AU2021102380A4 - Intelligent Mineral Identification Method and System - Google Patents
Intelligent Mineral Identification Method and System Download PDFInfo
- Publication number
- AU2021102380A4 AU2021102380A4 AU2021102380A AU2021102380A AU2021102380A4 AU 2021102380 A4 AU2021102380 A4 AU 2021102380A4 AU 2021102380 A AU2021102380 A AU 2021102380A AU 2021102380 A AU2021102380 A AU 2021102380A AU 2021102380 A4 AU2021102380 A4 AU 2021102380A4
- Authority
- AU
- Australia
- Prior art keywords
- characteristic
- mineral
- component information
- optical property
- training set
- Prior art date
- Legal status (The legal status 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 status listed.)
- Ceased
Links
- 229910052500 inorganic mineral Inorganic materials 0.000 title claims abstract description 159
- 239000011707 mineral Substances 0.000 title claims abstract description 159
- 238000000034 method Methods 0.000 title claims abstract description 19
- 239000013598 vector Substances 0.000 claims abstract description 81
- 230000003287 optical effect Effects 0.000 claims abstract description 65
- 239000013078 crystal Substances 0.000 claims abstract description 55
- 238000013527 convolutional neural network Methods 0.000 claims description 41
- 230000007935 neutral effect Effects 0.000 claims description 14
- 238000002372 labelling Methods 0.000 claims description 8
- 238000005457 optimization Methods 0.000 claims description 6
- 230000006870 function Effects 0.000 description 10
- 238000001514 detection method Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 239000011435 rock Substances 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- 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
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
- G06V10/464—Salient features, e.g. scale invariant feature transforms [SIFT] using a plurality of salient features, e.g. bag-of-words [BoW] representations
-
- 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
- G06V10/56—Extraction of image or video features relating to colour
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Spectrometry And Color Measurement (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
- Image Analysis (AREA)
Abstract
OF THE DISCLOSURE
The present disclosure relates to an intelligent mineral identification method and system.
Characteristic vectors are obtained by obtaining major component information, minor component
information, an optical property characteristic, a crystal form characteristic, a color characteristic
and a scratch characteristic of a mineral to be tested and calculating respective weights of the
major component information, the minor component information, the optical property
characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic.
A mineral identification model is trained with the characteristic vectors. The mineral to be tested
is input into the trained mineral identification model for classification and identification of the
mineral to be tested. The mineral identification model is trained with a plurality of identification
characteristics, so that the accuracy of mineral identification is increased.
16
-1/2
100 Obtain characteristic data of a mineral to be tested
200 Calculate respective weights of the major component information, the minor
component information, the optical property characteristic, the crystal form
characteristic, the color characteristic and the scratch characteristic
Multiply the major component information, the minor component
300 information, the optical property characteristic, the crystal form
characteristic, the color characteristic and the scratch characteristic by the
respective weights thereof to obtain characteristic vectors
400
Input the characteristic vectors into an intelligent mineral identification
model to obtain a predicted mineral name and a predicted confidence
500 Determine a classification result for the mineral to be tested based on the
predicted mineral name and the predicted confidence
FIG. 1
Description
-1/2
100 Obtain characteristic data of a mineral to be tested
200 Calculate respective weights of the major component information, the minor component information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic
Multiply the major component information, the minor component 300 information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic by the respective weights thereof to obtain characteristic vectors
400 Input the characteristic vectors into an intelligent mineral identification model to obtain a predicted mineral name and a predicted confidence
500 Determine a classification result for the mineral to be tested based on the predicted mineral name and the predicted confidence
FIG. 1
[01] The present disclosure relates to the technical field of intelligent mineral identification, and in particular, to an intelligent mineral identification method and system.
[02] With the development of detection techniques, much more mineralogical studies can be conducted by using large-scale instruments for detection and analysis. In this way, more accurate and reliable results for identification of minerals and rock composition can be obtained. However, when a mineralogical researcher not specializing in detection and analysis gets a detection report with various chemical components, the researcher may have to find reference to know the name of the required mineral. It is quite inefficient to find out the name of a mineral by manual checking, and the results may not be accurate enough. Many minerals belonging to the same category or undergoing isomorphous replacement of elements cannot be accurately differentiated by their principal components alone. Correct names of minerals can be finally obtained based on minor elements, optical property characteristics observed on a rock slice by using a microscope, and identification characteristics (such as crystal form, color, and scratch) of a hand specimen.
