定与岩石薄片对应的矿物特征的识别方法的流程示意图；
图14示出了本申请一个示例性实施例提供的一种与碎屑岩对应的岩石薄片的示意图；
图15示出了本申请一个示例性实施例提供的一种与碎屑岩对应的岩石识别模型的结构示意图；
图16示出了本申请一个示例性实施例提供的一种确定与组分区域对应的矿物特征的方法的流程示意图；
图17示出了本申请一个示例性实施例提供的一种矿物光谱数据库的结构示意图；
图18示出了本申请一个示例性实施例提供的一种对于几何特征及矿物特征进行确定的过程示意图；
图19示出了本申请一个示例性实施例提供的一种岩石薄片的鉴定结果的生成方法的示意图；
图20示出了本申请一个示例性实施例提供的一种岩石鉴定方法的过程示意图；
图21示出了本申请一个示例性实施例提供的一种薄片鉴定报告的内容示意图；
图22示出了本申请一个示例性实施例提供的一种岩石鉴定方法的过程示意图；
图23示出了本申请一个示例性实施例提供的一种岩石鉴定方法的过程示意图；
图24示出了本申请一个示例性实施例提供的一种侵入岩分类相图的示意图；
图25示出了本申请一个示例性实施例提供的一种岩石鉴定装置的结构框图；
图26示出了本申请一个示例性实施例提供的另一种岩石鉴定装置的结构框图；
图27示出了本申请一个示例性实施例提供的一种执行岩石鉴定方法的计算机设备的结构示意图。
具体实施方式
为使本申请的目的、技术方案和优点更加清楚，下面将结合附图对本申请实施方式作进一步地详细描述。
首先，对于本申请各个实施例中涉及的名词进行解释：
组分，指混合物中的各个成分。在本申请中，组分是指固态材料的成分。由于固态材料为混合物，在该混合物中，即会包括至少两种相互分离的组分。该固态材料也即被称为固态多组分混合材料。在一个示例中，混合物实现为金属丝，则金属丝的组分包括铅、镉、铋、锡中的至少一种；在另一个示例中，混合物实现为岩石，则岩石的组分包括但不限于陆源碎屑组分、非陆源碎屑组分、胶结物组分和杂基组分中的至少一种。在本申请中，组分指示岩石薄片中包含的组分，在岩石薄片图像中，对应不同的组分，有不同的组分区域。
图像处理技术，是用计算机对图像信息进行处理的技术，图像处理技术主要包括图像数字化、图像增强和复原、图像数据编码、图像分割和图像识别等。其中，图像分割是将图像分割成若干个特定、具有独特性质的区域并提出感兴趣目标的技术和过程。图像分割的方法包括基于阈值的分割方法、基于区域的分割方法、基于边缘的分割方法以及基于特定理论的分割方法中的至少一种。图像识别，是指利用计算机对图像进行处理、分析和理解，以识别各种不同模式的目标和对象的技术，是应用深度学习算法的一种实践应用。现阶段图像识别技术一般分为人脸识别与商品识别，人脸识别主要运用在安全检查、身份核验与移动支付中；商品识别主要运用在商品流通过程中，特别是无人货架、智能零售柜等无人零售领域。本申请实施例中，图像识别和图像分割技术均将被应用于岩石鉴定领域。
人工智能(Artificial Intelligence，AI)，是利用数字计算机或者数字计算机控制的机器模拟、延伸和扩展人的智能，感知环境、获取知识并使用知识获得最佳结果的理论、方法、技术及应用系统。人工智能企图了解智能的实质，并生产出一种能以人类智能相似的方式做出反应的智能机器。人工智能目的是使机器具有感知、推理与决策的功能。
人工智能技术是一门综合学科，涉及领域广泛。人工智能基础技术包括但不限于传感器技术、人工智能芯片技术、云计算技术、大数据处理技术、机电一体化技术。本申请实施例中所应用的人工智能技术为机器学习技术，且该机器学习设备应用于计算机设备当中。
机器学习(Machine Learning，ML)，是一门多领域交叉学科，涉及概率论、统计学、算法复杂度理论等多个学科领域。机器学习学科专门用于研究计算机怎样模拟或实现人类的学习行为，以使计算机获取新的知识，重新组织已有的知识结构，进而改善自身性能。机器学习通常与深度学习相结合，机器学习和深度学习通常包括人工神经网络、置信网络、强化学习、迁移学习、归纳学习、示教学习等技术。
光谱，是复色光经过色散系统分光后，被色散开的单色光按波长大小而一次排列的图案。光谱的全称为光学频谱。在本申请实施例中，光谱是光谱数据采集设备针对组分区域中的测量点进行采集得到的。以组分为矿物组分为例，当矿物受电磁辐射照射时，其内部的分子会产生量子化的能级之间的跃迁，进而导致发射、吸收或散射、辐射现象的产生，从而生成光谱。光谱实现为光谱数据的可视化表征。也即，在生成光谱的同时，光谱数据采集设备也会同时生成对应的光谱数据，并将此类光谱数据发送至其他计算机设备当中。
数据库，是按照数据结构来组织、存储和管理数据的虚拟仓库，是一个长期存储在计算机内的、有组织的、可共享的、统一管理的大量数据的集合。在本申请实施例中，对应分子光谱，则计算机设备可以调取光谱数据数据库，以对于光谱数据进行确认以及解析。
面扫描，即面扫描分析，是利用测试仪器对于材料组分进行精确鉴定的主流方法。通常做法是，将材料组分切成薄片或制成表面抛光平整的样品，再采用面扫描的方式对其中的组分进行准确鉴定。面扫描所采用的测试仪器包括但不限于诸如红外光谱与拉曼光谱的分子光谱、X射线荧光光谱(X-ray Fluorescence，XRF)、X射线能谱分析(Energy Dispersive X-ray Spectroscopy，EDS)等，且在本申请的一些实施例中，面扫描过程可以联用显微镜或扫描电子显微镜，将显微观察与微区分析相结合。
本申请针对相关技术中人工识别的岩石鉴定方式识别过程繁杂，专业要求极高，但准确率较低的情况，提供了一种岩石鉴定系统以及岩石鉴定方法，借助基于人工智能的图像识别技术，在岩石鉴定过程中生成多维的特征，提高了岩石鉴定的准确率。
图1示出了本申请一个示例性实施例提供的一种岩石鉴定系统的示意图，该装置包括图像采集设备101和计算机设备102，图像采集设备101与计算机设备102连接。
在本申请实施例中，计算机设备即为执行岩石鉴定方法的终端设备。该计算机设备具有数据发送、数据接收以及数据处理的功能。
图像采集设备为在计算机设备执行岩石鉴定方法的过程中，向计算机设备提供图像的设备。在本申请实施例中，图像采集设备具有图像摄制以及图像发送的功能。在一个示例中，图2示出了本申请一个示例性实施例提供的一种图像采集设备的结构示意图，该图像采集设备包括第一光源201、第一载物台202、第一物镜203、偏振装置204、电荷耦合器件(Charge-Coupled Device，CCD)相机205、第一底座206和第一支架207。请参考图2，第一载物台202位于第一支架207中央，第一物镜203位于第一支架的顶部，第一物镜203内具有检偏镜，第一物镜203的底部具有第一安装物镜转换器，第一安装物镜转换器与第一物镜203连接。CCD相机205位于第一物镜203的顶部，偏正装置204包括起偏镜和检偏镜，起偏镜和第一光源201位于第一底座上，第一底座206与第一支架207的底部连接。当图像采集设备200生成岩石薄片图像时，第一光源201启动并发出光线，该光线所对应的光路通过起偏镜和检偏镜，最终投射至岩石薄片上。岩石薄片位于第一载物台202顶部，也即，被置于第一载物台202上。CCD相机处于第一工作状态，也即，CCD相机对应物镜的位置，拍摄得到对应岩石薄片的显微图像，作为岩石薄片图像。
在本申请实施例中，图像采集设备与计算机设备之间建立有通信连接。基于该通信连接，图像采集设备可以将图像发送至计算机设备当中进行存储。在计算机设备接收到图像采集设备发送的图像后，即对于图像进行处理，得到与本方案相关的，适配岩石鉴定过程的特征，并基于上述适配岩石鉴定过程的特征进行岩石鉴定。
对应上述岩石鉴定系统，图3示出了本申请一个示例性实施例提供的一种岩石鉴定方法的流程图，以该方法应用于岩石鉴定系统的计算机设备中里进行说明，该方法包括：
步骤301，接收岩石薄片图像。
在一个示例中，在接收图像采集设备发送的岩石薄片图像之前，计算机设备向图像采集设备发送图像采集指令，在接收到图像采集指令后，图像采集设备对应该图像采集指令进行图像的采集，并将采集得到的图像发送至计算机设备。计算机设备对于图像采集设备发送的图像进行筛选以及处理，最终得到岩石薄片图像。岩石薄片为对岩石样本进行切割得到的薄片，也即，岩石薄片中的组分可以对岩石样本的组分进行代表与表征。对应地，岩石薄片图像中也即包括至少一个组分区域，每个组分区域中包括一种组分。在本申请实施例中，组分即指示与岩石对应的组分。请参考图4，由于岩石薄片图像400为对于岩石薄片进行显微拍摄得到的图像，故岩石薄片图像中包括组分区域401，在图4中，组分区域的数量为18个，该18个组分区域可以表征相同的组分，也可以表征互不相同的组分，也可以表示至少两个互不相同的组分，且存在组分区域中的组分相同的情况。本申请实施例对于组分区域的内容不做限定。在本申请实施例中，组分区域中包括的组分实现为矿物组分，也可以实现为结构组分。可选地，当组分区域中包括的内容为矿 物组分时，组分区域可以指示一类矿物。当组分区域中包括的内容为结构组分时，组分区域可以指示一类颗粒，或，一类填隙物。
需要说明的是，本申请实施例中的岩石薄片图像的数量为2张，分别为与岩石薄片对应的单偏光图像和正交偏光图像；或，本申请实施例中的岩石薄片数量为1张，且该图像是将岩石薄片对应的单偏光图像和正交偏光图像进行组合得到的图像。本申请实施例对于岩石薄片图像的实际表现形式不作限定，但岩石薄片图像需体现岩石薄片在普通光和偏振光光源下时表现出的各向异性。
步骤302，基于岩石薄片图像生成与岩石薄片对应的几何特征、矿物特征以及结构特征。
在本申请实施例中，在接收到岩石薄片图像的情况下，对应岩石薄片的性质，计算机设备将会对岩石薄片图像进行特征提取，对应岩石薄片的几何特征、矿物特征以及结构特征。可选地，几何特征用于指示岩石薄片基于组分的区域划分情况，矿物特征用于指示对岩石薄片进行矿物鉴定，得到的矿物鉴定结果，结构特征用于指示岩石样本的空间结构。在本申请实施例中，岩石薄片图像中的组分区域的分布图像可以表征几何特征，岩石薄片图像中组分区域的具体矿物类型可以表征矿物特征，岩石薄片图像所体现的空间结构性质可以表征结构特征。本申请实施例对于几何特征、矿物特征以及结构特征的具体内容不作限定。
步骤303，基于几何特征、矿物特征以及结构特征，生成岩石薄片的鉴定结果。
在本申请实施例中，在计算机设备确定岩石薄片图像表征的，也即，与岩石薄片对应的几何特征、矿物特征和结构特征后，即基于上述特征进行岩石薄片的鉴定结果的生成。鉴定结果中包括对于岩石样本的特征的文字化描述。在一个示例中，鉴定结果包括对于岩石样本的定名，以及对于几何特征、矿物特征和结构特征的文字化归纳。可选地，鉴定结果以列表的形式进行表示，或，鉴定结果还包括可视化图像内容，以图表的形式进行表示。本申请对于鉴定结构的生成方法不作限定。
综上所述，本申请实施例提供的方法，在获取岩石薄片图像后，对于岩石薄片图像进行基于几何特征、矿物特征以及结构特征的三个维度的特征提取，从微观组成和宏观表现的角度，多个特征维度对于岩石的性质进行确定，最终生成包括有文字化描述的鉴定结果。在对于岩石进行鉴定的过程中，在获取岩石薄片对应的微观可视化图像后，对该图像进行多个维度的特征提取，并在参考多个维度的特征的情况下对于岩石薄片进行鉴定，提高了岩石鉴定的准确率。
在本申请的一些实施例中，几何特征、矿物特征与结构特征之间存在相关性。在一个示例中。结构特征指示岩石的具体种类，而几何特征与矿物特征可以为岩石的具体种类的鉴定提供指导；在另一个示例中，矿物特征指示与组分区域对应的矿物，故在进行矿物特征的确定之前，计算机设备需要优先进行几何特征的确定。图5示出了本申请一个示例性实施例提供的一种确定几何特征、矿物特征以及结构特征的方法流程图，该方法可以替换如图3所示的步骤302，实现为步骤501至步骤506，该方法包括：
步骤501，基于岩石薄片图像生成初选岩类结构特征。
在本申请实施例中，结构特征包括初选岩类结构特征以及细分岩类结构特征。其中，初选岩类结构特征用于对于岩石薄片对应的岩石样本进行分类，且分类至初选岩石种类，初选岩石种类也即为沉积岩类、岩浆岩类和变质岩类。
在一个示例中，将岩石薄片图像输入初选岩类选择模型，输出得到初选岩类结果，初选岩类结果即可指示岩石薄片对应的初选岩类结构特征。初选岩类结构选择模型为基于初选岩类图像样本集构建的模型。在将岩石薄片图像输入初选岩类选择模型后，初选岩类选择模型会将岩石薄片图像与样本集中保存的样本图像进行对比，并进而确定初选岩类结构特征。
步骤502，基于初选岩类结构特征确定与岩石薄片对应的初选岩石种类。
在本申请实施例中，初选岩类结果包括沉积岩类、岩浆岩类和变质岩类，也即，初选岩类为根据成因划分得到的三种岩类。对应不同的初选岩类结构特征，即可得到初选岩石种类。在一个示例中，初选岩类结构特征实现为与对应的初选岩石种类的相似度。
步骤503，根据初选岩石种类，基于组分图像的位置对岩石薄片图像进行划分，得到岩石薄片分割图像，岩石薄片分割图像中包括至少两个分割区域。
