CN114387328A - RQD calculation method based on deep learning model and core image - Google Patents

RQD calculation method based on deep learning model and core image Download PDF

Info

Publication number
CN114387328A
CN114387328A CN202111582422.0A CN202111582422A CN114387328A CN 114387328 A CN114387328 A CN 114387328A CN 202111582422 A CN202111582422 A CN 202111582422A CN 114387328 A CN114387328 A CN 114387328A
Authority
CN
China
Prior art keywords
core
image
rock
rqd
cores
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.)
Pending
Application number
CN202111582422.0A
Other languages
Chinese (zh)
Inventor
张野
陈金桥
李炎隆
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian University of Technology
Original Assignee
Xian University of Technology
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
Application filed by Xian University of Technology filed Critical Xian University of Technology
Priority to CN202111582422.0A priority Critical patent/CN114387328A/en
Publication of CN114387328A publication Critical patent/CN114387328A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • 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
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • 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]
    • 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
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30132—Masonry; Concrete

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Geometry (AREA)
  • Geophysics And Detection Of Objects (AREA)

Abstract

The invention discloses an RQD (total Quadrature amplitude resolution) calculation method based on a deep learning model and a drill core image, which comprises the following steps of: establishing a borehole core image dataset; extracting image characteristics of a drill core image data set, inputting the image characteristics into a UNet depth network for training to obtain an EUNet model; performing semantic segmentation on the drilling rock core image by using an EUNet segmentation model to obtain a rock core area and a background area, and finding out a single-row drilling rock core image; extracting the outlines of all rock cores in the rock core image of the single-row drilling, counting the number of pixels of the rock core outline, making a pixel oscillogram, and judging the type of the rock core through the pixel oscillogram; determining the number of complete rock cores and the positions of the complete rock cores from the pixel oscillogram, and determining a research area; further determining the boundary line of the rock core; and calculating the length of the core according to the boundary line of the core to obtain the RQD. The intelligent quantitative analysis of the RQD is realized, and the efficiency of geological exploration is improved.