[03] A digital intelligent mineral identification and classification technique based on image identification is usually adopted in the prior art. Generally, this technique allows identification and classification of minerals by using a deep learning algorithm based on hyperspectral images or pictures thereof. In general, the hyperspectral images or pictures of a large amount of minerals are captured first for use in training of a convolutional neural network in the deep learning algorithm, and the trained convolutional neural network can be used to identify minerals. However, the training merely with image features may usually lead to low accuracy of identification results. Moreover, there is no obvious difference in characteristics between hyperspectral images or pictures of minerals of different categories. In other words, minerals of different categories cannot be differentiated merely based on images. As a result, training of the deep learning algorithm by directly using the pictures of minerals may result in a low accuracy.
[04] It is to be understood that any acknowledgement of prior art in this specification is not to be taken as an admission that this prior art forms part of the common general knowledge in Australia or elsewhere.
[05] An objective of the present disclosure is to provide an intelligent mineral identification method and system. A mineral identification model is trained with a plurality of identification characteristics for use in intelligent identification of minerals, so that the accuracy of mineral identification is increased.
[06] To achieve the above objective, the present disclosure provides the following solutions:
[07] An intelligent mineral identification method includes:
[08] obtaining characteristic data of a mineral to be tested, where the characteristic data comprises major component information, minor component information, an optical property characteristic, a crystal form characteristic, a color characteristic and a scratch characteristic;
[09] calculating respective weights of the major component information, the minor component information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic;
[10] multiplying the major component information, the minor component information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic by the respective weights thereof to obtain characteristic vectors;
[11] inputting the characteristic vectors into an intelligent mineral identification model to obtain a predicted mineral name and a predicted confidence; and
[12] determining a classification result for the mineral to be tested based on the predicted mineral name and the predicted confidence.
[13] Preferably, the intelligent mineral identification model may be determined by:
[14] obtaining the characteristic data of a mineral to be trained;
[15] labeling the characteristic data of the mineral to be trained in respect of category to obtain label information;
[16] determining a first training set and a second training set, where the first training set comprises an optical property characteristic in the characteristic data of the mineral to be trained and the corresponding label information thereof; and the second training set comprises major component information, minor component information, a crystal form characteristic, a color characteristic and a scratch characteristic in the characteristic data of the mineral to be trained and the respective label information thereof;
[17] calculating a weight of the optical property characteristic in the first training set, and multiplying the optical property characteristic in the first training set by the weight of the optical property characteristic to obtain a first characteristic vector;
[18] calculating respective weights of the major component information, the minor component information, the crystal form characteristic, the color characteristic and the scratch characteristic in the second training set, and multiplying the major component information, the minor component information, the crystal form characteristic, the color characteristic and the scratch characteristic by the respective weights thereof to obtain second characteristic vectors;
[19] building a convolutional neural network module and a fully-connected network module;
[20] inputting the first characteristic vector and the label information in the first training set into the convolutional neural network module for training; and
[21] inputting the second characteristic vectors and the label information in the second training set into the fully-connected network module for training, followed by combining and inputting the trained convolutional neural network module and the trained fully-connected network module into a fully-connected neutral network, where an output from the fully-connected neutral network is a prediction result.
[22] Preferably, the inputting the first characteristic vector and the label information in the first training set into the convolutional neural network module for training may specifically include:
[23] determining a loss function based on the first characteristic vector and the label information in the first training set; and
[24] training the convolutional neural network module by using a gradient descent optimization algorithm so as to minimize the loss function, thereby obtaining the trained convolutional neural network module.
[25] Preferably, the obtaining the characteristic data of a mineral to be trained may specifically include:
[26] extracting characteristic data from a mineral characteristic data base as the characteristic data of the mineral to be trained.
[27] Preferably, the major component information and the minor component information may be numeric data; the optical property characteristic may be pictorial data; and the crystal form characteristic, the color characteristic and the scratch characteristic may be text data.