可选地，在确定初选岩石种类后，计算机设备即可进一步确定与初选岩石种类对应的组分区域内容， 以及组分区域的常见形状。在本申请实施例中，计算机设备基于该组分区域内容以及组分区域的常见形状对于岩石薄片图像进行分割，得到岩石薄片分割图像，岩石薄片分割图像中包括至少两个分割区域。分割区域即用于指示岩石薄片图像基于其所具有的组分区域的分割情况。
步骤504，基于分割区域确定与岩石薄片对应的几何特征。
在确定分割区域后，即可基于分割区域，对于与岩石薄片对应的几何特征进行确定。在此过程中，计算机设备即将分割区域视为组分区域，进行几何特征的识别。在该过程中，经过计算机设备的对应处理后，几何特征可用于表征组分区域的面积、组分区域的最大直径、组分区域的形状中的至少一种。
步骤505，基于分割区域确定与岩石薄片对应的矿物特征。
在确定与岩石薄片对应的几何特征后，即可基于初选岩类结构特征，进一步确定与岩石薄片对应的矿物特征。在本申请实施例中，矿物特征用于指示各个分割区域中的组分的矿物种类。
需要说明的是，在本申请实施例为例的部分实施例中，首先确定与岩石薄片对应的几何特征，其次确定与岩石薄片对应的矿物特征；在本申请的一些其他实施例中，首先确定与岩石薄片对应的矿物特征，其次确定与岩石薄片对应的几何特征；在本申请的一些其他实施例中，同时确定与岩石薄片对应的几何特征以及矿物特征。本申请对于几何特征以及矿物特征的确定先后次序不做限定。
步骤506，基于几何特征与矿物特征，确定细分岩类结构特征。
在本申请实施例中，在确定几何特征与矿物特征后，即可进一步确定细分岩类结构特征，细分岩类结构特征用于指示岩石样本对应的细分岩石种类，也即，在成因的基础上，结合岩石组分区域和组分区域的矿物组成，进一步确定的种类。
可选地，将几何特征与矿物特征输入细分岩类选择模型中，输出得到细分岩类结果，细分岩类结果指示岩石薄片对应的细分岩类结构特征，细分岩类选择模型为基于几何-矿物特征交互样本集构建的模型，几何-矿物特征交互样本集指示几何特征与矿物特征的组合，与细分岩类结构特征之间的对应关系。在一个示例中，对应岩石样本的初选岩类结构特征指示岩石样本为沉积岩，而通过细分岩类选择模型，根据岩石样本薄片中的分割区域的性质以及分割区域的种类，得到与岩石样本对应的细分岩类结构特征指示岩石样本为沉积岩中的碎屑岩。
可选地，初选岩石种类与细分岩石种类均为鉴定结果中包括的内容。
综上所述，本申请实施例提供的方法，对应结构特征对于岩石鉴定的实际应用内容，将结构特征分类为初选岩类结构特征以及细分岩类结构特征，并在确定初选岩类结构特征后，进一步确定与岩石薄片图像对应的几何特征与结构特征，进而确定细分岩类结构特征，使在岩石鉴定的过程中，几何特征与结构特征的确定过程更加具有针对性，进一步提高了岩石鉴定的准确率。
接下来，对应图5所示的实施例中涉及的几何特征以及矿物特征的详细确定方法进行逐个说明。
如上实施例中所述，几何特征即为用于指示岩石薄片基于组分的区域划分情况。本申请中，将通过基于机器学习的神经网络模型，对于几何特征进行获取。图6示出了本申请一个示例性实施例提供的一种对岩石薄片图像进行分割的方法的流程示意图，该方法可以替换图5所示实施例中的步骤503至步骤504。以该方法应用于计算机设备中为例进行说明，请参考图6，该方法包括：
步骤601，获取样本薄片图像，样本薄片图像标注有样本分割结果。
本申请实施例重要说明的是岩石薄片图像分割模型的构建以及应用过程，该岩石薄片图像分割模型即为对于组分区域进行划分的模型。在本申请实施例中，在选择岩石薄片图像分割模型所应用的基础网络后，即通过样本训练的方式实现机器学习过程。步骤601即为获取样本薄片图像，也即，构建样本集的过程。
在本申请实施例中，样本薄片图像为对应岩石薄片图像的内容选取的图像。在一个示例中，样本薄片图像为与岩石薄片的初选岩石种类相对应的岩石薄片图像；在另一个示例中，样本薄片图像为与岩石薄片的细分岩石种类相对应的岩石薄片图像。本申请对于样本薄片图像所对应的具体岩石种类不作限定。
可选地，样本薄片图像标注有样本分割结果。该样本分割结果即为通过其他方式进行判断后，确定的样本薄片图像基于组分区域的分割区域的划分方式。
步骤602，将样本薄片图像输入岩石薄片图像分割模型，输出得到预测分割结果。
在本申请实施例中，岩石薄片图像分割模型为基于机器学习的Mask-RCNN网络模型。在一个示例中，请参考图7，该岩石薄片图像分割模型700中包括特征图像生成网络701，区域建议网络702以及结果生成网络703，该三个子网络在岩石薄片图像分割模型中相互连接。其中，特征图像生成网络701中包括特征提取子网络7011以及特征强化子网络7012，结果生成网络703中包括结果区域分类分支网络7031以及边框回归分支网络7032。
在通过岩石薄片图像分割模型获取分割结果时。在岩石薄片图像进入岩石薄片图像分割模型后，将经过特征图像生成网络，在特征提取子网络中，经过特征提取子网络的表层特征提取以及特征强化子网络的特征强化，即可得到与岩石薄片图像相对应的特征图像。随后，特征图像被输入区域建议网络，输出得到与待测薄片对应的初步分割特征图像。在经过便捷细化处理后，将初步分割特征图像输入区域分类分支网络中，通过区域分类分支网络对于初步分割特征图像进行性质分类，并通过边框回归分支网络对初步分割特征图像进行修正，即最终输出分割结果。
在本申请实施例中，针对构建完成的岩石薄片图像分割模型，通过输入样本薄片图像的方式，获取预测分割结果。该预测分割结果可以以叠加显示有分割线的样本岩石薄片的形式输出。
步骤603，将预测分割结果与样本分割结果进行差异比较，得到分割差异。
该过程即为将预测分割结果与样本分割结果进行比较，以确定二者的差异的过程。由于造成预测分割结果与样本分割结果之间的差异的原因为，岩石薄片图像分割模型的参数调整并未完成。在一个示例中，分割差异实现为岩石薄片分割图像中分割线的长度差异；在另一个示例中，分割差异实现为岩石薄片分割图像中分割线的形状差异。本申请对于分割差异的具体实现形式不做限定。
步骤604，基于分割差异对岩石薄片图像分割模型进行调整。
在本申请实施例中，基于分割差异的差异情况，确定岩石薄片图像分割模型的训练是否完成。若岩石薄片图像分割模型的训练未完成，则通过继续将样本薄片图像输入岩石薄片图像分割模型的方式，输出得到其他的预测分割结果，并进行差异比较，以执行模型的训练过程。可选地，在本申请的其他实施例中，还可以基于岩石薄片图像分割模型的已训练次数，确定岩石薄片图像分割模型的训练是否完成。在本申请实施例中，响应于分割差异指示预测分割结果与样本分割结果的差异阈值时，确定岩石薄片图像分割模型的训练完成；或，响应于岩石薄片图像分割模型在一个分割训练周期中的样本薄片图像输入次数达到分割次数阈值时，确定岩石薄片图像分割模型的训练完成。
需要说明的是，本申请中出现的各个“阈值”，可以实现为计算机程序中存储的阈值，也可以实现为在计算机设备执行相应步骤时，人工输入的阈值，本申请对于阈值的实际获取方式不做限定。
在确定岩石薄片图像分割模型的训练完成后，即通过输入待测的岩石薄片图像的方式，对于岩石薄片图像进行分割，以进行集合特征的提取。
步骤605，响应于岩石薄片图像分割模型训练完成，将岩石薄片图像输入岩石薄片图像分割模型中，输出得到岩石薄片分割图像。
步骤605所对应的过程即为生成与岩石薄片对应的岩石薄片分割图像的过程，该岩石薄片分割图像中包括了至少两个分割区域，如前文所述，分割区域指示了基于组分区域确定的分割情况。在岩石薄片图像分割模型训练完成的情况下，分割区域即可直接表示组分区域。如图8所示，在经过岩石薄片图像分割模型后得到的岩石薄片分割图像800中，即以分割线801的方式划分出了多个组分区域802。以图8为例，被分割线801全包围的区域即为组分区域802。
在得到岩石薄片分割图像后，即可生成与岩石薄片图像对应的分割结果。分割结果包括与岩石薄片图像对应的组分区域的数量、与岩石薄片图像对应的组分区域的分布密度以及与岩石薄片图像对应的组分区域的大小中的至少一种。在一个示例中，上述分割结果可以直接指示与岩石薄片对应的几何特征，或，上述分割结果可以经过处理，生成与岩石薄片对应的几何特征。
综上所述，本申请实施例提供的方法，在对于岩石薄片图像进行鉴定的过程中，通过基于Mask-RCNN网络的岩石薄片图像分割模型，对于岩石薄片图像进行分割，以得到与岩石薄片相对应的几何特征。在应用计算机中的图像分割方式代替人工分割方式的基础上，引入了用于图像处理的基于机器学习的 Mask-RCNN网络模型，对岩石薄片图像中的组分区域提供了更好的划分指导，提高了几何特征的获取效率。
可选地，在本申请中，图6所示的实施例即为一种计算机设备获取几何特征的实施例。在本申请的其他实施例中，几何特征还包括其他的获取方式。本申请对于几何特征的具体获取方式不做限定。
本申请实施例中，由于几何特征指示组分区域的划分情况，且每个组分区域中对应不同的组分，故在确定几何特征后，进行矿物特征的确定，也即，进行岩石薄片中的矿物分布情况的确定。
在本申请的一个实施例中，通过基于机器学习的神经网络模型进行矿物特征的确定。图9示出了本申请一个示例性实施例提供的一种矿物特征的识别方法的流程图，该方法可以替换实现为图5所示的实施例中的步骤505。以该方法应用于计算机设备中为例进行说明，该方法包括：
步骤901，基于分割区域确定组分图像，组分图像中包括一个组分区域。
在本申请实施例中，请参考图10，分割区域1001即为基于岩石薄片分割图像1000确定的分割结果，该分割区域1001即可用于进行组分图像1010的确定，组分图像1010中包括一个组分区域。由于岩石薄片图像分割模型已经过训练，故该组分区域与岩石薄片中包括的组分区域对应，直接表征岩石薄片中的组分。
步骤902，获取样本组分图像，样本组分图像标注有样本组分类型特征。
在本申请实施例中，通过组分种类识别模型，进行与组分区域对应的组分类型特征的确定，也即，进行与组分区域对应的矿物组成的确定。可选地，组分种类识别模型的构建方式与岩石薄片图像分割模型的构建方式相似。均为基于机器学习的样本训练。在本申请实施例中，样本组分图像选自于样本组分图像数据集，且每个样本组分图像中均包括且仅包括一个组分区域，该样本组分区域对应有样本组分类型特征。
步骤903，将样本组分图像输入组分种类识别模型，输出得到预测组分类型特征。
在本申请实施例中，预测组分类型特征即为处于训练过程中的组分种类识别模型对样本组分图像进行识别，得到的预测组分类型特征。在本申请实施例中，组分种类识别模型可选用的地层神经网络包括Mobilenet神经网络，Resnet神经网络，VGG(Visual Geometry Group Network)神经网络中的至少一种。
步骤904，将预测组分类型特征与样本组分类型特征进行差异比较，得到识别差异。
该过程即为确定预测组分类型特征与样本组分类型特征的差异识别过程。在一个示例中，组分类型特征表征为数值形式，不同的数值表征不同的组分类型，以及组分区域属于该组分类型的概率。在此情况下，识别差异可以指示预测组分类型特征与样本组分类型特征的差值，或，预测组分类型特征与样本组分类型特征的比值。本申请实施例对于组分类型特征的实际输出形式不作限定。
步骤905，基于识别差异对组分种类识别模型进行调整。
该过程即为根据组分种类识别差异对于组分种类识别模型进行调整，直至差异小于差异阈值，或组分种类识别模型的训练次数达到训练次数阈值。当满足上述条件，或满足上述条件之一时，即可确定组分种类识别模型的训练完成。
步骤906，响应于组分种类识别模型训练完成，将组分图像输入组分种类识别模型，输出得到组分图像的组分类型特征。
该过程即为对于组分类型特征进行输出的过程。需要说明的是，本申请实施例关注的是对于组分图像中的单个组分的识别，故当组分图像中不可避免地包括了两个组分区域时，组分种类识别模型会将上述两个组分区域进行统一识别，或，组分种类识别模型将在上述两个组分区域中选取与任一组分区域对应的图像区域，进行单独识别。对应一次组分识别的过程，组分种类识别模型将输出且仅输出一个组分类型特征。
步骤907，基于组分图像对应的组分类型特征，确定与岩石薄片对应的矿物特征。
可选地，计算机设备中存储有组分类型特征与矿物种类之间的对应关系，当组分类型特征被输出后，计算机设备即可从组分类型特征与矿物种类之间的对应关系中判断与组分区域对应的矿物种类。
在与岩石薄片图像对应的所有组分区域的矿物种类均被确定后，即可得到与岩石薄片对应的矿物特征。
综上所述，本申请实施例提供的方法，对应单个组分区域，通过将其输入组分种类识别模型中，进行组分识别，并将输出得到指示组分区域的矿物特征的组分类型特征。通过输入具有单个组分区域的组分图 像，借助基于机器学习的识别模型进行组分颗粒的详细类型的识别的方式，提高了对于岩石薄片中的组分的识别效率。