Description

RQD calculation method based on deep learning model and core image
Technical Field
The invention belongs to the technical field of image processing methods, and relates to an RQD (total-Quadrature-resolution) calculation method based on a deep learning model and a core image.
Background
Large and medium-sized hydraulic and hydroelectric engineering is mostly positioned in high mountains and canyons with complicated geological conditions, fault structures develop, and the integrality of rock masses in different areas has obvious difference. Accurate and objective judgment of geological conditions and rock integrity is an important prerequisite for design and construction of hydraulic and hydroelectric engineering.
The RQD is used as an important rock mass quality evaluation index and is widely applied to hydraulic engineering, and the RQD is also a basic parameter in rock mass multifactor evaluation systems such as RMR and Q-system. The traditional method for acquiring the RQD is obtained by calculating after geological workers measure the length of a drill core, but the manual RQD measurement wastes time and labor, and the geological exploration efficiency is reduced.
Disclosure of Invention
The invention aims to provide an RQD calculation method based on a deep learning model and a core image, and solves the problem of low geological exploration efficiency caused by manual RQD measurement in the prior art.
The invention adopts the technical scheme that an RQD calculation method based on a deep learning model and a drill core image comprises the following steps:
step 1, collecting a drilling core image, and establishing a drilling core image data set after correcting the drilling core image;
step 2, extracting image characteristics of the drilling rock core image data set, inputting the image characteristics into a UNet depth network for training to obtain an EUNet model;
step 3, performing semantic segmentation on the drill hole rock core image by using an EUNet segmentation model to obtain a rock core area and a background area, and finding out a single-row drill hole rock core image;
step 4, extracting the outlines of all rock cores in the rock core image of the single-row drill hole by applying a Canny edge detection algorithm, counting the number of pixels of the rock core outline, making a pixel oscillogram, and judging the type of the rock core through the pixel oscillogram;
step 5, determining the number of complete rock cores and the positions of the complete rock cores from the pixel oscillogram, and taking a crest group adjacent to the head and the tail of the complete rock core as a research area when the pixels of rock core areas at two ends of the complete rock core section are increased; if continuous wave crests and wave troughs appear at one end of the complete core section, the length of the research area is preferably 30 mm;
step 6, finding a research area from the semantically segmented image, and taking a central line of the background area as a boundary line of the rock core; when the cores are tightly attached together, fitting a partial boundary line by adopting a curve fitting method to obtain a boundary line of the cores;
and 7, calculating the length of the core according to the boundary line of the core, wherein the RQD is the ratio of the sum of the lengths of the cores with the lengths larger than 10cm in the footage to the total footage.
The invention is also characterized in that:
and after the step 2, judging the performance of the EUNet model, and performing the next step when the performance meets the requirement.
The method for judging the performance of the EUNet model comprises the following steps:
and (3) segmenting the drill core image by using an EUNet model, further calculating an F1_ score value and a IoU value, and judging the performance of the EUNet model according to the F1_ score value and the IoU value.
The judgment mode of the single-row drill core image in the step 3 is as follows: traversing pixels of each row in the segmentation result, counting the number of pixels of the core area in each row, and cutting out a single-row core by taking the row with zero number of pixels of the core area as a segmentation line to obtain a single-row drilling core image;
the method for calculating the core length comprises the following steps: the actual length represented by a single pixel in the drilling core image is the number of pixels in the length direction of the drilling core image after the physical length of the core box is compared with that of the corrected drilling core image; taking a horizontal center line of a single line of cores as a datum line for determining the length of the cores, and when the calculated cores are in the middle position, the length of the cores is the total length of pixels of a core area between a left boundary and a right boundary on the datum line; when the core is at two ends, the core length is the total length from the first core area pixel at the head or tail of the datum line to the core area pixel between the boundaries.
The invention has the beneficial effects that:
according to the RQD calculation method based on the depth learning model and the core image, accurate identification of the core part in the core image of the drilled hole is achieved by adopting a depth model migration algorithm, and the core image segmentation result is evaluated by adopting multiple indexes, so that the influence of subjective factors is avoided, and the identification efficiency of the core is improved; the method comprises the steps of segmenting a drilling rock core image in actual engineering by using a rock core segmentation model, segmenting a single-row rock core by scanning transverse pixels, obtaining a rock core profile map by profile detection, scanning longitudinal pixels of the profile map to obtain a pixel oscillogram, performing morphological analysis and logic analysis on a semantic segmentation image and the oscillogram, summarizing an internal distribution rule among pixels, exploring a rock core length calculation method, realizing intelligent quantitative analysis of RQD (total root-rank decomposition) and improving geological exploration efficiency; provides theoretical basis and method support for rapid evaluation of rock mass quality indexes in hydraulic engineering rock foundations.