[28] An intelligent mineral identification system includes:
[29] an obtaining unit for obtaining characteristic data of a mineral to be tested, where the characteristic data comprises major component information, minor component information, an optical property characteristic, a crystal form characteristic, a color characteristic and a scratch characteristic;
[30] a calculating unit for calculating respective weights of the major component information, the minor component information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic;
[31] a vector obtaining unit for multiplying the major component information, the minor component information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic by the respective weights thereof to obtain characteristic vectors;
[32] a predicting unit for inputting the characteristic vectors into an intelligent mineral identification model to obtain a predicted mineral name and a predicted confidence; and
[33] a determining unit for determining a classification result for the mineral to be tested based on the predicted mineral name and the predicted confidence.
[34] Preferably, the intelligent mineral identification system may further include a model determining unit which specifically includes:
[35] an obtaining module for obtaining the characteristic data of a mineral to be trained;
[36] a labeling module for labeling the characteristic data of the mineral to be trained in respect of category to obtain label information;
[37] a training set determining module for determining a first training set and a second training set, where the first training set comprises an optical property characteristic in the characteristic data of the mineral to be trained and the corresponding label information thereof; and the second training set comprises major component information, minor component information, a crystal form characteristic, a color characteristic and a scratch characteristic in the characteristic data of the mineral to be trained and the respective label information thereof;
[38] a first vector determining module for calculating a weight of the optical property characteristic in the first training set, and multiplying the optical property characteristic in the first training set by the weight of the optical property characteristic to obtain a first characteristic vector;
[39] a second vector determining module for calculating respective weights of the major component information, the minor component information, the crystal form characteristic, the color characteristic and the scratch characteristic in the second training set, and multiplying the major component information, the minor component information, the crystal form characteristic, the color characteristic and the scratch characteristic by the respective weights thereof to obtain second characteristic vectors;
[40] a building module for building a convolutional neural network module and a fully-connected network module;
[41] a first model determining module for inputting the first characteristic vector and the label information in the first training set into the convolutional neural network module for training; and
[42] a second model determining module for inputting the second characteristic vectors and the label information in the second training set into the fully-connected network module for training, followed by combining and inputting the trained convolutional neural network module and the trained fully-connected network module into a fully-connected neutral network, where an output from the fully-connected neutral network is a prediction result.
[43] Preferably, the first model determining module specifically includes:
[44] a loss function determining submodule for determining a loss function based on the first characteristic vector and the label information in the first training set; and
[45] a training submodule for training the convolutional neural network module by using a gradient descent optimization algorithm so as to minimize the loss function, thereby obtaining the trained convolutional neural network module.
[46] Preferably, the obtaining module may include:
[47] an obtaining submodule for extracting characteristic data from a mineral characteristic data base as the characteristic data of the mineral to be trained.
[48] Preferably, the major component information and the minor component information may be numeric data; the optical property characteristic may be pictorial data; and the crystal form characteristic, the color characteristic and the scratch characteristic may be text data.
[49] According to specific embodiments of the present disclosure, the present disclosure has the following technical effects:
[50] The present disclosure provides an intelligent mineral identification method and system. Characteristic vectors are obtained by obtaining major component information, minor component information, an optical property characteristic, a crystal form characteristic, a color characteristic and a scratch characteristic of a mineral to be tested and calculating respective weights of the major component information, the minor component information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic. A mineral identification model is trained with the characteristic vectors. The mineral to be tested is input into the trained mineral identification model for classification and identification of the mineral to be tested. The mineral identification model is trained with a plurality of identification characteristics, so that the accuracy of mineral identification is increased.
[51] In order to explain the technical solutions in embodiments of the present disclosure or in the prior art more clearly, the accompanying drawings required in the embodiments will be described below briefly. Apparently, the accompanying drawings in the following description show merely some embodiments of the present disclosure, and other drawings can be derived from these accompanying drawings by those of ordinary skill in the art without creative efforts.
[52] FIG. 1 is a flowchart of an intelligent mineral identification method according to an embodiment of the present disclosure.
[53] FIG. 2 is a schematic diagram of calculating characteristic vectors according to an embodiment of the present disclosure.
[54] FIG. 3 is a module connection diagram of an intelligent mineral identification system according to an embodiment of the present disclosure.
[55] The technical solutions in embodiments of the present disclosure will be described below clearly and completely with reference to the accompanying drawings in the embodiments of the present disclosure. Apparently, the described embodiments are merely a part rather than all of the embodiments of the present disclosure. All other embodiments derived from the embodiments of the present disclosure by a person of ordinary skill in the art without creative efforts shall fall within the protection scope of the present disclosure.
[56] The present disclosure aims to train a mineral identification model with a plurality of identification characteristics for use in intelligent identification of minerals, so that the accuracy of mineral identification is increased.
[57] To make the foregoing objective, features, and advantages of the present disclosure clearer and more comprehensible, the present disclosure will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[58] FIG. 1 is a flowchart of an intelligent mineral identification method according to an embodiment of the present disclosure. As shown in FIG. 1, the intelligent mineral identification method of the present disclosure includes:
[59] Step 100: obtain characteristic data of a mineral to be tested, where the characteristic data comprises major component information, minor component information, an optical property characteristic, a crystal form characteristic, a color characteristic and a scratch characteristic.
[60] Step 200: calculate respective weights of the major component information, the minor component information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic.
[61] Step 300: multiply the major component information, the minor component information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic by the respective weights thereof to obtain characteristic vectors.
[62] Step 400: input the characteristic vectors into an intelligent mineral identification model to obtain a predicted mineral name and a predicted confidence.
[63] Alternatively, the intelligent mineral identification model may include a mineral category identification model and a confidence prediction model.
[64] Step 500: determine a classification result for the mineral to be tested based on the predicted mineral name and the predicted confidence.
[65] FIG. 2 is a schematic diagram of calculating characteristic vectors according to an embodiment of the present disclosure. As shown in FIG. 2, the characteristic vectors are obtained by multiplying identification characteristics by respective weights thereof in the present disclosure. In the characteristic vectors according to the present disclosure, the major component information and the minor component information may be numeric data; the optical property characteristic may be pictorial data; and the crystal form characteristic, the color characteristic and the scratch characteristic may be text data.
[66] Alternatively, major characteristics involved in the present disclosure mainly include major components and minor components of a mineral, an optical property characteristic of the mineral, and visual text information of the mineral, such as crystal form, color and scratches thereof. The characteristics of major components and minor components are mainly organized numerically; the optical property characteristic is presented as an image; and the visual information is presented by natural statements. For the pictorial data, image characteristics are extracted by CNN; and natural text characteristics are characterized and extracted by word vectors.
[67] Specifically, a specialist mineral data analysis system is adopted in the present disclosure to analyze characteristics to obtain respective weights thereof.
[68] Alternatively, there may be one or more characteristic vectors.
[69] Preferably, the intelligent mineral identification model may be determined by a process as follows:
[70] The characteristic data of a mineral to be trained is obtained.
[71] The characteristic data of the mineral to be trained is labeled in respect of category to obtain label information.
[72] A first training set and a second training set are determined. The first training set includes an optical property characteristic in the characteristic data of the mineral to be trained and the corresponding label information thereof; and the second training set includes major component information, minor component information, a crystal form characteristic, a color characteristic and a scratch characteristic in the characteristic data of the mineral to be trained and the respective label information thereof.
[73] A weight of the optical property characteristic in the first training set is calculated, and the optical property characteristic in the first training set is multiplied by the weight of the optical property characteristic to obtain a first characteristic vector.
[74] Respective weights of the major component information, the minor component information, the crystal form characteristic, the color characteristic and the scratch characteristic in the second training set are calculated, and the major component information, the minor component information, the crystal form characteristic, the color characteristic and the scratch characteristic are multiplied by the respective weights thereof to obtain second characteristic vectors.
[75] A convolutional neural network module and a fully-connected network module are built.
[76] The first characteristic vector and the label information in the first training set are input into the convolutional neural network module for training.
[77] Specifically, the trained convolutional neural network module is determined as the mineral category identification model.
[78] The second characteristic vector and the label information in the second training set are input into the fully-connected network module for training.
[79] As an alternative embodiment, the trained fully-connected network module is determined as the confidence prediction model.