在本申请的另一些实施例中，通过基于分类识别模型进行矿物特征的初步确定，随后通过矿物数据采集设备，对于分类识别模型的初步确定结果进行验证。在此情况下，岩石鉴定系统中即还包括矿物数据采集设备。图11示出了本申请一个示例性实施例提供的一种岩石鉴定系统的结构框图，该岩石鉴定系统包括图像采集设备1101、计算机设备1102以及矿物数据采集设备1103，其中，图像采集设备1101以及矿物数据采集设备1103分别与计算机设备1102连接。对应地，图12示出了本申请一个示例性实施例提供的一种矿物数据采集设备的装置示意图，该矿物数据采集设备包括第二光源1201、第二载物台1202、第二物镜1203、光电信号转换器1204、第二底座1205和第二支架1206；第二光源1201对应有第二光路1207，第二光路1207的终点投射至第二载物台1202；第二载物台1202位于第二支架1207中央，第二物镜1203位于第二支架1207的顶部，第二物镜1203的底部具有第二安装物镜转换器，第二安装物镜转换器与第二物镜1203连接；光电信号转换器1205于第二物镜1203的顶部；当矿物数据采集设备1200生成矿物数据时，岩石薄片位于第二载物台1202顶部，光电信号转换器1204处于第二工作状态，以进行光谱数据的获取。在本申请实施例中，第二工作状态即为光电信号转换器1204进行数据采集时的状态。
需要说明的是，矿物数据采集设备与图像采集设备可以实现为同一设备，也即，矿物数据采集设备与图像采集设备共用第一载物台、第一物镜、第一底座以及第一支架。CCD相机和光电信号转换器均位于第一物镜上，矿物数据采集设被和图像采集设备的集成设备可以通过获取岩石薄片图像和光谱数据，并且将岩石薄片图像和光谱数据共同发送至计算机设备当中。
基于如上所述的岩石鉴定系统，图13示出了本申请一个示例性实施例提供的一种确定与岩石薄片对应的矿物特征的识别方法的流程示意图。该方法可以替换实现为与图5对应的实施例中的步骤505，以该方法应用于计算机设备中为例进行说明，该方法包括：
步骤1301，将岩石薄片分割图像输入分类识别模型中，输出得到单晶态组分集合图像和非单晶态组分集合图像。
本申请实施例中，通过分类识别模型，对于岩石薄片分割图像中，属于不同晶态类型的组分区域进行区分。
在通常情况下，岩石体现为多组分混合物，在该混合物中，根据组分区域内的晶体的宏观形态，组分可以被划分为单晶态组分以及非单晶态组分。可选地，非单晶态组分包括多晶态组分和非晶态组分。在一个示例中，分类识别模型中包括晶态组分分类模型，以将组分区分为单晶态组分以及非单晶态组分，且分类识别模型中还包括非单晶态组分分类模型，将非单晶态组分区分为多晶态组分和非晶态组分，并对应设置单晶态组分识别模型、多晶态组分识别模型以及非晶态组分识别模型，以进行组分区域对应的具体矿物种类的识别。
在此情况下，针对不同的岩石种类，可以对分类识别模型中的分类子模型以及识别子模型的具体种类进行设置，以适配不同岩石种类中包含的不同组分区域的情况。接下来，以岩石薄片的岩石种类为碎屑岩为例进行说明。
对应碎屑岩这一岩石种类，碎屑岩中的组分区域所指示的组分包括碎屑组分与填隙物组分。其中，碎屑是沉积岩或沉积物的一种组分，是母岩机械风华的产物；填隙物是位于各个碎屑之间的连接物质。随后，在进行进一步细分时，将碎屑组分划分为陆源碎屑组分和非陆源碎屑组分。其中，陆源碎屑是指陆源区母岩经物理风化或机械破坏而形成的碎屑物质，主要包括石英、长石、岩屑等，非陆源碎屑是指非陆源区母岩经物理风化或机械破坏而形成的碎屑物质，主要包括内源屑、火山碎屑等。同时，将填隙物组分划分为胶结物组分和杂基组分。其中，杂基指示填充于碎屑颗粒之间的细小机械混入物。在本申请实施例中，杂基可以实现为粉砂、粘土物质、细分、碳酸盐灰泥中的至少一种。胶结物指示填隙物中除杂基以外的化学沉淀位置。再次，将胶结物组分根据其组成的物质成分，进一步划分为硅质钙质组分以及泥质组分。并将陆源碎屑组分根据其碎屑形态，划分为石英长石组分以及岩屑组分。也即，对应碎屑岩这一岩类，分类识 别模型需要具有对于石英长石碎屑区域、硅质钙质胶结物区域、非陆源碎屑区域、岩屑碎屑区域、杂基填隙物区域和泥质胶结物区域的识别能力。图14示出了本申请一个示例性实施例提供的一种与碎屑岩对应的岩石薄片的示意图，该碎屑岩的岩石薄片中包括各种组分，请参考图14。该碎屑岩的岩石薄片1400中包括石英碎屑区域1401、硅质钙质胶结物区域1402、非陆源碎屑区域1403、岩屑碎屑区域1404、杂基填隙物区域1405以及泥质胶结物区域1406。
请参考图15，对应岩石种类为碎屑岩的情况，该岩石识别模型1500中包括分类子模型组以及识别子模型组，分类子模型组中包括组分分类子模型1511、碎屑分类子模型1512、陆源碎屑分类子模型1513、填隙物分类子模型1514和胶结物分类子模型1515，识别子模型组中包括非陆源碎屑识别子模型1521、岩屑碎屑识别子模型1522、石英长石碎屑识别子模型1523、硅质钙质胶结物识别子模型1524、泥质胶结物识别子模型1525和杂基填隙物识别子模型1526。其中，组分分类子模型1511与碎屑分类子模型以及1512以及填隙物分类子模型1514连接，碎屑分类子模型1512分别与非陆源碎屑识别子模型1521以及陆源碎屑分类子模型1513连接，陆源碎屑分类子模型1513分别与岩屑碎屑识别子模型1522以及石英长石碎屑识别子模型1523连接，填隙物分类子模型1514分别与胶结物分类子模型1515以及杂基填隙物识别子模型1526连接，胶结物分类子模型1515分别与硅质钙质胶结物识别子模型1524以及泥质胶结物识别子模型1525连接。
在本示例中，当将岩石薄片分割图像输入岩石识别模型中后，各个分类子模型将会通过将岩石薄片分割图像进行划分以及重组的方式，生成对应分类的子图像，以在岩石薄片分割图像中进行对应区域的确定。各个识别子模型对应各个子图像，进行具体组分种类的确定。在该过程中，也即实现为：将碎屑岩岩石薄片输入组分分类子模型中，输出得到碎屑子图像和填隙物子图像；将碎屑子图像输入碎屑分类子模型中，输出得到非陆源碎屑子图像和陆源碎屑子图像；将陆源碎屑子图像输入陆源碎屑分类子模型中，输出得到岩屑碎屑子图像和石英长石碎屑子图像；将填隙物子图像输入填隙物分类子模型中，输出得到杂基填隙物子图像和胶结物子图像；将胶结物子图像输入胶结物分类子模型中，输出得到硅质钙质胶结物子图像和泥质胶结物子图像；将非陆源碎屑子图像输入非陆源碎屑识别子模型中，输出得到非陆源碎屑识别结果，并基于非陆源碎屑识别结果确定非陆源碎屑区域；将岩屑子图像输入岩屑碎屑识别子模型中，输出得到岩屑碎屑识别结果，并基于岩屑碎屑识别结果确定岩屑区域；将石英长石子图像输入石英长石碎屑识别子模型中，输出得到石英长石碎屑识别结果，并基于石英长石碎屑识别结果确定石英长石碎屑区域；将杂基填隙物子图像输入杂基填隙物识别子模型中，输出得到杂基填隙物识别结果，并基于杂基填隙物识别结果确定杂基填隙物区域；将硅质钙质胶结物子图像输入硅质钙质胶结物识别子模型中，输出得到硅质钙质胶结物识别结果，并基于硅质钙质胶结物识别结果确定硅质钙质胶结物区域；将泥质胶结物子图像输入泥质胶结物识别子模型中，输出得到泥质胶结物识别结果，并基于泥质胶结物识别结果确定泥质胶结物区域基于石英长石碎屑区域与硅质钙质胶结物区域，确定单晶态组分区域；基于非陆源碎屑区域、岩屑碎屑区域、泥质胶结物区域以及杂基填隙物区域确定非单晶态组分区域。
对应碎屑岩的实际情况，石英长石碎屑区域以及硅质钙质胶结物区域属于单晶态组分区域；非陆源碎屑区域、岩屑碎屑区域、泥质胶结物区域以及杂基填隙物区域属于非单晶态组分区域。
也即，在本申请实施例中，单晶态组分集合图像中包括至少一个单晶态组分区域，以及与单晶态组分区域对应的单晶态组分识别结果，非单晶态组分区域中包括至少一个非单晶态组分区域，以及与非单晶态组分区域对应的非单晶态组分识别结果。
需要说明的是，本申请实施例中的分类识别模型，或组成分类识别模型的各个子模型，均为基于深度学习的神经网络模型。
步骤1302，基于单晶态组分集合图像向矿物数据采集设备发送扫描指令。
由于单晶态组分区域对应的图像特征并不明显，故针对单晶态组分集合图像，需要通过矿物数据采集设备进行扫描数据的采集。可选地，扫描数据即为矿物数据采集设备采集到的光谱数据。在本申请实施例中，矿物数据采集设备还可以实现为包括但不限于诸如红外光谱与拉曼光谱的分子光谱、X射线荧光光谱(X-ray Fluorescence，XRF)、X射线能谱分析(Energy Dispersive X-ray Spectroscopy，EDS)中的至少一种设备。计算机设备向矿物数据采集设备发送扫描指令，该扫描指令即向矿物数据采集设备指示单晶态组 分集合图像对应的组分区域的位置。
可选地，由于在岩石薄片的制作过程中会不可避免地引入杂质，且与单晶态组分对应的矿物以发生蚀变现象(指矿物受外界影响，使成分发生改变，生成新的矿物的过程的现象，在一个示例中，长石蚀变为粘土矿物)、双晶现象(指两个或两个以上同种晶体的规则连生，晶体的连接处可能对数据采集产生干扰的现象，在一个示例中，斜长石中包括聚片双晶现象)、包裹体现象(矿物中有一相或多相物质组成的，并与主晶矿物具有相的界限的封闭系统的现象，在一个示例中，石英具有金红石包裹体)、裂纹现象(指矿物因受外界应力作用而导致的断裂的现象，在一个示例中，石英具有裂纹)，影响矿物数据采集设备的数据采集结果，故，在进行扫描指令的生成之前，首先将单晶态组分集合图像输入区域筛选模型，得到区域筛选结果，以确定单晶态组分集合图像中适合进行扫描的部分，也即，区域筛选结果指示了单晶态组分集合图像中的干净区域。在此情况下，计算机设备基于区域筛选结果生成扫描指令，并向矿物数据采集设备发送扫描指令。
步骤1303，接收矿物数据采集设备基于扫描指令反馈的扫描数据。
在矿物数据采集设备接收到扫描指令后。在一个示例中，矿物数据采集设备基于扫描指令进行物镜的移动，或，进行载物台的移动，使置于载物台上的岩石薄片对应的区域与物镜相对，并通过光源的照射与光电信号转换器的信号转换，进行扫描数据的采集，并反馈至计算机设备当中。在本申请实施例中，扫描数据可以实现为光谱数据。
步骤1304，基于扫描数据对单晶态组分识别结果进行验证，得到验证结果。
扫描数据可以从化学成分维度，或，晶体结构维度对于单晶态组分集合图像中的每个组分区域中的组分类型进行确定，其中，单晶态组分集合图像中的每个单晶态组分区域均对应有指示其化学成分的扫描数据，故扫描数据可以对单晶态组分集合中的每个单晶态组分中所具有的单晶态识别结果进行确定。
步骤1305，基于非单晶态组分识别结果、单晶态组分识别结果以及验证结果，确定与岩石薄片对应的矿物特征。
当验证结果与单晶态组分识别结果相同时，确定验证结果指示验证通过，并确定单晶态组分识别结果不变；
当验证结果与单晶态组分结果识别不同时，确定验证结果指示验证不通过。在本申请实施例中，并基于识别结果生成规则确定单晶态组分识别结果。该识别结果生成规则是基于矿物数据采集设备对于单晶态组分识别结果的影响，生成的确定规则。该规则的产生的原因是，矿物数据采集设备的扫描数据生成结果受到矿物数据采集设备的具体种类的影响较大。
在确定各个组分区域对应的矿物类型后，即可将各个组分区域对应的矿物类型进行汇总，进而确定与岩石薄片对应的矿物特征。
综上所述，本申请实施例提供的方法，在对岩石薄片分割图像通过面扫描的方式进行组分确定之前，通过分类识别模型，将岩石薄片分割图像基于晶态进行初步识别，并在选取与单晶态组分区域对应的区域后，对于组分识别结果进行二次验证，提高了最终生成的矿物特征的准确率。
在本申请的另一些实施例中，在通过矿物数据采集设备采集到与组分区域对应的光谱数据后，通过在光谱数据数据库中进行检索比对的方式，进行矿物特征的确定。图16示出了本申请一个示例性实施例提供的一种确定与组分区域对应的矿物特征的方法的流程示意图，该方法可以替换实现为与图5对应的实施例中的步骤505。以该方法应用于计算机设备中为例进行说明，该方法包括：
步骤1601，基于分割区域确定组分图像，组分图像中包括一个组分区域。
该过程即为根据岩石薄片分割图像中的分割区域，对于组分区域进行分割并确定的过程。
步骤1602，基于组分图像向矿物数据采集设备发送光谱数据获取指令。
针对每一个独立的组分区域，在选取组分区域内的光谱数据采集点后，向矿物数据采集设备发送光谱数据获取指令。
步骤1603，接收矿物数据采集设备基于光谱数据获取指令发送的光谱数据。
在本申请实施例中，矿物数据采集设备即获取对应的光谱数据，并向计算机设备进行反馈。
步骤1604，基于光谱数据，在矿物光谱数据数据库中确定与组分区域对应的初选矿物种类。
在本申请实施例中，计算机设备中存储有矿物光谱数据数据库，该矿物光谱数据数据库中存储有光谱数据与初选矿物种类的对应关系。
在本申请实施例中，初选矿物种类为基于矿物的代表性性质进行划分得到的种类。可选地，计算机设备对应光谱数据数据库中的光谱数据样本集进行分类，根据不同矿物的分类数量，确定与矿物对应的稀有度。可选地，在一个初选矿物种类的划分下，还包括至少两个细分矿物种类。
在本申请实施例中，获取初选矿物种类的过程如下：