Drawings
FIG. 1 is a flow chart of the RQD calculation method of the present invention based on a deep learning model and borehole core images;
FIG. 2 is a schematic diagram of an image correction process of the RQD calculation method based on the deep learning model and the borehole core image of the present invention:
FIG. 3 is a schematic structural diagram of a UNet model in the RQD calculation method based on the deep learning model and the borehole core image;
FIG. 4 is a schematic diagram of changes of different evaluation indexes in the EUNet model training process in the RQD calculation method based on the deep learning model and the borehole core image;
FIG. 5 is a comparison graph of borehole core images and semantic segmentation results in the RQD calculation method based on the depth learning model and the borehole core images according to the present invention;
FIG. 6a is a comparison graph of single-row core images and semantic segmentation results in the RQD calculation method based on the deep learning model and the borehole core images according to the present invention;
FIG. 6b is a single core row segmentation position determination diagram in the RQD calculation method based on the depth learning model and the borehole core image according to the present invention;
FIG. 7 is a comparison graph of a single-rank core semantic segmentation result and a Canny edge detection result in the RQD calculation method based on the deep learning model and the borehole core image;
FIG. 8 is a waveform diagram of pixels in the RQD calculation method based on a deep learning model and a borehole core image according to the present invention;
FIG. 9 is a schematic diagram of the determination of the study region in the RQD calculation method based on the deep learning model and the borehole core image according to the present invention;
FIG. 10 is a schematic diagram of a fitted core boundary curve in a study region in the RQD calculation method based on a depth learning model and borehole core images according to the present invention;
FIG. 11 is a diagram of a method for determining the core length in the RQD calculation method based on the depth learning model and the borehole core image according to the present invention.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
The RQD calculation method based on the depth learning model and the borehole core image, as shown in FIG. 1, comprises the following steps:
step 1, collecting a drilling core image, correcting the drilling core image, and establishing a drilling core image data set, wherein the drilling core image data set comprises a training set and a testing set;
step 1.1, collecting borehole core images, wherein the borehole core images comprise complete core images and broken, flaky and fractured cores;
step 1.2, correcting images and unifying sizes of the drill core images; the image rectification method comprises the following steps: determining the position of the core box through four corner points of the core box, taking four sides of a rectangular frame determined by the four points as an object, stretching and rotating an image in the frame to enable a view field and the image to be horizontal, and referring to fig. 2; the uniform image size is 640 × 480;
step 1.3, marking a core part in the drill core image processed in the step 1.2 to manufacture an image tag; and (5) taking 80% of the core images as a training set and 20% of the core images as a verification set, and completing the preparation of the obtained core data set. In this example, 280 photographs of the core of the borehole were collected, 240 training sets and 40 testing sets were selected.
Step 2, extracting image features in a training set by using EfficientNet 5 as a pre-training model, inputting the image features into a UNet deep network for training, as shown in FIG. 3, setting 200 epochs in the training process, setting the batch size (the number of samples selected in one training) to be 4, and consuming 3.2 hours in the model training process to obtain an EUNet model; segmenting the core image of the verified centralized drilled hole by using an EUNet model, and further calculating F1_ score value and IoU value, wherein different evaluation index changes in the model training process are shown in FIG. 4; and setting that the trained model can be applied to segmentation of the borehole core image when the value of F1_ score is more than 95% and the value of IoU is more than 85%, wherein F1_ score converges to 0.973 and IoU converges to 0.947 in the embodiment, and the identification accuracy requirement is met. The F1_ score and IoU values were calculated as follows:
Figure BDA0003426623830000051
Figure BDA0003426623830000052
Figure BDA0003426623830000053
Figure BDA0003426623830000061
wherein TP represents that the label is positive and the prediction is positive; FP represents that the label is negative and the prediction is positive; TN represents that the label is negative and the prediction is negative; FN means tag is positive, predicted negative; DT represents the prediction result range, GT represents the real result range of the object; n represents intersection, u represents union; precision is Precision and Recall is Recall.
Step 3, performing semantic segmentation on the drill core image by using an EUNet segmentation model to obtain a core region and a background region, wherein a in the graph is the drill core image and b in the graph is a semantic segmentation result graph as shown in FIG. 5; in the segmentation result in the embodiment, the core area is white, and the background area is black; traversing pixels of each line in the segmentation result, counting the number of white pixels in each line, and cutting a single-row core by taking the line with zero number of white pixels as a segmentation line to obtain a single-row drill core image, as shown in fig. 6a and 6b, wherein a in fig. 6a is a drill core image, and b in the drawing is a semantic segmentation result;