[80] The trained convolutional neural network and the trained fully-connected network module are combined and input into a fully-connected neutral network, where an output from the fully-connected neutral network is a prediction result.
[81] As an alternative embodiment, a categorizer pretraining unit is added ahead of the convolutional neural network module and includes a module for generating a data set for neural network training. The module for generating a data set for neural network training is configured to preprocess the optical property characteristic in the characteristic data of a mineral to be trained for use in neutral network training and create the first training set. A convolutional neural network training module performs neutral network calculation with inputs, i.e., the first training set and the label information output from the module for generating a data set for neural network training.
[82] Alternatively, after the completion of the training of the convolutional neural network module and the fully-connected network module, a picture of a mineral to be tested is input into the convolutional neural network module to obtain a one-dimensional characteristic value vector or an unnormalized one-dimensional probability value vector. The fully-connected network module has multiple layers. The second characteristic vectors are input into a fully-connected network of two or more layers and an unnormalized one-dimensional probability value vector is output. The one-dimensional characteristic value vector or the unnormalized one-dimensional probability value vector obtained from the convolutional neural network are then spliced with the unnormalized one-dimensional probability value vector obtained from the fully-connected network to provide a new one-dimensional vector. The new one-dimensional vector is input into a fully-connected network of one or more layers, and final results are output, i.e., probabilities of belonging to different categories and a mineral name.
[83] Preferably, the inputting the first characteristic vector and the label information in the first training set into the convolutional neural network module for training specifically includes the following steps:
[84] A loss function is determined based on the first characteristic vector and the label information in the first training set.
[85] The convolutional neural network module is trained by using a gradient descent optimization algorithm so as to minimize the loss function, thereby obtaining the mineral category identification model.
[86] Preferably, the obtaining the characteristic data of a mineral to be trained may specifically includes the following step:
[87] Characteristic data is extracted from a mineral characteristic data base as the characteristic data of the mineral to be trained.
[88] FIG. 3 is a module connection diagram of an intelligent mineral identification system according to an embodiment of the present disclosure. As shown in FIG. 3, the intelligent mineral identification system of the present system includes:
[89] an obtaining unit for obtaining characteristic data of a mineral to be tested, where the characteristic data comprises major component information, minor component information, an optical property characteristic, a crystal form characteristic, a color characteristic and a scratch characteristic;
[90] a calculating unit for calculating respective weights of the major component information, the minor component information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic;
[91] a vector obtaining unit for multiplying the major component information, the minor component information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic by the respective weights thereof to obtain characteristic vectors;
[92] a predicting unit for inputting the characteristic vectors into an intelligent mineral identification model to obtain a predicted mineral name and a predicted confidence; and
[93] a determining unit for determining a classification result for the mineral to be tested based on the predicted mineral name and the predicted confidence.
[94] Preferably, the intelligent mineral identification system further includes a model determining unit which specifically includes:
[95] an obtaining module for obtaining the characteristic data of a mineral to be trained;
[96] a labeling module for labeling the characteristic data of the mineral to be trained in respect of category to obtain label information;
[97] a training set determining module for determining a first training set and a second training set, where the first training set comprises an optical property characteristic in the characteristic data of the mineral to be trained and the corresponding label information thereof; and the second training set comprises major component information, minor component information, a crystal form characteristic, a color characteristic and a scratch characteristic in the characteristic data of the mineral to be trained and the respective label information thereof;
[98] a first vector determining module for calculating a weight of the optical property characteristic in the first training set, and multiplying the optical property characteristic in the first training set by the weight of the optical property characteristic to obtain a first characteristic vector;
[99] a second vector determining module for calculating respective weights of the major component information, the minor component information, the crystal form characteristic, the color characteristic and the scratch characteristic in the second training set, and multiplying the major component information, the minor component information, the crystal form characteristic, the color characteristic and the scratch characteristic by the respective weights thereof to obtain second characteristic vectors;
[100] a building module for building a convolutional neural network module and a fully-connected network module;
[101] a first model determining module for inputting the first characteristic vector and the label information in the first training set into the convolutional neural network module for training; and
[102] a second model determining module for inputting the second characteristic vectors and the label information in the second training set into the fully-connected network module for training, followed by combining and inputting the trained convolutional neural network module and the trained fully-connected network module into a fully-connected neutral network, where an output from the fully-connected neutral network is a prediction result.
[103] Preferably, the first model determining module specifically includes:
[104] a loss function determining submodule for determining a loss function based on the first characteristic vector and the label information in the first training set; and
[105] a training submodule for training the convolutional neural network module by using a gradient descent optimization algorithm so as to minimize the loss function, thereby obtaining the trained convolutional neural network module.
[106] Preferably, the obtaining module includes:
[107] an obtaining submodule for extracting characteristic data from a mineral characteristic data base as the characteristic data of the mineral to be trained.
[108] Preferably, the major component information and the minor component information are numeric data; the optical property characteristic are pictorial data; and the crystal form characteristic, the color characteristic and the scratch characteristic are text data.
[109] The present disclosure has the following beneficial effects.
[110] (1) According to the present disclosure, a mineral identification model is trained with a plurality of identification characteristics for use in intelligent identification of minerals, so that the accuracy of mineral identification is increased.
[111] (2) By using a mineral characteristic data base (including chemical components, crystal growth characteristics, and optical property characteristics of minerals and visual identification characteristics of hand specimens) resulting from years of work of the applicant, intelligent identification of unknown minerals can be achieved by means of a big data based machine learning. As a result, automatic detection and identification of mineral names are realized, and the degree of automation of identification is increased.
[112] The embodiments of the present specification are described in a progressive manner. Each embodiment focuses on the difference from other embodiment, and the same and similar parts between the embodiments may refer to each other. For the system disclosed in the embodiment, since the system corresponds to the method disclosed in the embodiment, the description is relatively simple, and reference can be made to the method description.
[113] In this specification, several specific embodiments are used for illustration of the principles and implementations of the present disclosure. The description of the foregoing embodiments is used to help illustrate the method of the present disclosure and the core ideas thereof. In addition, those of ordinary skill in the art can make various modifications in terms of specific implementations and the scope of application in accordance with the ideas of the present disclosure. In conclusion, the content of this specification shall not be construed as a limitation to the present disclosure.
Claims (5)
1. An intelligent mineral identification method, comprising: obtaining characteristic data of a mineral to be tested, wherein the characteristic data comprises major component information, minor component information, an optical property characteristic, a crystal form characteristic, a color characteristic and a scratch characteristic; calculating respective weights of the major component information, the minor component information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic; multiplying the major component information, the minor component information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic by the respective weights thereof to obtain characteristic vectors; inputting the characteristic vectors into an intelligent mineral identification model to obtain a predicted mineral name and a predicted confidence; and determining a classification result for the mineral to be tested based on the predicted mineral name and the predicted confidence.
2. The intelligent mineral identification method according to claim 1, wherein the intelligent mineral identification model is determined by: obtaining the characteristic data of a mineral to be trained; labeling the characteristic data of the mineral to be trained in respect of category to obtain label information; determining a first training set and a second training set, wherein the first training set comprises an optical property characteristic in the characteristic data of the mineral to be trained and the corresponding label information thereof; and the second training set comprises major component information, minor component information, a crystal form characteristic, a color characteristic and a scratch characteristic in the characteristic data of the mineral to be trained and the respective label information thereof; calculating a weight of the optical property characteristic in the first training set, and multiplying the optical property characteristic in the first training set by the weight of the optical property characteristic to obtain a first characteristic vector; calculating respective weights of the major component information, the minor component information, the crystal form characteristic, the color characteristic and the scratch characteristic in the second training set, and multiplying the major component information, the minor component information, the crystal form characteristic, the color characteristic and the scratch characteristic by the respective weights thereof to obtain second characteristic vectors; building a convolutional neural network module and a fully-connected network module; inputting the first characteristic vector and the label information in the first training set into the convolutional neural network module for training; and inputting the second characteristic vectors and the label information in the second training set into the fully-connected network module for training, followed by combining and inputting the trained convolutional neural network module and the trained fully-connected network module into a fully-connected neutral network, wherein an output from the fully-connected neutral network is a prediction result; wherein the inputting the first characteristic vector and the label information in the first training set into the convolutional neural network module for training specifically comprises: determining a loss function based on the first characteristic vector and the label information in the first training set; and training the convolutional neural network module by using a gradient descent optimization algorithm so as to minimize the loss function, thereby obtaining the trained convolutional neural network module; wherein the obtaining the characteristic data of the mineral to be trained specifically comprises: extracting characteristic data from a mineral characteristic data base as the characteristic data of the mineral to be trained.
3. The intelligent mineral identification method according to claim 1, wherein the major component information and the minor component information are numeric data; the optical property characteristic is pictorial data; and the crystal form characteristic, the color characteristic and the scratch characteristic are text data.
4. An intelligent mineral identification system, comprising: an obtaining unit for obtaining characteristic data of a mineral to be tested, wherein the characteristic data comprises major component information, minor component information, an optical property characteristic, a crystal form characteristic, a color characteristic and a scratch characteristic; a calculating unit for calculating respective weights of the major component information, the minor component information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic; a vector obtaining unit for multiplying the major component information, the minor component information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic by the respective weights thereof to obtain characteristic vectors; a predicting unit for inputting the characteristic vectors into an intelligent mineral identification model to obtain a predicted mineral name and a predicted confidence; and a determining unit for determining a classification result for the mineral to be tested based on the predicted mineral name and the predicted confidence.
5. The intelligent mineral identification system according to claim 4, further comprising a model determining unit which specifically comprises: an obtaining module for obtaining the characteristic data of a mineral to be trained; a labeling module for labeling the characteristic data of the mineral to be trained in respect of category to obtain label information; a training set determining module for determining a first training set and a second training set, wherein the first training set comprises an optical property characteristic in the characteristic data of the mineral to be trained and the corresponding label information thereof; and the second training set comprises major component information, minor component information, a crystal form characteristic, a color characteristic and a scratch characteristic in the characteristic data of the mineral to be trained and the respective label information thereof; a first vector determining module for calculating a weight of the optical property characteristic in the first training set, and multiplying the optical property characteristic in the first training set by the weight of the optical property characteristic to obtain a first characteristic vector; a second vector determining module for calculating respective weights of the major component information, the minor component information, the crystal form characteristic, the color characteristic and the scratch characteristic in the second training set, and multiplying the major component information, the minor component information, the crystal form characteristic, the color characteristic and the scratch characteristic by the respective weights thereof to obtain second characteristic vectors; a building module for building a convolutional neural network module and a fully-connected network module; a first model determining module for inputting the first characteristic vector and the label information in the first training set into the convolutional neural network module for training; and a second model determining module for inputting the second characteristic vectors and the label information in the second training set into the fully-connected network module for training, followed by combining and inputting the trained convolutional neural network module and the trained fully-connected network module into a fully-connected neutral network, wherein an output from the fully-connected neutral network is a prediction result; wherein the first model determining module specifically comprises: a loss function determining submodule for determining a loss function based on the first characteristic vector and the label information in the first training set; and a training submodule for training the convolutional neural network module by using a gradient descent optimization algorithm so as to minimize the loss function, thereby obtaining the trained convolutional neural network module; wherein the obtaining module comprises: an obtaining submodule for extracting characteristic data from a mineral characteristic data base as the characteristic data of the mineral to be trained; wherein the major component information and the minor component information are numeric data; the optical property characteristic is pictorial data; and the crystal form characteristic, the color characteristic and the scratch characteristic are text data.
－1/2－ － 06 May 2021
100 Obtain characteristic data of a mineral to be tested
200 Calculate respective weights of the major component information, the minor component information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic 2021102380
Multiply the major component information, the minor component 300 information, the optical property characteristic, the crystal form characteristic, the color characteristic and the scratch characteristic by the respective weights thereof to obtain characteristic vectors
400 Input the characteristic vectors into an intelligent mineral identification model to obtain a predicted mineral name and a predicted confidence
500 Determine a classification result for the mineral to be tested based on the predicted mineral name and the predicted confidence
FIG. 1
－2/2－ － 06 May 2021
Major component Major component information weight
Minor component Minor component information weight 2021102380
Optical property Optical property Characteristic vectors characteristic characteristic weight
Color characteristic Color characteristic weight
Scratch characteristic Scratch characteristic weight
FIG. 2
Obtaining unit
Calculating unit
Vector obtaining unit
Predicting unit
Determining unit FIG. 3
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110423576.9 | 2021-04-20 | ||
| CN202110423576.9A CN113128404A (en) | 2021-04-20 | 2021-04-20 | Intelligent mineral identification method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| AU2021102380A4 true AU2021102380A4 (en) | 2021-06-24 |
Family
ID=76442118
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| AU2021102380A Ceased AU2021102380A4 (en) | 2021-04-20 | 2021-05-06 | Intelligent Mineral Identification Method and System |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN113128404A (en) |
| AU (1) | AU2021102380A4 (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115656053B (en) * | 2022-10-19 | 2024-05-31 | 山东大学 | Rock mineral content testing method and system |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102053268B (en) * | 2010-11-18 | 2012-07-25 | 中国石油天然气股份有限公司 | Method and device for identifying lithology of altered volcanic rock |
| CN110119753B (en) * | 2019-01-08 | 2023-05-23 | 长江岩土工程有限公司 | Lithology recognition method by reconstructed texture |
| CN110059765B (en) * | 2019-04-26 | 2021-02-02 | 中国地质大学（北京） | Intelligent mineral identification and classification system and method |
| CN111460947B (en) * | 2020-03-15 | 2022-05-31 | 吉林大学 | BP neural network-based method and system for identifying metal minerals under microscope |
| CN111784642B (en) * | 2020-06-10 | 2021-12-28 | 中铁四局集团有限公司 | Image processing method, target recognition model training method and target recognition method |
-
2021
- 2021-04-20 CN CN202110423576.9A patent/CN113128404A/en active Pending
- 2021-05-06 AU AU2021102380A patent/AU2021102380A4/en not_active Ceased
Also Published As
| Publication number | Publication date |
|---|---|
| CN113128404A (en) | 2021-07-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107273490B (en) | Combined wrong question recommendation method based on knowledge graph | |
| CN110309867B (en) | Mixed gas identification method based on convolutional neural network | |
| CN112597038B (en) | Software defect prediction method and system | |
| CN108334943A (en) | The semi-supervised soft-measuring modeling method of industrial process based on Active Learning neural network model | |
| CN111239137B (en) | Grain quality detection method based on transfer learning and adaptive deep convolution neural network | |
| CN109389418A (en) | Electric service client's demand recognition methods based on LDA model | |
| CN112836067B (en) | Intelligent searching method based on knowledge graph | |
| AU2021102380A4 (en) | Intelligent Mineral Identification Method and System | |
| JP2010128779A (en) | Method for extracting multiple regression equation | |
| CN110515836B (en) | Weighted naive Bayes method for software defect prediction | |
| CN111967501B (en) | Method and system for judging load state driven by telemetering original data | |
| CN111291189B (en) | Text processing method and device and computer readable storage medium | |
| CN113283467A (en) | Weak supervision picture classification method based on average loss and category-by-category selection | |
| CN112579777A (en) | Semi-supervised classification method for unlabelled texts | |
| CN116521872A (en) | Combined recognition method and system for cognition and emotion and electronic equipment | |
| CN112505337B (en) | Data processing method for assisting analysis of sample | |
| CN110045691A (en) | A kind of multitasking fault monitoring method of multi-source heterogeneous big data | |
| CN115222684A (en) | Surface defect detection method based on multitask incremental learning | |
| CN113869194A (en) | Variable parameter milling process signal marking method and system based on deep learning | |
| CN114548088A (en) | Geological text entity relation joint extraction method and system | |
| CN114153968A (en) | Few-sample financial text classification system based on word attribute position relation and Bayes | |
| CN117473096B (en) | Knowledge point labeling method fusing LATEX labels and model thereof | |
| CN116503674B (en) | Small sample image classification method, device and medium based on semantic guidance | |
| CN112990145B (en) | Group-sparse-based age estimation method and electronic equipment | |
| CN117635381B (en) | Method and system for evaluating computing thinking quality based on man-machine conversation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| FGI | Letters patent sealed or granted (innovation patent) | ||
| MK22 | Patent ceased section 143a(d), or expired - non payment of renewal fee or expiry |