首先基于光谱数据，在矿物光谱数据数据库中，确定与组分区域对应的初选矿物种类的一级类别，该矿物种类的一级类别为基于矿物的常见程度划分的类别，该一级类别中包括常见矿物类别与非常见矿物类别。在本申请实施例中，矿物的一级类别确定方式通过与一级类别频率阈值比较以实现。可选地，当矿物对应的光谱数据占组分区域中的光谱数据的比例大于一级类别频率阈值，也即与矿物对应的光谱数据的出现频率大于一级类别频率阈值时，确定该矿物的一级类别为常见矿物，当矿物对应的光谱数据的出现频率小于等于一级类别频率阈值时，确定该矿物的以及类别为非常见矿物。在本申请实施例中，基于光谱数据数据库中的矿物样本数量，确定常见类别与非常见矿物类别，也即，当光谱数据数据库中的矿物样本数量小于样本数量阈值时，确定矿物种类的一级类别为非常见矿物类别，当光谱数据数据库中的矿物样本数量大于等于样本数量阈值时，确定矿物种类的一级类别为非常见矿物类别。在此分类下，与常见矿物类别对应的初选矿物种类包括矿物族种类以及普通常见矿物种类，与非常见矿物类别对应的初选矿物种类包括包裹体矿物种类、强敏感矿物种类以及蚀变矿物种类。其中，普通常见矿物种类即指示在矿物光谱数据数据库中，出现频率大于某个初选矿物种类频率阈值的初选矿物种类；包裹体矿物种类即为矿物生长后，体现为包裹体的矿物所对应的矿物种类；强敏感矿物种类指示在光谱数据的获取过程中，因化学性质原因，对光谱数据采集设备的采集准确率产生影响，并使光谱数据采集设备关注自身的矿物所对应的矿物种类；蚀变矿物种类即为在矿物形成过程中产生识别现象的矿物所对应的矿物种类。
需要说明的是，上述过程中出现的“一级类别频率阈值”以及“矿物种类频率阈值”，可以是在计算机设备中预存的阈值，也可以是计算机设备执行对应步骤时，通过接收到的信号确定的阈值，本申请对于各个阈值的生成方式不作限定。
在本申请实施例中，矿物光谱数据数据库也对应具有多个子数据库。请参考图17，矿物光谱数据数据库1700中包括常见矿物类别子数据库1701以及非常见矿物类别子数据库1702，在常见矿物类别子数据库1701中，还包括普通常见矿物种类子数据库1711以及矿物族数据库1712，在非常见矿物类别子数据库1702中，还包括包裹体矿物种类数据库1721，强敏感矿物种类数据库1722以及蚀变矿物种类数据库1723。此外，矿物光谱数据数据库1700中还包括与杂质对应的杂质类别子数据库1703。由于杂质的类型较多，以岩石为碎屑岩为例，碎屑岩中包含的杂质成分为碳化硅，则该杂质类别子数据库1703为与碳化硅对应的识别数据库。
步骤1605，确定与初选矿物种类对应的矿物种类验证规则。
在本申请实施例中，对应上述初选矿物种类，矿物种类验证规则包括分类验证规则、直接验证规则以及选点重验规则。响应于初选矿物种类为矿物族种类，确定矿物种类验证规则为分类验证规则；响应于初选矿物种类为包裹体矿物种类，或，强敏感矿物种类，或，蚀变矿物种类，确定矿物种类验证规则为选点重验规则；响应于初选矿物种类为普通常见矿物种类，或，与初选矿物种类对应的一级类别为非常见矿物类别，且初选矿物类别不为包裹体矿物种类、强敏感矿物种类与蚀变矿物种类中的任意一种，确定矿物种类验证规则为直接验证规则。在本申请实施例中，选点重验规则指示计算机设备向光谱数据采集设备重新获取扫描数据的规则；分类验证规则指示选取与初选矿物种类相对应的子数据库，进行二次校验的规则；直接验证规则指示将初选矿物种类的结果直接作为验证后的矿物种类结果的规则。
其中，当初选矿物种类为包裹体矿物种类，或，强敏感矿物种类，或蚀变矿物种类时，对应的验证规则即为选点重验规则，也即，由于组分区域中晶体的特殊性质，需要在组分区域内重新进行选点。该重新选点的次数与计算机设备中预存的验证次数对应。
在本申请实施例中，分类验证规则为基于矿物光谱数据数据库中的子数据库，对于矿物种类进行进一 步细分的规则。可选地，当初选矿物种类为矿物族种类时，基于矿物族数据库中的子数据库对于组分区域的矿物种类进行细分，并得到验证结果。也即，该分类验证规则适用于矿物族种类的矿物。
在本申请实施例中，对应初选矿物种类为普通常见矿物种类的情况，或，对应矿物种类的一级类别为非常见矿物类别，且细分矿物种类不为包裹体矿物种类、强敏感矿物种类与蚀变矿物种类中的任意一种时，确定直接验证规则。直接验证规则也即为将初选矿物类别确定为细分矿物类别的规则。
步骤1606，基于矿物种类验证规则以及初选矿物种类确定组分区域对应的矿物种类。
该过程即为确定与组分区域对应的矿物种类的过程。
步骤1607，基于组分图像对应的矿物种类，确定与岩石薄片对应的矿物特征。
该过程即为基于各个组分区域对应的矿物种类，最终生成与岩石薄片对应的矿物特征的过程。
综上所述，本申请实施例提供的方法，在获取与组分区域对应的光谱数据后，对应数据库进行与组分区域的矿物种类的初步确定，并对应确定验证规则，在结合验证规则的基础上，最终确定与组分区域对应的矿物种类。在进行与组分区域对应的矿物种类的确定过程中，在基于数据库确定其初选种类的基础上，对应每种初选种类，确定相关的验证规则，进一步得到其细分矿物种类，提高了矿物种类鉴定的准确率，进而提高了岩石薄片的矿物特征的确定效率以及准确率。
需要说明的是，图16所示的实施例，图13所示的实施例以及图9所示的实施例作为并列的三种矿物特征的识别方式，在岩石鉴定的流程中，可以进行任意的叠加组合，以达到对于矿物特征的识别并验证的效果，进一步提高对于岩石薄片的矿物特征的确定准确率。也即，图9所示的实施例、图13所示的实施例以及图6所示的实施例为在岩石鉴定的过程中，获取矿物特征的三个实施例。
需要说明的是，在本申请的一些实施例中，与岩石薄片对应的矿物特征以及几何特征可以被同时确定。图18示出了本申请一个示例性实施例提供的一种对于几何特征及矿物特征进行确定的过程示意图，请参考图18，该方法包括：
步骤1801，获取样本薄片图像。
在本申请实施例中，岩石薄片图像识别模型为基于机器学习的神经网络模型。该神经网络模型可以由如图6所示实施例中的岩石薄片图像分割模型以及如图9所示实施例中的组分种类识别模型组合得到，或，该神经网络模型可以根据岩石薄片图像分割模型以及组分种类识别模型的参数进行构建。对应训练该岩石薄片图像识别模型的样本为样本薄片图像，该样本薄片图像标注有样本分割结果以及样本组分类型特征。
步骤1802，将样本薄片图像输入岩石薄片图像分割识别模型中，输出得到预测分割区域以及预测组分类型特征。
该过程即为通过尚未训练完成的岩石薄片图像是被模型对于样本薄片图像进行预测，得到预测分割区域以及预测组分类型特征的过程。
步骤1803，将预测分割区域与样本分割区域进行比对，得到样本分割差异，并将预测组分类型特征与样本组分类型特征进行比对，得到样本组分差异。
本申请实施例中，由于同时输出预测分割区域以及预测组分类型特征，故同时对样本分割差异以及样本组分差异进行确定。
步骤1804，基于样本分割差异以及样本组分差异，对岩石薄片图像分割识别模型进行调整。
该过程即为对于岩石薄片图像分割识别模型进行调整的过程。
步骤1805，将岩石薄片图像输入岩石薄片图像分割识别模型，输出得到与岩石薄片图像对应的至少两个分割区域以及与分割区域对应的组分类型特征。
该过程即为将待测的岩石薄片图像输入岩石薄片图像分割识别模型中，并输出得分割区域以及与区域对应的组分类型特征的过程。
步骤1806，结合分割区域以及组分类型特征，确定与岩石薄片对应的几何特征以及矿物特征。
在本申请实施例中，分割区域与几何特征对应，组分类型特征与矿物特征对应。
综上所述，本申请实施例提供的方法，对应作为岩石薄片图像分割识别模型的输入量的样本薄片图像，并在模型的训练完成的前提下，对于岩石薄片图像的几何特征与矿物特征进行同步的提取，以进一步提高 岩石鉴定过程中的特征提取的效率。
可选地，在本申请中，图18所示的实施例为一个计算机设备同时获得几何特征和矿物特征的实施例。
通过上述各个实施例中所述的方式，进行矿物特征以及几何特征的确定之后，在本申请实施例中，即通过几何特征、矿物特征与结构特征，生成对应岩石薄片的鉴定结果，也即，对岩石样本的特征的文字化描述。图19示出了本申请一个示例性实施例提供的一种岩石薄片的鉴定结果的生成方法的示意图，该方法可以替换实现为如图3所示的步骤303，以该方法应用于计算机设备中为例进行说明，该方法包括：
步骤1901，基于初选岩类特征与细分岩类特征，确定与鉴定结果对应的至少一个岩石薄片子鉴定结果的种类。
在本申请实施例中，岩石薄片子鉴定结果即为基于组分区域的形状，以及组分区域的矿物性质得到的，对于岩石薄片的具体种类具有表征功能的鉴定结果。可选地，岩石薄片鉴定结果包括粒径区间识别结果、最大颗粒粒径识别结果、颗粒分选性结果、颗粒磨圆度识别结果、矿物自形程度结果以及颗粒接触方式识别结果中的至少一种。
步骤1902，基于几何特征与矿物特征，生成岩石薄片子鉴定结果。
可选地，对应上述粒径区间识别结果、最大颗粒粒径识别结果、颗粒分选性结果、颗粒磨圆度识别结果、矿物自形程度结果以及颗粒接触方式识别结果的确定方式进行举例，以说明岩石薄片子鉴定结果的生成过程。
在一个示例中，岩石薄片鉴定结果中包括粒径区间识别结果，粒径区间识别结果指示岩石薄片中，与一类组分对应的组分区域的粒径的区间。则首先，基于矿物特征确定与粒径区间识别结果对应的粒径区间组分区域集合，粒径区间组分区域集合中包括至少两个组分区域，其次，基于粒径区间组分区域集合中的组分区域对应的几何特征，确定与粒径区间组分区域中的组分区域对应的粒径，最后，基于组分区域对应的粒径，确定粒径区间识别结果。
在一个示例中，岩石薄片鉴定结果中包括最大颗粒粒径识别结果，最大颗粒粒径识别结果指示岩石薄片中，与一类组分对应的组分区域中粒径最大的组分区域的粒径。在此情况下，基于矿物特征确定与最大颗粒粒径识别结果对应的最大颗粒组分区域，并基于几何特征确定与组分区域对应的最大颗粒粒径，并基于最大颗粒粒径生成最大颗粒粒径识别结果。
在一个示例中，岩石薄片鉴定结果中包括颗粒分选性结果，颗粒分选性结果指示岩石薄片中与一类颗粒对应的颗粒粒径的分布情况。在次情况下，首先，基于矿物特征确定与颗粒分选性识别结果对应的分选性组分区域集合，分选性组分区域集合中包括至少两个组分区域；其次，基于分选性组分区域集合中的组分区域对应的几何特征，确定颗粒粒径分级对应关系，颗粒粒径分级对应关系中包括至少两级颗粒粒径级别，以及处于颗粒粒径级别中的组分区域的数量；最终，基于颗粒粒径分级对应关系，确定颗粒分选性结果。
在一个示例中，岩石薄片鉴定结果中包括颗粒磨圆度识别结果，该颗粒磨圆度识别结果即指示组分区域的磨圆度，且磨圆度对应有磨圆度级别。在此情况下，首先，基于矿物特征确定与颗粒磨圆度识别结果对应的磨圆度识别组分区域集合，磨圆度识别组分区域集合中包括至少两个组分区域；其次，基于磨圆度组分识别区域集合中的组分区域对应的几何特征，确定磨圆度分级对应关系，磨圆度分级对应关系中包括至少两个磨圆度级别，以及处于磨圆度级别中的组分区域的数量；最后，基于磨圆度分级对应关系，确定颗粒磨圆度识别结果。
在一个示例中，岩石薄片鉴定结果中包括矿物自形程度结果，矿物自形程度结果指示组分区域矿物基于自身结晶习性发育的程度。在此情况下，首先，基于矿物特征确定与矿物自形程度结果对应的自形程度组分区域集合，自形程度组分区域集合中包括至少两个自形程度组分区域；其次，基于自形程度组分区域集合中的组分区域对应的几何特征，确定组分区域的形状；最终，基于组分区域的形状，确定矿物自形程度结果。
在一个示例中，岩石薄片鉴定结果中包括颗粒接触方式识别结果。颗粒接触方式识别结果指示相邻两个组分区域的接触状态。在此情况下，首先，基于矿物特征确定至少两组组分区域对，组分区域对中包括 两个相邻的组分区域；其次，基于几何特征，确定组分区域对中的两个组分区域的交并比；最后，基于至少两个组分区域的交并比，确定颗粒接触方式识别结果。
对上述过程进行总结，即在确定了岩石薄片子鉴定结果的种类后，基于矿物种类对于岩石薄片图像中，与该岩石薄片子鉴定结果对应的组分区域进行确定，并基于几何特征，从组分区域的形状、形态、大小等方面确定岩石薄片子鉴定结果。
步骤1903，基于细分岩类特征，确定与岩石薄片名称对应的命名规则。
在本申请实施例中，不同岩石种类对应有不同的命名规则，故需要基于细分岩类特征，进行岩石薄片名称的确定。
步骤1904，基于命名规则以及岩石薄片子鉴定结果，生成岩石薄片名称。
该过程即为最终确定岩石薄片名称的过程。可选地，计算机设备可以根据岩石薄片子鉴定结果生成对应的关键词，并结合细分岩类特征，最终确定岩石薄片对应的名称。
可选地，对于岩石薄片的鉴定结果，本申请的一些实施例中，还会生成岩石样本鉴定报告。在该岩石样本鉴定报告中，包括了岩石薄片名称以及各个岩石薄片子鉴定结果。在本申请实施例中，计算机设备根据不同的细分岩类特征，预存了至少两种岩石样本鉴定报告。
综上所述，本申请实施例提供的方法，在获取初选岩类特征与细分岩类特征后，进行与岩石种类对应的岩石薄片鉴定子结果的选择，并且对应选择得到的岩石薄片鉴定子结果，使最终生成的岩石薄片鉴定结果的内容与岩石薄片的种类对应，进一步提高岩石薄片鉴定结果的准确率。
如上实施例中所述，图19对应的实施例即说明了通过几何特征、矿物特征与结构特征获得岩石薄片鉴定子结果，进而获得最终的岩石薄片鉴定结果的过程。
在一个示例中，对应碎屑岩薄片鉴定过程，图20示出了本申请一个示例性实施例提供的一种岩石鉴定方法的过程示意图，该过程包括：
步骤2001，采集碎屑岩薄片图像。
在本申请实施例中，碎屑岩薄片样本图像包括了单偏光图像以及正交偏光图像。可选地，在图像采集设备下连续采集碎屑岩薄片样品的图像，再进行拼接，最终得到与碎屑岩薄片样本对应的单偏光图像以及正交偏光图像。
步骤2002，确定碎屑岩薄片图像的初选岩石种类。
在本申请实施例中，通过岩石分类模型，确定碎屑岩薄片图像的初选岩石种类为沉积岩。
步骤2003，确定碎屑岩薄片的细分岩石种类。
在本申请实施例中，通过岩石结构分类模型，确定碎屑岩薄片的细分岩石种类为碎屑岩，且具有砂状结构。
步骤2004，将碎屑岩薄片图像输入碎屑岩分割模型，获得分割后的碎屑岩薄片图像。
该过程即为通过岩石薄片分割模型，对碎屑岩薄片图像进行分割，得到分割图像的过程。在进行分割后，计算机设备即可对于碎屑岩薄片图像对应的碎屑岩几何特征进行确定。
步骤2005，根据分割后的碎屑岩薄片图像进行组分区域的提取，并通过组分种类识别模型获取组分类型特征。
该过程即为获取与碎屑岩薄片图像对应的矿物特征的过程。
步骤2006，将组分类型特征在碎屑岩薄片图像上进行映射，得到碎屑岩薄片样本中间图像。
该过程即为将基于组分类型特征和分割图像在碎屑岩薄片图像上进行标注，得到可以表征碎屑岩薄片的几何特征和矿物特征的中间图像。
在本申请的其他实施例中，组分种类识别模型将直接输出该碎屑岩薄片样本中间图像。
步骤2007，在碎屑岩薄片样本中间图像上，统计相应组分区域的面积。
在本申请实施例中，基于对于各个组分区域的矿物种类识别结果，通过统计组分区域面积的方式，确定不同组分的相对含量。在一个示例中，基于组分的矿物种类进行划分，得到硅质胶结物占比2.0％，钙质胶结物占比10.5％，杂基填隙物占比1％，岩屑碎屑占比13.84％，长石碎屑占比20.76％，石英碎屑占比51.9％。 由于石英碎屑、长石碎屑与岩屑碎屑均属于陆源碎屑，硅质、钙质胶结物、杂基填隙物均属于填隙物，故可进一步确定陆源碎屑占比86.5％，填隙物占比13.5％，即可确定该碎屑岩薄片图像中不包含非陆源碎屑占比，颗粒与填隙物的相对含量为颗粒占比86.5％，填隙物占比13.5％。
步骤2008，对石英碎屑区域、长石碎屑区域、硅质胶结物区域和钙质胶结物区域对应的子区域图像分别计算，并进行干净区域筛选。
由于石英碎屑区域、长石碎屑区域、硅质胶结物区域和钙质胶结物区域对应的组分特征为单晶态，故需要进行可供进行扫描的干净区域的筛选。可选地，该过程通过基于机器学习的模型实现。
步骤2009，确定面扫描坐标图。
该过程即为确定具体扫描区域位置的过程，在本申请实施例中，扫描方式为面扫描。
步骤2010，对面扫描坐标进行计算，规划最优路径，生成扫描指令。
可选地，该过程即为计算机设备确定扫描指令的过程，由于需要面扫描的区域在实际情况中并不相邻，故需要进行最优路径的规划。
步骤2011，通过扫描指令控制矿物数据采集设备采集扫描数据。
该过程即为将扫描数据发送至矿物数据采集设备，使矿物数据采集设备生成对应的扫描数据的过程，可选地，扫描数据为光谱数据。
步骤2012，基于扫描数据对于生成矿物信息。
该过程即为基于扫描数据生成基于矿物数据采集设备发送的扫描数据，确定由扫描数据表征的，组分区域的矿物成分的过程。
步骤2013，对比矿物信息与分类识别结果，在每个组分区域上获得最终结果。
该过程即为将矿物成分与扫描数据进行比对，最终获得结果的过程。可选地，在该过程中，根据矿物数据采集设备的设备种类，进行最终的验证结果的确定。
步骤2014，基于最终结果对碎屑岩薄片样本中间图像上对应组分区域的识别结果进行替换。
该过程即为通过面扫描的验证结果，对于碎屑岩薄片样本中间图像上的，通过机器学习模型得到的组分区域矿物识别结果进行替换的过程。由于光谱数据对于矿物成分的鉴定更加准确，故本步骤中，以面扫描的验证结果为准确定组分区域的矿物特征。
步骤2015，确定更新后的碎屑岩薄片样本最终图像。
结合步骤2007中的示例，在更新后，硅质胶结物的占比为2.0％，钙质胶结物对应的实际形态为方解石，占比为10.5％，杂基填隙物占比为1％，岩屑碎屑占比为13.84％，石英碎屑占比69.2％，长石碎屑占比3.46％，也即，陆源碎屑占比仍为86.5％，非陆源碎屑占比仍为0％。对应地，颗粒占比86.5％，填隙物占比13.5％
步骤2016，确定与碎屑岩薄片对应的薄片子鉴定结果。
在本申请实施例中，薄片子鉴定结果包括粒径区间识别结果、颗粒分选性结果、最大颗粒粒径识别结果、相对组分含量结果、颗粒磨圆度识别结果以及颗粒接触方式识别结果。经过基于矿物特征的组分区域选取以及基于几何特征的结果生成，可以确定，在一个示例中，颗粒的主要粒径区间为，细砂(0.125-0.25mm)占比10％，中砂(0.25-0.5mm)占比30％，粗砂(0.5-2mm)占比60％；最大颗粒粒径为1.8mm；颗粒分选性为中；85％的组分区域的颗粒磨圆度为次棱；15％的组分区域的颗粒磨圆度为次圆；75％的组分区域之间为线接触，25％的组分区域之间为点接触。
步骤2017，基于薄片子鉴定结果确定薄片名称以及薄片鉴定报告。
请参考图21，在薄片鉴定报告2100中，包括碎屑岩定名区域2110，碎屑岩定名区域中包括基于薄片子鉴定结果对于碎屑岩岩石薄片进行的定名。本申请实施例中，碎屑岩的定名为“中粗粒岩屑石英砂岩”。可选地，该定名的依据为计算机设备中存储的行业标准，也即行业标准SY/T 5368-2016《岩石薄片鉴定》。在薄片鉴定报告2100中，还包括指示碎屑岩样本获取位置以及获取方式的薄片样品信息区域2120、指示碎屑岩样本的矿物构成的组分特征区域2130、指示碎屑岩样本的结构特征的结构特征区域2140以及岩石薄片图像显示区域2150。
综上所述，本申请实施例提供的方法，对应岩石薄片样本为碎屑岩的情况，在基于结构特征的识别， 确定种类为碎屑岩后，通过基于机器学习的图像分割以及图像识别，并通过光谱数据的获取对于图像识别得到的识别结果进行验证，最终生成鉴定结果的方法，将光谱数据中表征的组分区域的化学特征以及图像中表征的组分区域的图像特征相结合，对于碎屑岩薄片进行鉴定，提高了对于碎屑岩薄片进行鉴定的效率以及准确率。
如上实施例中所述，图20即为对碎屑岩这一岩类，通过几何特征、矿物特征与结构特征的确定，并基于三类特征，确定碎屑岩岩石薄片鉴定子结果，最终得到碎屑岩岩石薄片鉴定结果的过程。
在一个示例中，针对岩浆岩薄片鉴定过程，图22示出了本申请一个示例性实施例提供的一种岩石鉴定方法的过程示意图，该过程包括：
步骤2201，获取岩浆岩薄片样本图像。
图23示出了本申请一个示例性提供的一种岩浆岩薄片图像的示意图。该岩浆岩薄片2300中包括至少一个橄榄石组分区域2301、普通辉石组分区域2302、基性斜长石组分区域2303、紫苏辉石组分区域2304和磁铁矿组分区域2305。
在本申请实施例中，与图20所示的实施例相似，岩浆岩薄片图像也包括与岩浆岩薄片样本对应的单偏光图像以及正交偏光图像。
步骤2202，确定与岩浆岩薄片图像对应的初选岩石种类。
在本申请实施例中，根据岩浆岩薄片图像的表现形式，确定其为岩浆岩，据此，几何特征与矿物特征也即与岩浆岩对应的特征。
在本申请的其他实施例中，可以通过模型识别的方式，在确定岩石薄片图像为岩浆岩薄片图像的情况下，进一步确定该岩浆岩的结构的细分结构为辉长结构。也即，在确定初选岩石种类后，存在直接进一步确定细分岩石种类的情况。据此，几何特征与矿物特征也即与具有辉长结构的岩浆岩对应的特征。
步骤2203，将岩浆岩薄片样本图像输入岩浆岩分割模型，获得分割后的岩浆岩薄片图像。
该过程即为对岩浆岩薄片图像进行分割，得到分割结果的过程。
步骤2204，根据分割后的岩浆岩薄片图像进行组分区域的提取，并通过组分种类识别模型获取组分类型特征。
该过程即为基于分割结果对于组分区域进行提取，并进一步识别组分区域对应的矿物组成的过程。
步骤2205，将分割后的岩浆岩薄片图像中对应的组分区域进行干净区域筛选，确定面扫描区域。
在本申请实施例中，对应每一个组分区域，均需要通过获取扫描数据的方式，进行与组分区域对应的矿物组成的验证，故步骤2205即为通过干净区域筛选，确定面扫描区域的过程。
步骤2206，对面扫描坐标进行计算，规划最优路径，生成扫描指令。
该过程即为对应面扫描区域不相邻的情况，进行最优路径的规划，以生成发送至矿物数据采集设备的扫描指令的过程。
步骤2207，通过扫描指令控制矿物数据采集设备采集扫描数据。
该过程即为计算机设备控制矿物数据采集设备进行扫描数据的采集过程。
步骤2208，基于扫描数据生成矿物信息。
该过程即为从矿物数据采集设备中接收扫描数据，也即光谱数据，并进而根据光谱数据确定与组分区域对应的矿物种类的过程
步骤2209，对比矿物信息与分类识别结果，在每个组分上获得最终结果。
该过程即为通过验证，最终确定与各个组分区域相对应的矿物种类的过程。
步骤2210，确定岩浆岩薄片中包含的矿物类型。
在本申请实施例中，经过对于各个组分区域相对应的矿物种类的确定，即可确定与岩浆岩薄片中包含的矿物类型。在一个示例中，该岩浆岩薄片包含矿物种类有斜长石、普通辉石、紫苏辉石、橄榄石、黑云母、磁铁矿以及磷灰石。
步骤2211，基于包含的矿物类型，确定岩浆岩薄片对应的细分岩石种类。
基于步骤2210中获取的矿物种类，即可确定与该岩浆岩薄片对应的细分岩石种类为斜长石，亚族为 拉长石，属于基性斜长石。
需要说明的是，在本申请实施例中，针对岩浆岩这一岩种，与岩浆岩的细分结构特征的数量为至少两个。在一个示例中，除了指示岩石具体种类的结构特征之外，与岩浆岩对应的细分结构特征还包括反应边结构特征，该细分结构特征用于指示岩浆岩中包含的反应边类型。
步骤2212，基于岩石细分种类，确定与岩浆岩薄片对应的薄片子鉴定结果。
在本申请实施例的一个示例中，基于岩石细分种类，确定薄片子鉴定结果包括粒径识别结果、粒径形状结果、矿物自形程度结果、颗粒交互关系结果以及矿物含量结果。基于岩浆岩薄片图像指示的矿物特征，在该岩浆岩薄片中，斜长石含量为52.4％，普通辉石含量为34.9％，紫苏辉石含量为3.2％，橄榄石含量为7.6％，黑云母含量为1.6％，磁铁矿含量为0.4％，磷灰石含量为0.1％。
基于橄榄石与辉石对应的颗粒，以及辉石与黑云母对应的颗粒之间的交互关系，计算机设备可以确定橄榄石外边具有紫苏辉石镶边，具有反应边结构，进而指示该岩浆岩为基性侵入岩或超基性侵入岩。
此外，通过对于组分区域的粒径分析可得，斜长石粒径为(1.6-2.1)x(1.8-3.5)mm，普通辉石粒径为1.2-2.2mm，紫苏辉石粒径为1.1-1.7mm，橄榄石粒径为0.4-2.5mm，黑云母粒径为1.2-2.2mm，磁铁矿粒径为0.05-0.2mm，磷灰石粒径为0.05-0.1mm。经过统计，即可粒径与形状对应关系的判断，计算机设备即可确定，斜长石为粒状-短板状，普通辉石为不规则粒状及短柱状，紫苏辉石为粒状，橄榄石为粒状，黑云母为板片状，磁铁矿为粒状，磷灰石为短柱状。且与自形程度进行对应分析后，可确定斜长石为半自形-他形，普通辉石为他形-半自形，紫苏辉石为他形，橄榄石为他形，黑云母为自形-半自形，磁铁矿为自形，磷灰石为半自形-自形。
需要说明的是，针对岩浆岩这一岩种，确定薄片子鉴定结果的确定以及生成过程，即可视为对于结构特征的验证过程。在如上所述的实施例中，通过计算橄榄石与辉石，或，辉石与黑云母之间的相互关系，计算机设备即可确定岩浆岩薄片中具有反应边结构，与岩浆岩薄片对应的薄片鉴定子结果中包括指示其存在反应边的反应边结构。而在前叙的步骤中，在确定细分岩石种类的过程中，已确定岩石薄片中可能具有反应边结构，则此处的反应边结构确定过程即为对于细分岩石种类的验证过程。
步骤2213，基于薄片子鉴定结果确定薄片名称以及薄片鉴定报告。
如上实施例中所述，在通过薄片子鉴定结果确定薄片名称以及薄片鉴定报告的过程中，可以提取以下对于岩浆岩进行命名的要素：
要素1：在岩浆岩薄片图像中，与斜长石对应的组分区域指示斜长石的粒径属于中粒。
要素2：在岩浆岩薄片图像中，矿物特征指示其中的主要矿物为斜长石和辉石。
要素3：在主要矿物为斜长石和辉石的情况下，基于几何特征，确定斜长石和辉石均呈现为粒装。
要素4：在主要矿物为斜长石和辉石的情况下，基于几何特征，确定斜长石和辉石自形程度相近，均呈现半自形-他形。
基于上述要素，确定岩浆岩薄片的命名存在橄榄辉长岩的可能性。
在本申请实施例中，在计算机设备确定命名存在橄榄辉长岩的可能性后，可以通过如下示例中的方法，对于岩浆岩薄片的命名进行验证。需要说明的是，如下验证过程是在计算机设备内存储有相关的国家标准数据以及行业标准数据的情况下进行的。
示例方法：基于矿物特征以及几何特征的组合，通过组分区域的面积指代含量的方式，同时，定义Q为石英、鳞石英以及方石英的含量之和，A为碱性长石(即正长石、微斜长石、条纹长石、歪长石、透长石和钠长石的集合)的含量；P为斜长石和方柱石的含量之和；F为副长石类和副长石(即霞石、白榴石、钾霞石、假白榴石、方钠石、黝方石、蓝方石、钙霞石和方沸石的集合)的含量之和；M为镁铁矿物以及有关矿物的含量之和。由于在岩浆岩中，Q组矿物和F组矿物相互排斥，或Q存在，则F必缺失，反之亦然。在对应岩浆岩为超基性侵入岩或基性侵入岩，且Q＝0，A＝0，P＝52.4，F＝0，M＝47.6的情况下，请参考图24，在侵入岩分类相图2400上，根据P的相对含量为100％，即可确定与该岩浆岩薄片对应的分类区域为分类区域2401，该分类区域2401指示岩浆岩薄片对应的岩石名称可能包括闪长岩、辉长岩以及斜长岩中的至少一种。在此情况下，即可对于岩浆岩薄片的类型为辉长岩的情况进行验证。
可选地，在本申请的其他实施例中，验证的方式还包括通过细分结构特征所指示的辉长结构进行验证 等方法，本申请实施例在此不做赘述。
在此情况下，由于矿物特征指示岩浆岩薄片中的橄榄石含量为7.6％，大于5％，故在命名时需要体现该橄榄石特征，对应地，即可经过验证，最终确定对该岩浆岩薄片的命名为“橄榄辉长岩”。
需要说明的是，上述命名依据标准为计算机设备中存储的行业标准和国家标准，即行业标准SY/T 5368-2016《岩石薄片鉴定》，以及国家标准GB/T 17412.1-1998《岩石分类和命名方案——火成岩岩石分类和命名方案》。可选地，本申请实施例还提供有与岩浆岩薄片对应的薄片鉴定报告。
综上所述，本申请实施例提供的方法，对应岩石薄片样本为岩浆岩的情况，在基于结构特征的识别，确定种类为岩浆岩后，通过基于机器学习的图像分割以及图像识别，并通过光谱数据的获取对于图像识别得到的识别结果进行验证，最终生成鉴定结果的方法，将光谱数据中表征的组分区域的化学特征以及图像中表征的组分区域的图像特征相结合，确定岩浆岩的细分种类，并进而对于岩浆岩薄片所表征的结构特征与组分特征进行进一步鉴定，提高了对于碎屑岩薄片进行鉴定的效率以及准确率。
如上实施例中所述，图22为岩浆岩鉴定方法实施例，通过几何特征、矿物特征和结构特征获得岩浆岩岩石薄片鉴定子结果，进而获得最终鉴定结果的过程。
图25示出了本申请一个示例性实施例提供的一种岩石鉴定装置的结构框图，该装置包括：
接收模块2501，用于接收图像采集设备发送的岩石薄片图像，岩石薄片图像为对岩石薄片进行拍摄得到的图像，岩石薄片为对岩石样本进行切割得到的薄片，岩石薄片图像中包括至少一个组分区域；
生成模块2502，用于基于岩石薄片图像生成与岩石薄片对应的几何特征、矿物特征以及结构特征，其中，几何特征用于指示岩石薄片基于组分的区域划分情况，矿物特征用于指示岩石薄片中的矿物分布情况，结构特征用于指示岩石样本的空间结构；
基于几何特征、矿物特征以及结构特征，生成岩石薄片的鉴定结果，鉴定结果中包括对岩石样本的特征的文字化描述。
在一个可选的实施例中，结构特征包括初选岩类结构特征，初选岩类结构特征用于指示岩石样本对应的初选岩石种类，初选岩石种类包括沉积岩类、岩浆岩类和变质岩类；
生成模块2502，还用于基于岩石薄片图像生成初选岩类结构特征；
请参考图26，该装置，还包括确定模块2503，用于基于初选岩类结构特征确定与岩石薄片对应的初选岩石种类；
生成模块2502，还用于根据初选岩石种类，基于岩石薄片图像生成几何特征；
根据初选岩石种类，基于岩石薄片图像生成矿物特征。
在一个可选的实施例中，结构特征还包括细分岩类结构特征，细分岩类结构特征用于指示岩石样本对应的细分岩石种类；
确定模块2503，还用于基于几何特征与矿物特征，确定细分岩类结构特征。
在一个可选的实施例中，岩石薄片图像中包括至少两个表征组分区域的组分图像；
该装置，还包括划分模块2504，用于基于组分图像的位置对岩石薄片图像进行划分，得到岩石薄片分割图像，岩石薄片分割图像中包括至少两个分割区域，分割区域用于指示岩石薄片图像基于组分区域的分割情况；
确定模块2503，还用于基于分割区域确定与岩石薄片对应的几何特征；
基于分割区域确定与岩石薄片对应的矿物特征。
在一个可选的实施例中，该装置，还包括输入模块2505，用于将岩石薄片图像输入岩石薄片图像分割模型中，输出得到岩石薄片分割图像，岩石薄片图像分割模型为基于机器学习的Mask-RCNN网络模型。
在一个可选的实施例中，确定模块2503，还用于基于分割区域确定组分图像，组分图像中包括一个组分区域；
输入模块2505，还用于将组分图像输入组分种类识别模型，输出得到组分图像的组分类型特征，组分类型特征用于对与组分区域的矿物特征进行表征，组分种类识别模型为基于机器学习的神经网络模型；
确定模块2503，还用于基于组分图像对应的组分类型特征，确定与岩石薄片对应的矿物特征。
在一个可选的实施例中，岩石鉴定系统中还包括矿物数据采集设备，矿物数据采集设备与计算机设备连接；
输入模块2505，还用于将岩石薄片分割图像输入分类识别模型中，输出得到单晶态组分集合图像和非单晶态组分集合图像，单晶态组分集合图像中包括至少一个单晶态组分区域，以及与单晶态组分区域对应的单晶态组分识别结果，非单晶态组分区域中包括至少一个非单晶态组分区域，以及与非单晶态组分区域对应的非单晶态组分识别结果，分类识别模型为基于深度学习的神经网络模型；
该装置，还包括发送模块2506，用于基于单晶态组分集合图像向矿物数据采集设备发送扫描指令；
接收模块2501，还用于接收矿物数据采集设备基于扫描指令反馈的扫描数据；
该装置，还包括验证模块2507，用于基于扫描数据对单晶态组分识别结果进行验证，得到验证结果；
确定模块2503，还用于基于非单晶态组分识别结果、单晶态组分识别结果以及验证结果，确定与岩石薄片对应的矿物特征。
在一个可选的实施例中，输入模块2505，还用于将单晶态组分集合图像输入区域筛选模型，输出得到区域筛选结果，区域筛选结果用于指示单晶态组分集合图像中用于进行面扫描数据的部分；
生成模块2502，还用于基于区域筛选结果生成扫描指令；
发送模块2506，还用于向矿物数据采集设备发送扫描指令。
在一个可选的实施例中，确定模块2503，还用于基于扫描数据确定扫描识别结果；
响应于扫描识别结果与单晶态组分识别结果相同，确定验证结果指示验证通过，并确定单晶态组分识别结果不变；
响应于面扫描识别结果与单晶态组分识别结果不同，确定验证结果指示验证不通过，并基于识别结果生成规则确定单晶态组分识别结果，识别结果生成规则中包括基于矿物采集设备种类的规则。
在一个可选的实施例中，确定模块2503，还用于基于分割区域确定组分图像，组分图像中包括一个组分区域；
发送模块2506，还用于基于组分图像向矿物数据采集设备发送光谱数据获取指令；
接收模块2501，还用于接收矿物数据采集设备基于光谱数据获取指令发送的光谱数据；
确定模块2503，还用于基于光谱数据，在矿物光谱数据数据库中确定与组分区域对应的初选矿物种类；
确定与初选矿物种类对应的矿物种类验证规则；
基于矿物种类验证规则以及初选矿物种类确定组分区域对应的矿物种类；
基于组分图像对应的矿物种类，确定与岩石薄片对应的矿物特征。
在一个可选的实施例中，矿物光谱数据数据库包括矿物种类子数据库；
确定模块2503，还用于基于光谱数据，在矿物光谱数据数据库中确定与组分区域对应的初选矿物种类的一级类别，矿物种类的一级类别为基于矿物的常见程度划分的类别，一级类别包括常见矿物类别与非常见矿物类别，与常见矿物类别对应的初选矿物种类包括矿物族种类以及普通常见矿物种类，与非常见矿物类别对应的初选矿物种类包括包裹体矿物种类、强敏感矿物种类以及蚀变矿物种类；
基于矿物种类一级类别对应的矿物种类子数据库，确定与组分区域对应的初选矿物种类。
在一个可选的实施例中，矿物种类验证规则包括分类验证规则、直接验证规则以及选点重验规则；
确定模块2503，还用于响应于初选矿物种类为矿物族种类，确定矿物种类验证规则为分类验证规则；
响应于初选矿物种类为包裹体矿物种类，或，强敏感矿物种类，或，蚀变矿物种类，确定矿物种类验证规则为选点重验规则；
响应于初选矿物种类为普通常见矿物种类，或，与初选矿物种类对应的一级类别为非常见矿物类别，且初选矿物类别不为包裹体矿物种类、强敏感矿物种类与蚀变矿物种类中的任意一种，确定矿物种类验证规则为直接验证规则。
在一个可选的实施例中，岩石薄片图像中包括至少两个表征组分区域的组分图像；
输入模块2505，还用于将岩石薄片图像输入岩石薄片图像分割识别模型，输出得到与岩石薄片图像对应的至少两个分割区域以及与分割区域对应的组分类型特征，分割区域用于指示岩石薄片图像基于组分区域的分割情况，组分类型特征用于对分割区域中的组分区域的矿物特征进行表征，岩石薄片图像分割识别 模型为基于机器学习的神经网络模型；
确定模块2503，还用于结合分割区域以及组分类型特征，确定与岩石薄片对应的几何特征以及矿物特征。
在一个可选的实施例中，输入模块2505，还用于将岩石薄片图像输入初选岩类选择模型中，输出得到初选岩类结果，初选岩类结果指示岩石薄片对应的初选岩类结构特征，初选岩类结构选择模型为基于初选岩类图像样本集构建的模型。
在一个可选的实施例中，鉴定结果中包括细分岩石种类，输入模块2505，还用于将几何特征与矿物特征输入细分岩类选择模型中，输出得到细分岩类结果，细分岩类结果指示岩石薄片对应的细分岩石结构特征，细分岩类选择模型为基于几何-矿物特征交互样本集构建的模型，几何-矿物特征交互样本集指示几何特征与矿物特征的组合，与细分岩类结构特征之间的对应关系。
在一个可选的实施例中，鉴定结果中包括岩石薄片子鉴定结果以及岩石薄片名称中的至少一种；
确定模块2503，还用于基于初选岩类特征与细分岩类特征，确定与鉴定结果对应的至少一个岩石薄片子鉴定结果的种类，岩石薄片鉴定结果包括粒径区间识别结果、最大颗粒粒径识别结果、颗粒分选性结果、颗粒磨圆度识别结果、矿物自形程度结果以及颗粒接触方式识别结果中的至少一种；
生成模块2502，还用于基于几何特征与矿物特征，生成岩石薄片子鉴定结果；
确定模块2503，还用于基于岩石薄片鉴定结果确定岩石薄片名称。
综上所述，本申请实施例提供的装置，在获取岩石薄片图像后，对于岩石薄片图像进行基于几何特征、矿物特征以及结构特征的三个维度的特征提取，从微观组成和宏观表现的角度，综合多个特征维度对于岩石的性质进行确定，最终生成包括有文字化描述的鉴定结果。在对于岩石进行鉴定的过程中，在获取岩石薄片对应的微观可视化图像后，对该图像进行多个维度的特征提取，并在参考多个维度的特征的情况下对于岩石薄片进行鉴定，提高了岩石鉴定的准确率。
需要说明的是：上述实施例提供的岩石鉴定装置，仅以上述各功能模块的划分进行举例说明，实际应用中，可以根据需要而将上述功能分配由不同的功能模块完成，即将设备的内部结构划分成不同的功能模块，以完成以上描述的全部或者部分功能。
图27示出了本申请一个示例性实施例提供的一种执行岩石鉴定方法的计算机设备的结构示意图，该计算机设备包括：
处理器2701包括一个或者一个以上处理核心，处理器2701通过运行软件程序以及模块，从而执行各种功能应用以及数据处理。
接收器2702和发射器2703可以实现为一个通信组件，该通信组件可以是一块通信芯片。可选地，该通信组件可以实现包括信号传输功能。也即，发射器2703可以用于发射控制信号至图像采集设备以及扫描设备中，接收器2702可以用于接收对应的反馈指令。
存储器2704通过总线2705与处理器2701相连。
存储器2704可用于存储至少一个指令，处理器2701用于执行该至少一个指令，以实现上述方法实施例中的各个步骤。
本申请实施例还提供一种计算机可读存储介质，该可读存储介质中存储有至少一条指令、至少一段程序、代码集或指令集，以由处理器加载并执行以实现上述岩石鉴定方法。
本申请还提供了一种计算机程序产品或计算机程序，该计算机程序产品或计算机程序包括计算机指令，该计算机指令存储在计算机可读存储介质中。计算机设备的处理器从计算机可读存储介质读取该计算机指令，处理器执行该计算机指令，使得该计算机设备执行上述实施例中任一所述的岩石鉴定方法。
可选地，该计算机可读存储介质可以包括：只读存储器(ROM，Read Only Memory)、随机存取记忆体(RAM，Random Access Memory)、固态硬盘(SSD，Solid State Drives)或光盘等。其中，随机存取记忆体可以包括电阻式随机存取记忆体(ReRAM，Resistance Random Access Memory)和动态随机存取存储 器(DRAM，Dynamic Random Access Memory)。上述本申请实施例序号仅仅为了描述，不代表实施例的优劣。
本领域普通技术人员可以理解实现上述实施例的全部或部分步骤可以通过硬件来完成，也可以通过程序来指令相关的硬件完成，的程序可以存储于一种计算机可读存储介质中，上述提到的存储介质可以是只读存储器，磁盘或光盘等。
上述仅为本申请的可选实施例，并不用以限制本申请，凡在本申请的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本申请的保护范围之内。

Claims (23)

  1. 一种岩石鉴定方法，其特征在于，所述方法包括：
    接收岩石薄片图像，所述岩石薄片图像为对岩石薄片进行拍摄得到的图像，所述岩石薄片为对岩石样本进行切割得到的薄片，所述岩石薄片图像包括至少一个组分区域；
    基于所述岩石薄片图像生成与所述岩石薄片对应的几何特征、矿物特征以及结构特征，其中，所述几何特征用于指示所述岩石薄片的组分区域划分情况，所述矿物特征用于指示对所述岩石薄片中与所述组分区域对应的矿物种类分布情况，得到的矿物鉴定结果，所述结构特征用于指示所述岩石样本的岩石种类；
    基于所述几何特征、所述矿物特征以及所述结构特征，生成所述岩石薄片的鉴定结果。
  2. 根据权利要求1所述的方法，其特征在于，所述结构特征包括初选岩类结构特征，其中，在基于所述岩石薄片图像生成与所述岩石薄片对应的几何特征、矿物特征以及结构特征步骤中，包括：
    基于所述岩石薄片图像生成所述初选岩类结构特征，所述初选岩类结构特征用于指示所述岩石样本对应的初选岩石种类，所述初选岩石种类包括沉积岩类、岩浆岩类和变质岩类；
    基于所述初选岩类结构特征确定与所述岩石薄片对应的所述初选岩石种类；
    根据所述初选岩石种类，基于所述岩石薄片图像生成所述几何特征；
    根据所述初选岩石种类，基于所述岩石薄片图像生成所述矿物特征。
  3. 根据权利要求2所述的方法，其特征在于，所述结构特征还包括细分岩类结构特征，其中，在根据所述初选岩石种类，基于所述岩石薄片图像生成所述矿物特征之后，还包括：
    基于所述几何特征与所述矿物特征，确定所述细分岩类结构特征，所述细分岩类结构特征用于指示所述岩石样本对应的细分岩石种类。
  4. 根据权利要求1至3中任一项所述的方法，其特征在于，所述岩石薄片图像包括至少两个表征所述组分区域的组分图像，其中，在基于所述岩石薄片图像生成与所述岩石薄片对应的几何特征、矿物特征以及结构特征步骤中，包括：
    基于组分图像的位置对所述岩石薄片图像进行划分，得到岩石薄片分割图像，所述岩石薄片分割图像中包括至少两个分割区域，所述分割区域用于指示所述岩石薄片图像基于所述组分区域的分割情况；
    基于所述分割区域确定与所述岩石薄片对应的几何特征；
    基于所述分割区域确定与所述岩石薄片对应的矿物特征。
  5. 根据权利要求4所述的方法，其特征在于，在基于组分图像的位置对所述岩石薄片图像进行划分，得到岩石薄片分割图像步骤中，包括：
    将所述岩石薄片图像输入岩石薄片图像分割模型中，输出得到所述岩石薄片分割图像，所述岩石薄片图像分割模型为基于机器学习的Mask-RCNN网络模型。
  6. 根据权利要求4所述的方法，其特征在于，在基于所述分割区域确定与所述岩石薄片对应的矿物特征步骤中，包括：
    基于所述分割区域确定组分图像，所述组分图像中包括一个所述组分区域；
    将所述组分图像输入组分种类识别模型，输出得到所述组分图像的组分类型特征，所述组分类型特征用于对与所述组分区域的矿物特征进行表征，所述组分种类识别模型为基于机器学习的神经网络模型；
    基于所述组分图像对应的所述组分类型特征，确定与所述岩石薄片对应的所述矿物特征。
  7. 根据权利要求4所述的方法，其特征在于，在基于所述分割区域确定与所述岩石薄片对应的矿物特征步骤中，包括：
    将所述岩石薄片分割图像输入分类识别模型中，输出得到单晶态组分集合图像和非单晶态组分集合图像，所述单晶态组分集合图像中包括至少一个单晶态组分区域，以及与所述单晶态组分区域对应的单晶态组分识别结果，所述非单晶态组分区域中包括至少一个非单晶态组分区域，以及与所述非单晶态组分区域对应的非单晶态组分识别结果，所述分类识别模型为基于深度学习的神经网络模型；
    基于所述单晶态组分集合图像向矿物数据采集设备发送扫描指令；
    接收所述矿物数据采集设备基于所述扫描指令反馈的扫描数据；
    基于所述扫描数据对所述单晶态组分识别结果进行验证，得到验证结果；
    基于所述非单晶态组分识别结果、所述单晶态组分识别结果以及所述验证结果，确定与所述岩石薄片对应的矿物特征。
  8. 根据权利要求7所述的方法，其特征在于，在基于所述单晶态组分集合图像向所述矿物数据采集设备发送扫描指令步骤中，包括：
    将所述单晶态组分集合图像输入区域筛选模型，输出得到区域筛选结果，所述区域筛选结果用于指示所述单晶态组分集合图像中用于进行面扫描数据的部分；
    基于所述区域筛选结果生成所述扫描指令；
    向所述矿物数据采集设备发送所述扫描指令。
  9. 根据权利要求7所述的方法，其特征在于，在基于所述扫描数据对所述单晶态组分识别结果进行验证，得到验证结果步骤中，包括：
    基于所述扫描数据确定扫描识别结果；
    响应于所述扫描识别结果与所述单晶态组分识别结果相同，确定所述验证结果指示所述验证通过，并确定所述单晶态组分识别结果不变；
    响应于所述面扫描识别结果与所述单晶态组分识别结果不同，确定所述验证结果指示所述验证不通过，并基于识别结果生成规则确定所述单晶态组分识别结果，所述识别结果生成规则中包括基于所述矿物采集设备种类的规则。
  10. 根据权利要求7所述的方法，其特征在于，在基于所述分割区域确定与所述岩石薄片对应的矿物特征步骤中，还包括：
    基于所述分割区域确定组分图像，所述组分图像中包括一个组分区域；
    基于所述组分图像向所述矿物数据采集设备发送光谱数据获取指令；
    接收所述矿物数据采集设备基于所述光谱数据获取指令发送的光谱数据；
    基于所述光谱数据，在矿物光谱数据数据库中确定与所述组分区域对应的初选矿物种类；
    确定与所述初选矿物种类对应的矿物种类验证规则；
    基于所述矿物种类验证规则以及所述初选矿物种类确定所述组分区域对应的矿物种类；
    基于组成所述岩石薄片分割图像的所述组分图像对应的所述矿物种类，确定与所述岩石薄片对应的所述矿物特征。
  11. 根据权利要求10所述的方法，其特征在于，所述矿物光谱数据数据库包括矿物种类子数据库，其中，在基于所述光谱数据，在矿物光谱数据数据库中确定与所述组分区域对应的初选矿物种类步骤中，包括：
    基于所述光谱数据，在所述矿物光谱数据数据库中确定与所述组分区域对应的所述初选矿物种类的一级类别，所述矿物种类的一级类别为基于所述矿物的常见程度划分的类别，所述一级类别包括常见矿物类别与非常见矿物类别，与所述常见矿物类别对应的所述初选矿物种类包括矿物族种类以及普通常见矿物种类，与所述非常见矿物类别对应的所述初选矿物种类包括包裹体矿物种类、强敏感矿物种类以及蚀变矿物种类；
    基于所述矿物种类一级类别对应的矿物种类子数据库，确定与所述组分区域对应的所述初选矿物种类。
  12. 根据权利要求11所述的方法，其特征在于，所述矿物种类验证规则包括分类验证规则、直接验证规则以及选点重验规则，其中，在确定与所述初选矿物种类对应的矿物种类验证规则步骤中，包括：
    响应于所述初选矿物种类为所述矿物族种类，确定所述矿物种类验证规则为所述分类验证规则；
    响应于所述初选矿物种类为所述包裹体矿物种类，或，所述强敏感矿物种类，或，所述蚀变矿物种类，确定所述矿物种类验证规则为所述选点重验规则；
    响应于所述初选矿物种类为所述普通常见矿物种类，或，与所述初选矿物种类对应的一级类别为所述非常见矿物类别，且所述初选矿物类别不为所述包裹体矿物种类、所述强敏感矿物种类与所述蚀变矿物种类中的任意一种，确定所述矿物种类验证规则为所述直接验证规则。
  13. 根据权利要求4所述的方法，其特征在于，所述岩石薄片图像包括至少两个表征组分区域的组分图像，其中，在基于所述岩石薄片图像生成与所述岩石薄片对应的几何特征、矿物特征以及结构特征步骤 中，包括：
    将所述岩石薄片图像输入岩石薄片图像分割识别模型，输出得到与所述岩石薄片图像对应的至少两个分割区域以及与所述分割区域对应的组分类型特征，所述分割区域用于指示所述岩石薄片图像基于组分区域的分割情况，所述组分类型特征用于对所述分割区域中的所述组分区域的矿物特征进行表征，所述岩石薄片图像识别模型为基于机器学习的神经网络模型；
    结合所述分割区域以及所述组分类型特征，确定与所述岩石薄片对应的几何特征以及矿物特征。
  14. 根据权利要求3所述的方法，其特征在于，所述鉴定结果包括初选岩石种类，其中，在基于所述初选岩类结构特征确定与所述岩石薄片对应的初选岩石种类步骤中，包括：
    将所述岩石薄片图像输入初选岩类选择模型中，输出得到初选岩类结果，所述初选岩类结果指示所述岩石薄片对应的所述初选岩类结构特征，所述初选岩类结构选择模型为基于初选岩类图像样本集构建的模型。
  15. 根据权利要求14所述的方法，其特征在于，所述鉴定结果还包括所述细分岩石种类，其中，在基于所述几何特征与所述矿物特征，确定所述细分岩类结构特征步骤中，包括：
    将所述几何特征与所述矿物特征输入细分岩类选择模型中，输出得到细分岩类结果，所述细分岩类结果指示所述岩石薄片对应的所述细分岩类结构特征，所述细分岩类选择模型为基于几何-矿物特征交互样本集构建的模型，所述几何-矿物特征交互样本集指示所述几何特征与所述矿物特征的组合，与所述细分岩类结构特征之间的对应关系。
  16. 根据权利要求15所述的方法，其特征在于，所述鉴定结果还包括岩石薄片子鉴定结果以及岩石薄片名称中的至少一种，其中，在基于所述几何特征、所述矿物特征以及所述结构特征，生成所述岩石薄片的鉴定结果步骤中，包括：
    基于所述初选岩类特征与所述细分岩类特征，确定与所述鉴定结果对应的至少一个岩石薄片子鉴定结果的种类，所述岩石薄片子鉴定结果中包括粒径区间识别结果、最大颗粒粒径识别结果、颗粒分选性结果、颗粒磨圆度识别结果、矿物自形程度结果以及颗粒接触方式识别结果中的至少一种；
    基于所述几何特征与所述矿物特征，生成所述岩石薄片子鉴定结果；
    基于所述岩石薄片子鉴定结果确定所述岩石薄片名称。
  17. 一种岩石鉴定系统，其特征在于，所述岩石鉴定系统包括相互连接的图像采集设备与计算机设备，其中，
    所述图像采集设备，用于生成岩石薄片图像并向所述计算机设备发送所述岩石薄片图像；
    所述计算机设备，用于接收所述图像采集设备发送的岩石薄片图像，并基于所述岩石薄片图像生成与所述岩石薄片对应的几何特征、矿物特征以及结构特征，以及基于所述几何特征、所述矿物特征以及所述结构特征，生成所述岩石薄片的鉴定结果，其中，所述岩石薄片图像为对岩石薄片进行拍摄得到的图像，所述岩石薄片为对岩石样本进行切割得到的薄片，所述岩石薄片图像包括至少一个组分区域，所述几何特征用于指示所述岩石薄片的组分区域划分情况，所述矿物特征用于指示所述岩石薄片中与所述组分区域对应的矿物种类分布情况，所述结构特征用于指示所述岩石样本的岩石种类。
  18. 根据权利要求17所述的系统，其特征在于，所述图像采集设备包括：
    第一光源；
    第一底座和第一支架，所述第一底座与所述第一支架的底部连接；
    第一载物台，所述第一载物台位于所述第一支架的中央；
    第一物镜，所述第一物镜位于所述第一支架的顶部，所述第一物镜内具有检偏镜，所述第一物镜的底部具有第一安装物镜转换器，所述第一安装物镜转换器与所述第一物镜连接；
    偏振装置，所述偏振装置包括起偏镜和检偏镜，所述起偏镜和所述第一光源位于所述第一底座上；
    电荷耦合器件CCD相机，所述CCD相机位于所述第一物镜的顶部，其中，
    当所述图像采集设备生成所述岩石薄片图像时，所述岩石薄片位于所述第一载物台顶部，所述CCD相机处于第一工作状态。
  19. 根据权利要求18所述的系统，其特征在于，所述岩石鉴定系统还包括与所述计算机设备通信连接的矿物数据采集设备，其中，所述矿物数据采集设备包括：
    第二光源，所述第二光源对应有第二光路；
    第二底座和第二支架；
    第二载物台，所述第二载物台位于所述第二支架中央，其中，所述第二光路的终点投射至所述第二载物台；
    第二物镜；
    所述第二物镜位于所述第二支架的顶部，所述第二物镜的底部具有第二安装物镜转换器，所述第二安装物镜转换器与所述第二物镜连接；
    光电信号转换器，所述光电信号转换器位于所述第二物镜的顶部，且与所述计算机设备通信连接；以及
    测量装置，其中，
    当所述矿物数据采集设备生成矿物数据时，所述岩石薄片位于所述第二载物台顶部，所述光电信号转换器处于第二工作状态。
  20. 根据权利要求19所述的系统，其特征在于，所述矿物数据采集设备与所述图像采集设备共用所述第一载物台、所述第一物镜、所述第一底座以及所述第一支架。
  21. 一种岩石鉴定装置，其特征在于，所述装置包括：
    接收模块，用于接收图像采集设备发送的岩石薄片图像，所述岩石薄片图像为对岩石薄片进行拍摄得到的图像，所述岩石薄片为对岩石样本进行切割得到的薄片，所述岩石薄片图像包括至少一个组分区域；
    生成模块，用于基于所述岩石薄片图像生成与所述岩石薄片对应的几何特征、矿物特征以及结构特征，其中，所述几何特征用于指示所述岩石薄片的组分区域划分情况，所述矿物特征用于指示所述岩石薄片中与所述组分区域对应的的矿物种类分布情况，所述结构特征用于指示所述岩石样本的岩石种类；
    所述生成模块，用于基于所述几何特征、所述矿物特征以及所述结构特征，生成所述岩石薄片的鉴定结果。
  22. 一种计算机设备，其特征在于，所述计算机设备包括处理器和存储器，所述存储器存储有至少一条指令、至少一段程序、代码集或指令集，所述至少一条指令、至少一段程序、代码集或指令集由所述处理器加载并执行以实现如权利要求1至16中任一项所述的岩石鉴定方法。
  23. 一种计算机可读存储介质，其特征在于，所述可读存储介质存储有至少一条指令、至少一段程序、代码集或指令集，所述至少一条指令、至少一段程序、代码集或指令集由所述处理器加载并执行以实现如权利要求1至16中任一项所述的岩石鉴定方法。
PCT/CN2021/121840 2021-02-08 2021-09-29 岩石鉴定方法、系统、装置、终端及可读存储介质 WO2022166232A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
US18/264,654 US20240054766A1 (en) 2021-02-08 2021-09-29 Rock identification method, system and apparatus, terminal, and readable storage medium
EP21924229.4A EP4290470A1 (en) 2021-02-08 2021-09-29 Rock identification method, system and apparatus, terminal, and readable storage medium
JP2023548202A JP2024508688A (ja) 2021-02-08 2021-09-29 岩石を特定する方法、システム、装置、端末および可読記憶媒体

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN202110181773.4A CN112730326A (zh) 2021-02-08 2021-02-08 一种岩石薄片智能鉴定装置及方法
CN202110181773.4 2021-02-08
CN202110182027 2021-02-10
CN202110182027.7 2021-02-10

Publications (1)

Publication Number Publication Date
WO2022166232A1 true WO2022166232A1 (zh) 2022-08-11

Family

ID=77755662

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/121840 WO2022166232A1 (zh) 2021-02-08 2021-09-29 岩石鉴定方法、系统、装置、终端及可读存储介质

Country Status (5)

Country Link
US (1) US20240054766A1 (zh)
EP (1) EP4290470A1 (zh)
JP (1) JP2024508688A (zh)
CN (6) CN113435456A (zh)
WO (1) WO2022166232A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113688956A (zh) * 2021-10-26 2021-11-23 西南石油大学 一种基于深度特征融合网络的砂岩薄片分割和识别方法
CN114565820A (zh) * 2022-03-01 2022-05-31 中科海慧（北京）科技有限公司 一种基于时空大数据分析的矿产样本识别系统

Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN203490417U (zh) * 2013-09-22 2014-03-19 成都西图科技有限公司 用于岩石薄片观测的新型显微镜
CN104112126A (zh) * 2014-08-06 2014-10-22 南京大学镇江高新技术研究院 一种大理岩显微薄片自动鉴别方法
CN104134069A (zh) * 2014-08-06 2014-11-05 南京大学 一种页岩显微薄片自动鉴别方法
CN105354600A (zh) * 2015-09-28 2016-02-24 南京大学 一种砂岩显微薄片的自动分类方法
CN106485223A (zh) * 2016-10-12 2017-03-08 南京大学 一种砂岩显微薄片中岩石颗粒的自动识别方法
CN106677708A (zh) * 2016-11-24 2017-05-17 上海工程技术大学 具备岩石薄片鉴定功能的石油勘探用钻井钻头系统及方法
CN108318515A (zh) * 2018-01-09 2018-07-24 南京大学 一种基于扫描电镜能谱分析的单颗粒矿物相自动识别及定量分析方法
CN109283148A (zh) * 2018-09-30 2019-01-29 核工业北京地质研究院 一种基于光谱信息自动识别岩石矿物的方法
CN109523566A (zh) * 2018-09-18 2019-03-26 姜枫 一种砂岩薄片显微图像的自动分割方法
CN109800728A (zh) * 2019-01-28 2019-05-24 济南浪潮高新科技投资发展有限公司 一种基于深度学习的矿物质快速识别的方法
CN110286141A (zh) * 2019-07-15 2019-09-27 中国石油集团渤海钻探工程有限公司 一种基于逻辑回归的自动岩性定名方法
CN111007064A (zh) * 2019-12-13 2020-04-14 常州大学 一种基于图像识别的录井岩性智能识别方法
CN111160158A (zh) * 2019-12-17 2020-05-15 山东大学 偏光显微镜下岩石图像智能识别系统及方法
WO2020225592A1 (en) * 2019-05-09 2020-11-12 Abu Dhabi National Oil Company (ADNOC) Automated method and system for categorising and describing thin sections of rock samples obtained from carbonate rocks

Family Cites Families (31)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1317569C (zh) * 2004-06-29 2007-05-23 中国国土资源航空物探遥感中心 高光谱矿物分层谱系识别方法
MX2007002630A (es) * 2004-09-07 2007-07-13 Petromodel Ehf Aparato y metodo para analisis de tamano, forma y angularidad, para el analisis composicional de particulas minerales y de roca.
CN104181603A (zh) * 2014-07-24 2014-12-03 中国石油大学(华东) 碎屑岩沉积成岩综合相识别方法
CN106294525A (zh) * 2015-06-25 2017-01-04 中国石油化工股份有限公司 一种录井柱状剖面信息提取方法和系统
CN106780536A (zh) * 2017-01-13 2017-05-31 深圳市唯特视科技有限公司 一种基于对象掩码网络的形状感知实例分割方法
CN106875406B (zh) * 2017-01-24 2020-04-14 北京航空航天大学 图像引导的视频语义对象分割方法及装置
US11287548B2 (en) * 2017-11-29 2022-03-29 Landmark Graphics Corporation Geological sediment provenance analysis and display system
CN108510493A (zh) * 2018-04-09 2018-09-07 深圳大学 医学图像内目标对象的边界定位方法、存储介质及终端
US11010883B2 (en) * 2018-04-17 2021-05-18 Saudi Arabian Oil Company Automated analysis of petrographic thin section images using advanced machine learning techniques
CN108805879A (zh) * 2018-05-24 2018-11-13 电子科技大学 一种基于Spiking神经网络的图像分割方法
US10891462B2 (en) * 2018-06-29 2021-01-12 Saudi Arabian Oil Company Identifying geometrical properties of rock structure through digital imaging
CN110873722A (zh) * 2018-09-03 2020-03-10 中国石油化工股份有限公司 一种岩心矿物组分鉴别方法
CN109741358B (zh) * 2018-12-29 2020-11-06 北京工业大学 基于自适应超图学习的超像素分割方法
CN109612943B (zh) * 2019-01-14 2020-04-21 山东大学 基于机器学习的隧洞岩石石英含量测试系统及方法
CN109856029B (zh) * 2019-02-01 2021-07-30 中海石油(中国)有限公司上海分公司 一种基于图像分析的孔隙度评价方法
CN109949317B (zh) * 2019-03-06 2020-12-11 东南大学 基于逐步对抗学习的半监督图像实例分割方法
CN110095388A (zh) * 2019-04-18 2019-08-06 中国石油大学(北京) 碎屑岩颗粒结构的确定方法及装置
CN110443862B (zh) * 2019-06-28 2022-10-14 中国地质科学院矿产资源研究所 基于无人机的岩性填图方法及系统、电子设备
CN110443303B (zh) * 2019-08-04 2023-07-11 中国矿业大学 基于图像分割和分类的煤岩显微组分智能识别方法
CN110490880A (zh) * 2019-08-16 2019-11-22 重庆邮电大学 一种基于局部视觉线索的髋关节x光图像分割方法及系统
CN111027538A (zh) * 2019-08-23 2020-04-17 上海撬动网络科技有限公司 一种基于实例分割模型的集装箱检测方法
CN110837114B (zh) * 2019-10-16 2022-02-01 中国石油天然气股份有限公司 粗面质火山碎屑岩识别方法、装置及电子设备
CN111191741A (zh) * 2020-01-15 2020-05-22 中国地质调查局发展研究中心 一种岩石识别深度学习模型岩石分类约束继承性损失方法
CN111220616B (zh) * 2020-01-21 2021-06-01 山东大学 基于长石特征的隧洞内碎屑岩抗风化能力判别系统与方法
CN111340784B (zh) * 2020-02-25 2023-06-23 安徽大学 一种基于Mask R-CNN图像篡改检测方法
CN111382676B (zh) * 2020-02-25 2023-12-22 南京大学 一种基于注意力机制的沙粒图像分类方法
CN111563445A (zh) * 2020-04-30 2020-08-21 徐宇轩 一种基于卷积神经网络的显微镜下岩性识别方法
CN111640125B (zh) * 2020-05-29 2022-11-18 广西大学 基于Mask R-CNN的航拍图建筑物检测和分割方法及装置
CN111652142A (zh) * 2020-06-03 2020-09-11 广东小天才科技有限公司 基于深度学习的题目分割方法、装置、设备和介质
CN112084660B (zh) * 2020-09-10 2022-05-31 西南石油大学 基于岩电解释模型对深层/超深层碳酸盐岩沉积微相精细划分的方法
CN112132200A (zh) * 2020-09-17 2020-12-25 山东大学 基于多维岩石图像深度学习的岩性识别方法及系统

Patent Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN203490417U (zh) * 2013-09-22 2014-03-19 成都西图科技有限公司 用于岩石薄片观测的新型显微镜
CN104112126A (zh) * 2014-08-06 2014-10-22 南京大学镇江高新技术研究院 一种大理岩显微薄片自动鉴别方法
CN104134069A (zh) * 2014-08-06 2014-11-05 南京大学 一种页岩显微薄片自动鉴别方法
CN105354600A (zh) * 2015-09-28 2016-02-24 南京大学 一种砂岩显微薄片的自动分类方法
CN106485223A (zh) * 2016-10-12 2017-03-08 南京大学 一种砂岩显微薄片中岩石颗粒的自动识别方法
CN106677708A (zh) * 2016-11-24 2017-05-17 上海工程技术大学 具备岩石薄片鉴定功能的石油勘探用钻井钻头系统及方法
CN108318515A (zh) * 2018-01-09 2018-07-24 南京大学 一种基于扫描电镜能谱分析的单颗粒矿物相自动识别及定量分析方法
CN109523566A (zh) * 2018-09-18 2019-03-26 姜枫 一种砂岩薄片显微图像的自动分割方法
CN109283148A (zh) * 2018-09-30 2019-01-29 核工业北京地质研究院 一种基于光谱信息自动识别岩石矿物的方法
CN109800728A (zh) * 2019-01-28 2019-05-24 济南浪潮高新科技投资发展有限公司 一种基于深度学习的矿物质快速识别的方法
WO2020225592A1 (en) * 2019-05-09 2020-11-12 Abu Dhabi National Oil Company (ADNOC) Automated method and system for categorising and describing thin sections of rock samples obtained from carbonate rocks
CN110286141A (zh) * 2019-07-15 2019-09-27 中国石油集团渤海钻探工程有限公司 一种基于逻辑回归的自动岩性定名方法
CN111007064A (zh) * 2019-12-13 2020-04-14 常州大学 一种基于图像识别的录井岩性智能识别方法
CN111160158A (zh) * 2019-12-17 2020-05-15 山东大学 偏光显微镜下岩石图像智能识别系统及方法

Also Published As

Publication number Publication date
JP2024508688A (ja) 2024-02-28
CN113435457A (zh) 2021-09-24
CN113435460A (zh) 2021-09-24
CN113435458A (zh) 2021-09-24
CN113435456A (zh) 2021-09-24
CN113435459A (zh) 2021-09-24
CN113537235A (zh) 2021-10-22
EP4290470A1 (en) 2023-12-13
US20240054766A1 (en) 2024-02-15

Similar Documents

Publication Publication Date Title
WO2022166232A1 (zh) 岩石鉴定方法、系统、装置、终端及可读存储介质
CN105445160B (zh) 一种沥青混合料的空隙特征及其提取方法
CN107004123A (zh) 迭代式缺陷滤除工艺
CN109211904A (zh) 一种沥青混合料二维内部结构检测系统及检测方法
Muszynski et al. Particle shape estimates of uniform sands: visual and automated methods comparison
CN109540929B (zh) 一种测定盆地砂岩型铀矿成矿年龄的方法
Van Ruitenbeek et al. Measuring rock microstructure in hyperspectral mineral maps
CN103576127B (zh) 用于多天线的地面穿透雷达的合并显示
CN115019171B (zh) 非接触式围岩快速分区分级方法
Aligholi et al. Automated mineral identification algorithm using optical properties of crystals
Bhagvati et al. Morphological analysis of pavement surface condition
CN112150430A (zh) 一种利用岩石细观结构数字图像的数值分析方法
Aprile et al. Classification of mineral inclusions in ancient ceramics: comparing different modal analysis strategies
CN104680185A (zh) 基于边界点重分类的高光谱图像分类方法
Seelos et al. RADIUS–rapid particle analysis of digital images by ultra‐high‐resolution scanning of thin sections
KR101370496B1 (ko) 복합매질로 이루어진 시편에 대한 X-ray CT 영상의 최소 단위에 존재하는 각 순수매질의 부피비 측정방법
Chen et al. Rapid extraction of pavement aggregate gradation based on point clouds using deep learning networks
Ferrer et al. Evaluation of a simple and affordable image-based procedure to measure particle size distribution
CN103048257A (zh) 一种砂土种类识别与颗粒组分分析装置
Moaveni Advanced image analysis and techniques for degradation characterization of aggregates
Safaei Pixel and region-based image processing algorithms for detection and classification of pavement cracks
CN110517220A (zh) 一种基于激光三维数据的集料表面数量检测方法
CN108986078A (zh) 一种基于凸包几何特征的粗集料棱角性量化方法
CN105953766A (zh) 基于岩芯图像的砾石磨圆度测量方法
CN112949463B (zh) 一种集料级配快速检测模型的建立、检测方法及系统

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 2023548202

Country of ref document: JP

WWE Wipo information: entry into national phase

Ref document number: 18264654

Country of ref document: US

WWE Wipo information: entry into national phase

Ref document number: 2021924229

Country of ref document: EP

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2021924229

Country of ref document: EP

Effective date: 20230908