step 4, extracting the outlines of all cores in a single-row drill core image by applying a Canny edge detection algorithm, drawing the core outlines by using single-width white pixels, as shown in fig. 7, wherein a in fig. 7 is a drill core image, b in the drawing is a Canny edge detection result, counting the number of pixels of the core outlines, making a pixel oscillogram, and judging the core type by using the pixel oscillogram, specifically, regarding white pixels in the oscillogram which are continuously 2 as complete cores, regarding the rest parts as crushing sections, as shown in fig. 8, wherein a in the graph is a single-row core semantic segmentation result, b in the graph is a Canny edge detection result, and c in the graph is a oscillogram;
step 5, determining the number of complete rock cores and the positions of the complete rock cores from the pixel oscillogram, and simultaneously determining a research area; specifically, in the pixel oscillogram, when the pixels of the white areas at the two ends of the complete core segment are increased, the crest groups adjacent to the head and the tail of the complete core segment are taken as research areas, as shown in fig. 9, the core marked with numbers in the graph is a determined complete core block body needing length calculation, and the rest is a crushing segment; if continuous wave crests and wave troughs appear at one end of the complete core section, the boundary of a research area is difficult to determine, when the research area is too long, the subsequent processing is not facilitated, and the length of the research area is preferably 30mm after a large amount of processed core data are summarized;
step 6, finding a research area from the semantically segmented image, taking a background area (black area) as a separation area between cores, and taking the central line of the black area as a boundary line of the cores; when the cores are tightly attached together, the EUNet model cannot completely separate the cores, and a complete boundary line of the two cores cannot be obtained, so that a curve fitting method is adopted to obtain the boundary line of the cores by fitting partial boundary lines, as shown in FIG. 10; in the embodiment, the quartic polynomial curve is adopted, so that the rock core boundary line can be accurately fitted;
step 7, the actual length represented by a single pixel in the drilling core image is the number of pixels in the length direction of the drilling core image after the physical length of the core box is compared with that of the corrected drilling core image; taking a horizontal center line of a single line of cores as a datum line for determining the length of the cores, and when the calculated cores are in the middle position, the length of the cores is the total length of pixels of a core area between a left boundary and a right boundary on the datum line; when the core is at two ends, the core length is the total length from the first core area pixel at the head or tail of the datum line to the core area pixel between the boundaries, as shown in fig. 11; then RQD is the ratio of the sum of the core lengths in the footage greater than 10cm to the total footage.
The method has great significance for exploring the quality of the side slope and the underground rock mass in water conservancy, highway, mining and other projects, can directly expose the underground rock mass by adopting a drilling coring method, evaluates the quality of the rock mass, and is widely applied to the projects. The drilled rock cores are loaded into a standard rock core box according to the drilling sequence and photographed and recorded, so that the geological engineering personnel can conveniently perform RQD record and analysis of lithology and other rock indexes. The shot core picture can reflect the position and size information of the core, the information in the core can be interpreted through digital image processing, and the core RQD is obtained. However, when the core picture is shot, shooting directions and angles of different core boxes are different, and dust, mud and the like on the engineering site are attached to the surface of the core or the core box, so that the image recognition difficulty is greatly increased. The invention can overcome the problems of angle and position deviation when the rock core is shot by applying the image correction algorithm; establishing an EUNet deep learning model to realize accurate segmentation of the rock core; and providing a refined analysis flow to realize the intelligent calculation of the RQD.
Through the mode, the RQD calculation method based on the depth learning model and the core image realizes accurate identification of the core part in the core image of the drilled hole by adopting a depth model migration algorithm, and evaluates the core image segmentation result by adopting multiple indexes, avoids the influence of subjective factors, and improves the identification efficiency of the core; the method comprises the steps of segmenting a drilling rock core image in actual engineering by using a rock core segmentation model, segmenting a single-row rock core by scanning transverse pixels, obtaining a rock core profile map by profile detection, scanning longitudinal pixels of the profile map to obtain a pixel oscillogram, performing morphological analysis and logic analysis on a semantic segmentation image and the oscillogram, summarizing an internal distribution rule among pixels, exploring a rock core length calculation method, realizing intelligent quantitative analysis of RQD (total root-rank decomposition) and improving geological exploration efficiency; provides theoretical basis and method support for rapid evaluation of rock mass quality indexes in hydraulic engineering rock foundations.
Examples
The method of the invention is adopted to calculate the RQD of 40 drill core images of a test set, the length of 1302 core blocks is calculated in an accumulated mode, the result is shown in table 1, a single-box core is taken as a research object, the manual measurement result (M) and the automatic quantification result (A) of the RQD, and the Absolute Error (AE) and the Relative Error (RE) of the RQD prediction are listed in the table, wherein AE is | M-A |, and RE is |, AE/M. Compared with manual measurement, the average value of the absolute errors of the RQD predicted values is 1.48%, the maximum value is 3.74%, and the RQD quantization precision requirement can be met.
TABLE 1 comparison of manual and automatic RQD quantification results
Figure BDA0003426623830000081
Figure BDA0003426623830000091

Claims (5)

1. The RQD calculation method based on the deep learning model and the borehole core image is characterized by comprising the following steps of:
step 1, collecting a drilling core image, and establishing a drilling core image data set after correcting the drilling core image;
step 2, extracting image characteristics of the borehole core image data set, inputting the image characteristics into a UNet depth network for training to obtain an EUNet model;
step 3, performing semantic segmentation on the drill hole rock core image by using the EUNet segmentation model to obtain a rock core area and a background area, and finding out a single-row drill hole rock core image;
step 4, extracting the outlines of all rock cores in the rock core image of the single-row drill hole by applying a Canny edge detection algorithm, counting the number of pixels of the rock core outline, making a pixel oscillogram, and judging the type of the rock core through the pixel oscillogram;
step 5, determining the number of the complete rock cores and the positions of the complete rock cores from the pixel oscillogram, and taking a crest group adjacent to the head and the tail of the complete rock core as a research area when the pixels of the rock core areas at the two ends of the complete rock core section are increased; if continuous wave crests and wave troughs appear at one end of the complete core section, the length of the research area is preferably 30 mm;
step 6, finding the research area from the semantically segmented image, and taking the central line of the background area as the boundary line of the rock core; when the cores are tightly attached together, fitting a partial boundary line by adopting a curve fitting method to obtain a boundary line of the cores;
and 7, calculating the length of the core according to the boundary line of the core, wherein the RQD is the ratio of the sum of the lengths of the cores with the lengths larger than 10cm in the footage to the total footage.
2. The method for calculating the RQD based on the deep learning model and the borehole core image according to claim 1, wherein after the step 2, the EUNet model is subjected to performance judgment, and when the performance meets the requirement, the next step is performed.
3. The RQD calculation method based on the deep learning model and the borehole core image as claimed in claim 2, wherein the method for judging the performance of the EUNet model is as follows:
and segmenting the drill core image by using the EUNet model, further calculating an F1_ score value and a IoU value, and judging the performance of the EUNet model according to the F1_ score value and the IoU value.
4. The RQD calculation method based on the deep learning model and the borehole core image as claimed in claim 1, wherein the determination manner of the single-row borehole core image in step 3 is as follows: and traversing pixels of each row in the segmentation result, counting the number of pixels of the core area in each row, and cutting out a single-row core by taking the row with zero number of pixels of the core area as a segmentation line to obtain a single-row drilling core image.
5. The RQD calculation method based on the deep learning model and the borehole core image as claimed in claim 1, wherein the calculation method of the core length is as follows: the actual length represented by a single pixel in the drilling core image is the number of pixels in the length direction of the drilling core image after the physical length of the core box is compared with that of the corrected drilling core image; taking a horizontal center line of a single line of cores as a datum line for determining the length of the cores, and when the calculated cores are in the middle position, the length of the cores is the total length of pixels of a core area between a left boundary and a right boundary on the datum line; when the core is at two ends, the core length is the total length from the first core area pixel at the head or tail of the datum line to the core area pixel between the boundaries.
CN202111582422.0A 2021-12-22 2021-12-22 RQD calculation method based on deep learning model and core image Pending CN114387328A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111582422.0A CN114387328A (en) 2021-12-22 2021-12-22 RQD calculation method based on deep learning model and core image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111582422.0A CN114387328A (en) 2021-12-22 2021-12-22 RQD calculation method based on deep learning model and core image

Publications (1)

Publication Number Publication Date
CN114387328A true CN114387328A (en) 2022-04-22

Family

ID=81197485

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111582422.0A Pending CN114387328A (en) 2021-12-22 2021-12-22 RQD calculation method based on deep learning model and core image

Country Status (1)

Country Link
CN (1) CN114387328A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115035297A (en) * 2022-06-16 2022-09-09 东北大学 Automatic recording method, system, device and medium for drilling core RQD
CN116071725A (en) * 2023-03-06 2023-05-05 四川蜀道新能源科技发展有限公司 Pavement marking recognition method and system
CN116109906A (en) * 2023-04-10 2023-05-12 四川省地质矿产勘查开发局一0六地质队 Artificial intelligent rock mass RQD calculation method based on combination of MaskRCNN and U-Net
CN117706068A (en) * 2024-02-06 2024-03-15 湖南省通盛工程有限公司 Bridge basement rock RQD determination method, system and storage medium

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115035297A (en) * 2022-06-16 2022-09-09 东北大学 Automatic recording method, system, device and medium for drilling core RQD
CN116071725A (en) * 2023-03-06 2023-05-05 四川蜀道新能源科技发展有限公司 Pavement marking recognition method and system
CN116071725B (en) * 2023-03-06 2023-08-08 四川蜀道新能源科技发展有限公司 Pavement marking recognition method and system
CN116109906A (en) * 2023-04-10 2023-05-12 四川省地质矿产勘查开发局一0六地质队 Artificial intelligent rock mass RQD calculation method based on combination of MaskRCNN and U-Net
CN116109906B (en) * 2023-04-10 2023-06-23 四川省地质矿产勘查开发局一0六地质队 Artificial intelligent rock mass RQD calculation method based on combination of MaskRCNN and U-Net
CN117706068A (en) * 2024-02-06 2024-03-15 湖南省通盛工程有限公司 Bridge basement rock RQD determination method, system and storage medium
CN117706068B (en) * 2024-02-06 2024-04-19 湖南省通盛工程有限公司 Bridge basement rock RQD determination method, system and storage medium

Similar Documents

Publication Publication Date Title
CN114387328A (en) RQD calculation method based on deep learning model and core image
CN113610778B (en) Bridge surface crack detection method and system based on semantic segmentation
CN106290388A (en) A kind of insulator breakdown automatic testing method
CN107807387B (en) Acquisition methods when seismic first break neural network based is walked
CN107833213A (en) A kind of Weakly supervised object detecting method based on pseudo- true value adaptive method
CN110689512B (en) Method for quickly splicing and fusing annular images of panoramic video in hole into image
CN110619258B (en) Road track checking method based on high-resolution remote sensing image
CN111145157B (en) Road network data automatic quality inspection method based on high-resolution remote sensing image
CN110333543B (en) Low-resistance body interpretation and imaging method and system based on reflection coefficient analysis
CN110133639B (en) Dowel bar construction quality detection method
CN101697229A (en) Method for extracting region of interest of medical image
CN112819066A (en) Res-UNet single tree species classification technology
CN104992429A (en) Mountain crack detection method based on image local reinforcement
CN113628227A (en) Coastline change analysis method based on deep learning
CN114022474A (en) Particle grading rapid detection method based on YOLO-V4
CN115597494B (en) Precision detection method and system for prefabricated part preformed hole based on point cloud
CN105069395A (en) Road marking automatic identification method based on terrestrial three-dimensional laser scanning technology
CN113658245B (en) Method for acquiring rock mass structural plane occurrence information from drilling image
CN115393605A (en) Core RQD digital statistical method, equipment and terminal based on image recognition technology
CN106296678A (en) RQD based on boring optical image technology analyzes method
CN114662915A (en) Method for rapidly acquiring rock core sampling rate and RQD value
CN117934870A (en) Automatic core identification method and system based on texture assistance
CN111091049B (en) Road surface obstacle detection method based on reverse feature matching
CN116597193A (en) Tunnel intelligent construction-oriented tunnel face unmanned geological logging method and system
CN117152344A (en) Tunnel surrounding rock structural surface analysis method and system based on photo reconstruction point cloud

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination