Intelligent measurement method for underground mine hard rock pillar surface trace spacing

Abstract

The invention relates to an intelligent measurement method for underground mine hard rock pillar surface trace spacing. The method comprises the following steps: collecting massive original trace feature images; after preprocessing, dividing the image data into a training set and a test set according to a set proportion; constructing a convolutional neural network based on a deep learning framework, and performing training and testing to obtain a prediction model; inputting the target image into the prediction model, and predicting to obtain a two-dimensional trace distribution diagram; firstly, according to the angle difference principle, an angle threshold value is input, fracture segmentation is carried out on a trace, then a fracture threshold value is introduced to carry out secondary detection, and the number of trace segments consistent with an actual investigation result is obtained; according to an angle similarity principle, a classification angle is set, all trace segments are grouped and evaluated, and an optimal grouping number is determined; and carrying out manual scanning line calibration on all the traces and the traces in each independent group, and determining the measurement values of the optimal intra-group spacing and the total spacing. The method can accurately and efficiently measure the trace spacing on the surface of the hard rock pillar.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent monitoring, and particularly relates to an intelligent measurement method for the surface trace spacing of hard rock pillars in underground mines. Background Art

[0002] Trace spacing is one of the important characteristics for evaluating the quality and performance of surface exposure and underground rock masses. Macroscopically, a rock mass is an aggregate of several independent rock blocks, and trace spacing is a direct description of the block size. For hard rock pillars, in the underground environment with high ground stress, high ground temperature, high well depth, and strong mining disturbance, the initial damage degree often depends on the size of the blocks containing developed joints and fractures. Therefore, accurate measurement of the surface trace spacing of hard rock pillars is beneficial to further improve the evaluation accuracy of pillar and stope stability.

[0003] In traditional technologies, the trace spacing on the pillar surface is usually measured based on manual on-site methods. Specifically, geologists or professional technicians use contact devices such as tape measures and French curves to complete complex and lengthy measurement operations on the pillar surface. However, in the dim light environment of underground mines and under the serious interference of unknown dangers such as roof falls and rockfalls, the instability of contact devices and the uncertainty of manual operations further reduce the measurement accuracy of traces, which is particularly prominent when measuring the trace spacing on the surface of ultra-long pillars. Therefore, obtaining two-dimensional high-definition digital images of the pillar surface based on non-contact measurement technology and relying on image processing and intelligent automatic / semi-automatic analysis of image information to extract traces and complete the measurement of the spacing has become a better, more reliable, and safer technical means.

[0004] However, extracting accurate trace information from the two-dimensional images of complex rock masses has always been a frontier and difficult problem in the field of geotechnical engineering. At present, the extraction method based on edge detection is applicable to extracting trace information from the images of exposed rock masses with uniform target distribution and small number. Once the information content in the image is large and the target features are not easy to identify, the extraction method based on deep learning framework to predict the target can capture the trace information features more deeply and reduce unnecessary noise interference. However, it is undeniable that the performance of the deep learning model depends on the quantity and quality of samples, which is difficult to achieve in the complex and dangerous underground mine environment. Therefore, establishing a high-performance prediction model based on limited general-quality images is the basis for ensuring the measurement of trace spacing. On the other hand, although the predicted traces are continuous and complete, the one-to-one correspondence between the traces and the discontinuous planes is ignored. How to segment the traces at key positions to accurately describe the plane boundary is the premise for calculating the spacing. In addition, due to the disordered distribution of discontinuity surfaces, accurate classification of all traces in advance is required for measuring the trace spacing in similar planes. Most of the existing classification methods are developed based on the clustering concept, which usually only focuses on the similarity between traces and ignores the discontinuity of traces. Therefore, it is inevitable to have measurement errors in the actual application process.

[0005] Based on the above problems, there is an urgent need to propose a new intelligent measurement method for the trace spacing on the surface of hard rock pillars in underground mines, so as to realize the intelligent calculation of the trace spacing on the pillar surface and provide reliable technical support for the stability evaluation of hard rock pillars in underground mines. Summary of the Invention

[0006] In view of the problems existing in the above-mentioned prior art, the present invention provides an intelligent measurement method for the trace spacing on the surface of hard rock pillars in underground mines. The method is simple in implementation process, low in implementation cost, high in measurement efficiency and accuracy. It can realize the intelligent and automatic calculation process of the trace spacing on the surface of hard rock pillars in underground mines, accurately and efficiently measure the trace spacing data on the surface of hard rock pillars, be conducive to further improving the accuracy of the stability evaluation of pillars and stope, and provide effective technical support for the stability evaluation of hard rock pillars in underground mines.

[0007] In order to achieve the above object, the present invention provides an intelligent measurement method for the trace spacing on the surface of hard rock pillars in underground mines, which specifically includes the following steps:

[0008] Step 1: Collect a large number of original trace feature images of pillars in different states on the surface of underground stopes;

[0009] Step 2: Preprocess the collected original trace feature images, recombine the preprocessed images to complete the construction of the image database, and divide the image data in the image database into a training set and a test set according to a set ratio;

[0010] Step 3: Construct a convolutional neural network based on a deep learning framework, train the convolutional neural network using the training set to obtain a high-precision convolutional neural network, and test the performance of the high-precision convolutional neural network using the test set to obtain a prediction model for predicting the trace distribution in the image;

[0011] Step 4: Take the target image to be analyzed as input data and input it into the prediction model, and use the prediction model to predict the trace features of the target image to obtain a predicted distribution map of the traces;

[0012] Step 5: First, according to the principle of angle difference, input an angle threshold, and perform fracture segmentation on all the traces obtained through prediction to initially obtain several trace segments with a certain pixel length; then introduce a fracture threshold to perform secondary detection on the several initially obtained trace segments to obtain the number of trace segments consistent with the actual survey results;

[0013] Step 6: According to the principle of angle similarity, set a classification angle, group and evaluate all the trace segments, and determine the optimal number of groups and the number of traces included in each group based on the principle of minimizing the classification error;

[0014] Step 7: Manually perform scan line calibration on all the traces and the traces within each individual group, and determine the measured values of the optimal intra-group spacing and the total spacing based on the calibration results.

[0015] As an optimization, the different states in Step 1 include a stable state, a temporarily stable state, and an unstable state.

[0016] Furthermore, in order to obtain rich original trace feature images and thus ensure that a prediction model with higher prediction accuracy can be obtained, a high-performance prediction model is established based on limited general-quality images. The specific process of Step 2 is as follows:

[0017] S21: Conduct on-site screening of the original trace feature images on the surface of the ore pillar, and eliminate images with too high or too low exposure;

[0018] S22: Manually mark the trace features in each image, and repeatedly verify whether the marks are reasonable by geological experts or other professional and technical personnel;

[0019] S23: Use data augmentation techniques to perform flipping, moving, rotating at a set angle, local magnification, and mirroring operations on each image to expand the number of original trace images, and thus complete the construction of the image database;

[0020] S24: Allocate the image data in the database, and divide 80% of the image data into the training set and 20% of the image data into the test set. At the same time, calculate the trace and non-trace pixel distributions of the images in the training set and the test set. If the pixel distribution difference between the two sets meets the pixel distribution requirements, it is determined that the data allocation result is reasonable; otherwise, re-allocate the image data in the database until the pixel distribution between the two sets meets the pixel distribution requirements.

[0021] Further, in order to capture the trace information features more deeply and reduce unnecessary noise interference to ensure the prediction accuracy, in step three, first construct the total loss function, and then train the convolutional neural network. During the training process, evaluate whether the performance of the trained convolutional neural network meets the requirements through the preset accuracy index.

[0022] Further, accurately segmenting the trace at the key position so that it can truly and accurately describe the plane boundary is the premise for accurately calculating the spacing. In order to establish a one-to-one correspondence between the trace and the discontinuous plane, a segmentation algorithm is used to segment the trace. At the same time, in order to further strengthen the accuracy of the trace break, a break threshold is incorporated to evaluate the segmentation effect, so as to achieve accurate measurement of the trace spacing on the pillar surface. Therefore, the specific process of step five is as follows:

[0023] S51: Confirm the number of pixel points contained in each trace. Starting from the starting point, calculate the included angle of the line segments formed by three consecutive pixel points.

[0024] S52: Set different angle thresholds based on the angle-based break algorithm and compare them with all the line segment included angles obtained in S51 one by one to determine whether the trace needs to be interrupted. If the angle threshold is greater than the current line segment included angle, break the trace at the midpoint of the line segment forming the included angle and obtain the trace break diagram. At the same time, use this midpoint as the starting point for calculating the next line segment included angle value; otherwise, continue to compare the next line segment included angle until all traces are traversed and the corresponding trace break diagrams are obtained.

[0025] S53: Set the break threshold and import all the trace segments obtained in S52, and delete the trace segments with lengths less than the specified break threshold to obtain a new trace break diagram.

[0026] S54: Let geological experts or other professional technicians compare the actual survey and the distribution of the new trace segments obtained in S53. If there is a situation where non-negligible trace segments are deleted, it is determined that the selection of the break threshold is excessive. Re-execute S53 to obtain a new trace break diagram through the re-adjusted break threshold. At the same time, by combining the adjustment of different angle thresholds, ensure that the tiny traces that do not contribute significantly to the spacing measurement are not extracted too much until the optimal angle threshold and break threshold are determined to complete the trace break evaluation and obtain the final trace break diagram.

[0027] Furthermore, since the distribution of discontinuity planes is chaotic, measuring the spacing between traces distributed in similar planes requires accurate classification of all trace segments in advance. Most existing classification methods are developed based on the clustering concept, which basically only focuses on the similarity between traces and ignores the discontinuity of traces. Therefore, it is inevitable to cause errors in actual on-site applications. To accurately and efficiently classify the trace segments distributed in similar planes, the following grouping algorithm is used for processing. Thus, the specific process of step six is as follows:

[0028] S61: Determine the optimal classification angle value, classified angle;

[0029] S62: Import the trace fracture map obtained after segmentation based on the optimal angle threshold and cropping based on the fracture threshold, and calculate the difference in the angles between any two trace segments in the map and the horizontal axis;

[0030] S63: Divide the two trace segments with the calculated difference less than the optimal classification angle value, classified angle, into the same group. By analogy, traverse all trace segments. After the classification is completed, determine the final number of trace groups.

[0031] Furthermore, to determine the optimal number of groups, a strategy based on classification error is adopted to determine the optimal number of groups. In S61 of step six, first calculate the classification error Error according to formula (1), and then determine the optimal classification angle value, classified angle, according to the principle of minimizing the classification error Error;

[0032]

[0033] where θ i m is the angle of the m-th trace segment in the i-th trace group; averθ i is the average angle of all trace segments in the i-th trace group, and I is the maximum number of trace groups.

[0034] Furthermore, to accurately and efficiently measure the spacing information, in step seven, during the in-group spacing measurement process, make the scanning line perpendicular to most of the trace segments in the same group. At the same time, during the total spacing measurement process, make the scanning line perpendicular to most of the trace segments among all traces.

[0035] As an optimization, in S23 of step two, set the angle to 45 degrees or 90 degrees or 180 degrees or 270 degrees.

[0036] The present invention provides a semi - automated and human - machine interactive spacing measurement method for complex traces on the surface of hard rock pillars in underground mines. First, a large number of characteristic images of the pillar surface taken in underground stopes are collected to train a convolutional neural network based on a deep learning framework, obtaining a prediction model for predicting the trace distribution in the image. This model can automatically capture the trace information features from the target image to be analyzed and helps reduce unnecessary noise interference. Secondly, the prediction model is used to predict the target image to be analyzed to obtain the predicted distribution map of the traces, and a fracture algorithm based on angles is used for segmentation. At the same time, fracture threshold deletion is incorporated to remove trace segments that do not conform to the actual investigation, and the final new trace fracture map is obtained by adjusting the angle threshold and fracture threshold, integrating the actual investigation results and expert evaluation. Thus, the present invention effectively integrates the concept of fracture threshold into the traditional fracture algorithm based on angle threshold, and then realizes the accurate segmentation of trace segments. Accurately segmenting the traces at key positions can help truly and accurately reflect the situation of the plane boundary, so that the traces can be corresponded to the discontinuous planes one by one, ensuring the accurate measurement of the spacing between the traces on the surface of the pillar. Then, based on the principle of minimizing classification error, the optimal classification angle is determined, and a grouping algorithm based on the classification angle is used to divide the trace segments into different sets to obtain the optimal number of groups. Thus, the present invention develops a new grouping algorithm based on the classification angle to reasonably divide the trace segments, and proposes a criterion for minimizing error to enhance the accuracy of trace grouping. It can accurately and efficiently classify the trace segments distributed in similar planes, and at the same time, ensure that the discontinuity of the traces is not ignored, and further ensure that the spacing between the traces distributed in similar planes can be accurately measured. Finally, artificial scan - line calibration is performed on all traces and the traces within each group to obtain the intra - group and total spacing values of the traces.

[0037] This method has a simple implementation process, low implementation cost, high measurement efficiency, and high measurement accuracy. It can realize the intelligent and automated calculation process of the spacing between the traces on the surface of hard rock pillars in underground mines, and can accurately and efficiently measure the trace spacing data on the surface of hard rock pillars. It is beneficial to further improve the accuracy of the stability evaluation of pillars and stopes, and can provide effective technical support for the stability evaluation of hard rock pillars in underground mines. Description of the Drawings

[0038] Figure 1 is the flowchart of the present invention;

[0039] Figure 2 is the state diagram of the enhanced processing of the trace image data of the present invention;

[0040] Figure 3 is the trace fracture diagram of the present invention;

[0041] Figure 4 Schematic diagram of trace segment grouping in the present invention;

[0042] Figure 5 Schematic diagram of calibration of scanning lines for measuring trace spacing in the present invention. Detailed implementation manners

[0043] The present invention will be further described below in conjunction with embodiments.

[0044] For the convenience of those of ordinary skill in the art to understand and implement the present invention, reference is made to Figures 1 to 5 to further describe the present invention;

[0045] The present invention provides an intelligent measurement method for the surface trace spacing of hard rock pillars in underground mines, specifically including the following steps:

[0046] Step 1: Collect a large number of original trace feature images on the surfaces of pillars in different states in underground stopes; preferably, collect the original trace feature images in the underground stope using an image acquisition device under sufficient light conditions, and ensure that each image contains as much trace feature information as possible;

[0047] Step 2: Preprocess the collected original trace feature images, recombine the preprocessed images to complete the construction of an image database, and divide the image data in the image database into a training set and a test set according to a set ratio;

[0048] Step 3: Construct a classical convolutional neural network based on a deep learning framework, train the convolutional neural network using the training set to obtain a high-precision convolutional neural network, and test the performance of the high-precision convolutional neural network using the test set to obtain a prediction model (high-precision convolutional neural network) for predicting the trace distribution in the image;

[0049] Step 4: Input the target image to be analyzed as input data into the prediction model, and use the prediction model to predict the trace features of the target image to obtain a predicted distribution map of the traces;

[0050] Step 5: First, according to the angle difference principle, input an angle threshold, perform fracture segmentation on all the traces obtained through prediction to initially obtain several trace segments with a certain pixel length; then introduce a fracture threshold to perform secondary detection on the initially obtained several trace segments to obtain the number of trace segments consistent with the actual exploration results;

[0051] Step 6: According to the angle similarity principle, set a classification angle, group and evaluate all the trace segments, and determine the optimal number of groups and the number of traces it contains based on the principle of minimizing classification error;

[0052] Step 7: Manually calibrate the scan lines for all the traces and the traces within each individual group, and determine the optimal in-group spacing and the measured value of the total spacing based on the calibration results.

[0053] As an optimization, the different states in Step 1 include a stable state, a temporarily stable state, and an unstable state.

[0054] The performance of the prediction model depends on the quantity and quality of the samples, which is difficult to achieve in the complex and dangerous underground mine environment. Therefore, establishing a high-performance prediction model based on limited general-quality images is the basis for ensuring the measurement of the trace spacing. To obtain rich original trace feature images based on limited image data and thus ensure a prediction model with higher prediction accuracy, the specific process of Step 2 is as follows:

[0055] S21: Conduct on-site screening of the original trace feature images on the surface of the ore pillar, and eliminate the images with excessive or insufficient exposure;

[0056] S22: Manually mark the trace features in each image, and repeatedly verify whether the marks are reasonable by geological experts or other professional and technical personnel;

[0057] S23: As Figure 2 shown, use data augmentation techniques to flip, move, rotate at a set angle, locally magnify, and mirror each image to expand the quantity of the original trace images, and thus complete the construction of the image database;

[0058] S24: To avoid overfitting during the model training process, distribute the image data in the database, and divide 80% of the image data into the training set and 20% of the image data into the test set; meanwhile, conduct distribution statistics on the trace and non-trace pixels of the images in the above two sets. Specifically, calculate the trace and non-trace pixel distributions of the images in the training set and the test set. If the pixel distribution difference between the two sets meets the pixel distribution requirements, the data distribution result is considered reasonable; otherwise, re-distribute the image data in the database until the pixel distribution between the two sets meets the pixel distribution requirements.

[0059] To ensure the prediction accuracy, in Step 3, first construct the total loss function, and then train the convolutional neural network. During the training process, evaluate whether the performance of the trained convolutional neural network meets the requirements through a preset accuracy index.

[0060] Splitting the trace at key positions so that it can accurately and truly describe the plane boundary is a prerequisite for accurately calculating the spacing. To establish a one-to-one correspondence between the traces and the discontinuous planes, a splitting algorithm is adopted to process the traces. At the same time, to further enhance the accuracy of trace fracture, a fracture threshold is incorporated to evaluate the splitting effect, so as to achieve accurate measurement of the spacing between the traces on the ore pillar surface. Therefore, the specific process of step five is as follows:

[0061] S51: Confirm the number of pixel points contained in each trace. Starting from the starting point, calculate the included angle of the line segments formed by three consecutive pixel points;

[0062] S52: Set different angle thresholds based on the angle-based fracture algorithm and compare them with all the included angles of the line segments obtained in S51 one by one to determine whether the trace needs to be interrupted. If the angle threshold is greater than the current included angle of the line segment, break the trace at the midpoint of the line segment forming the included angle and obtain the trace fracture diagram. At the same time, use this midpoint as the starting point for calculating the next included angle value of the line segment; otherwise, continue to compare the next included angle of the line segment; until all traces are traversed and the corresponding trace fracture diagrams are obtained;

[0063] S53: Set the fracture threshold and import all the trace segments obtained in S52. Delete the trace segments with lengths less than the specified fracture threshold to obtain a new trace fracture diagram;

[0064] S54: Let geological experts or other professional and technical personnel compare the actual survey and the distribution of the new trace segments obtained in S53. If there is a situation where non-negligible trace segments are deleted, it is considered that the selection of the fracture threshold is excessive. Re-execute S53 to obtain a new trace fracture diagram through the re-adjusted fracture threshold. At the same time, by combining the adjustment of different angle thresholds, ensure that small traces that do not contribute significantly to the spacing measurement are not extracted too much until the optimal angle threshold and fracture threshold are determined to complete the trace fracture evaluation and obtain the final trace fracture diagram.

[0065] Since the distribution of discontinuous surfaces is chaotic, accurate classification of all trace segments needs to be carried out in advance for measuring the spacing between traces distributed in similar planes. Most of the existing classification methods are developed based on the clustering concept, which basically only focuses on the similarity between traces and ignores the discontinuity of traces. Therefore, it is inevitable to cause errors in actual field applications. To accurately and efficiently classify the trace segments distributed in similar planes, as Figure 4 shown, the following grouping algorithm is adopted for processing. Therefore, the specific process of step six is as follows:

[0066] S61: Determine the optimal classification angle value, classified angle;

[0067] S62: Import the trace breakage map obtained after segmentation based on the optimal angle threshold and cropping based on the breakage threshold, and calculate the difference in the angles between any two trace segments in the map and the horizontal axis;

[0068] S63: Group two trace segments with a calculated difference less than the optimal classification angle value classified angle into the same group, and so on. Traverse all trace segments. After the classification is completed, determine the final number of trace groups.

[0069] To determine the optimal number of groups, a strategy based on classification error is adopted to determine the optimal number of groups. In S61 of step six, first calculate the classification error Error according to formula (1), and then determine the optimal classification angle value classified angle according to the principle of minimizing the classification error Error;

[0070]

[0071] In the formula, θ i m is the angle of the m-th trace segment in the i-th trace group; averθ i is the average angle of all trace segments in the i-th trace group, and I is the maximum number of trace groups.

[0072] To accurately and efficiently measure the spacing information, as Figure 5 shown, in step seven, during the in-group spacing measurement process, make the scan line perpendicular to most of the trace segments in the same group, that is, the scan line should be perpendicular to as many trace segments in the same group as possible. At the same time, during the total spacing measurement process, make the scan line perpendicular to most of the trace segments among all the traces, that is, the number of trace segments in a perpendicular position relationship with the scan line should be as large as possible.

[0073] As an optimization, in S23 of step two, set the angle to 45 degrees or 90 degrees or 180 degrees or 270 degrees.

[0074] Example:

[0075] Figure 3 (a) is an example of an actual application. The image to be analyzed is a two-dimensional trace image with a size of 1000*1000 pixels, taken from the surface of a stable ore pillar in a certain underground mine. Since the ore pillar is located in a large space where mining tasks are being carried out, and the surface trace distribution is dense and there are too many short traces, the risk of manual measurement increases and the measurement accuracy does not meet the standard. Therefore, the intelligent measurement method in the present invention is used to measure the trace spacing.

[0076] First, import the image to be analyzed into the trained prediction model to obtain the predicted trace distribution map. It can be seen that the distribution of the predicted traces is continuous among different planes. Therefore, perform a fracture segmentation operation on all the above predicted traces. Set the angle threshold to 140 degrees. If the included angle of the line segment formed by three pixel points in any trace is less than the angle threshold, then break the trace at the position of the middle pixel point, and then use this pixel point as the starting point for the next included angle of the line segment to be detected, and so on until the trace is broken Figure 3 (b). To avoid deleting important trace segments and extracting too many insignificant tiny traces, set the fracture threshold to 2 pixels. If the length of the trace segment after fracture is less than the fracture threshold, then delete this trace segment. Evaluating all trace segments in this way, the final trace fracture can be obtained Figure 3 (c). Subsequently, set the classification angle value to 12.5 degrees, calculate the angle difference between any two trace segments. If the angle difference is less than the classification angle, then divide the above two trace segments into the same group. Repeat this operation until all trace segments are grouped. At the same time, calculate the grouping error, and obtain the optimal trace grouping according to the principle of the minimum classification error Figure 4 , finally, manually calibrate the scan lines and calculate the intra-group spacing and total spacing values of the traces respectively.

[0077] The present invention provides a semi - automated and human - machine interactive spacing measurement method for complex traces on the surface of hard rock pillars in underground mines. First, a large number of characteristic images of the pillar surface taken in underground stopes are collected to train a convolutional neural network based on a deep learning framework, obtaining a prediction model for predicting the trace distribution in the image. This model can automatically capture the trace information features from the target image to be analyzed and helps reduce unnecessary noise interference. Secondly, the prediction model is used to predict the target image to be analyzed to obtain the predicted distribution map of the traces, and a fracture algorithm based on angles is used for segmentation. At the same time, fracture threshold deletion is incorporated to remove trace segments that do not conform to the actual survey, and the final new trace fracture map is obtained by adjusting the angle threshold and fracture threshold, integrating the actual survey results and expert evaluation. Thus, the present invention effectively integrates the concept of fracture threshold into the traditional fracture algorithm based on angle threshold, thereby achieving accurate segmentation of trace segments. Accurate segmentation of traces at key positions can help accurately reflect the situation of the plane boundary, so that the traces can be corresponded to the discontinuous planes one by one, ensuring accurate measurement of the trace spacing on the surface of the pillar subsequently. Then, based on the principle of minimizing classification error, the optimal classification angle is determined, and a grouping algorithm based on the classification angle is used to divide the trace segments into different sets to obtain the optimal number of groups. Thus, the present invention develops a new grouping algorithm based on the classification angle to reasonably divide the trace segments, and proposes a criterion for minimizing error to enhance the accuracy of trace grouping. It can accurately and efficiently classify the trace segments distributed in similar planes, and at the same time, ensure that the discontinuity of the traces is not ignored, and further ensure that the trace spacing distributed in similar planes can be accurately measured. Finally, artificial scanning line calibration is performed on all traces and the traces within each group to obtain the intra - group and total spacing values of the traces.

[0078] The implementation process of this method is simple, with low implementation cost, high measurement efficiency and high measurement accuracy. It can realize the intelligent and automated calculation process of the trace spacing on the surface of hard rock pillars in underground mines, and can accurately and efficiently measure the trace spacing data on the surface of hard rock pillars, which is conducive to further improving the accuracy of the stability evaluation of pillars and stopes, and can provide effective technical support for the stability evaluation of hard rock pillars in underground mines.

Claims

1. An intelligent measurement method for the surface trace spacing of hard rock pillars in an underground mine, which is characterized in that it specifically includes the following steps: Step 1: Collect original trace feature images of the surfaces of pillars in different states in a large number of underground mines; Step 2: preprocess the collected original trace feature images, reassemble the preprocessed images to complete the construction of the image database, and divide the image data in the image database into a training set and a test set according to a set ratio; Step 3: Build a convolutional neural network based on a deep learning framework, use the training set to train the convolutional neural network to obtain a high-precision convolutional neural network, and use the test set to test the performance of the high-precision convolutional neural network to obtain a prediction model for predicting the distribution of traces in the image; Step 4: Input the target image to be analyzed as input data into the prediction model, use the prediction model to predict the trace features of the target image, and obtain a predicted distribution map of the trace; Step 5: First, according to the angle difference principle, input the angle threshold, perform break segmentation on all predicted traces, and initially obtain several trace segments with a certain pixel length; Then the fracture threshold is introduced to conduct secondary detection on several trace segments obtained initially, and the number of trace segments consistent with the actual survey results is obtained; Step 6: According to the principle of angle similarity, set the classification angle, group and evaluate all trace segments, and determine the optimal number of groups and the number of trace lines they contain based on the principle of minimizing classification errors; Step 7: Perform manual scan line calibration on all traces and traces in each individual group, and determine the optimal intra-group spacing and total spacing measurement values ​​based on the calibration results.

2. The intelligent measurement method for the surface trace spacing of hard rock pillars in underground mines according to claim 1 is characterized in that: The different states in step one include a stable state, a temporarily stable state and an unstable state.

3. The intelligent measurement method for the surface trace spacing of hard rock pillars in underground mines according to claim 1 is characterized in that: The specific process of step 2 is as follows: S21: Screening the original trace feature images on the surface of the pillars on site, and removing images with excessive or insufficient exposure; S22: Manually mark the trace features in each image, and have geological experts or other professional technicians repeatedly check whether the marking is reasonable; S23: Using data enhancement technology to flip, move, rotate at a set angle, partially enlarge and mirror each image to expand the number of original trace images, thereby completing the construction of the image database; S24: Allocate the image data in the database, and divide 80% of the image data into a training set, and divide 20% of the image data into a test set; at the same time, calculate the trace and non-trace pixel distribution of the images in the training set and the test set. If the pixel distribution difference between the two sets meets the pixel distribution requirements, the data allocation result is considered reasonable. Otherwise, the image data in the database is reallocated until the pixel distribution between the two sets meets the pixel distribution requirements.

4. The intelligent measurement method for the surface trace spacing of hard rock pillars in underground mines according to claim 1 or 2, characterized in that: In step three, the total loss function is first constructed, and then the convolutional neural network is trained. During the training process, the preset accuracy index is used to evaluate whether the performance of the convolutional neural network after training meets the requirements.

5. The intelligent measurement method for the surface trace spacing of hard rock pillars in underground mines according to claim 1 is characterized in that: The specific process of step five is as follows: S51: confirm the number of pixels contained in each trace, and calculate the angle of the line segment formed by three consecutive pixels starting from the starting point; S52: different angle thresholds are set based on the angle breaking algorithm and compared with all the line segment angles obtained in S51 one by one to determine whether the trace needs to be interrupted. If the angle threshold is greater than the current line segment angle, the trace is disconnected at the middle point of the line segment angle and a trace breaking diagram is obtained. At the same time, the middle point is used as the starting point for calculating the next line segment angle value. Otherwise, the comparison of the next line segment angle is continued; until all traces are traversed and the corresponding trace breaking diagram is obtained; S53: setting a fracture threshold, importing all trace segments obtained in S52, deleting trace segments whose length is less than the specified fracture threshold, and obtaining a new trace fracture map; S54: Geological experts or other professional technicians compare the distribution of new trace segments obtained in actual survey with that obtained in S53. If trace segments that cannot be ignored are deleted, it is determined that the fracture threshold is over-selected, and S53 is re-executed to obtain a new trace fracture map through the readjusted fracture threshold. At the same time, by combining the adjustment of different angle thresholds, it is ensured that tiny traces that do not contribute significantly to the spacing measurement are not extracted too much, until the optimal angle threshold and fracture threshold are determined to complete the trace fracture evaluation and obtain the final trace fracture map.

6. The intelligent measurement method for the surface trace spacing of hard rock pillars in underground mines according to claim 4 is characterized in that: The specific process of step six is ​​as follows: S61: Determine the best classified angle value; S62: importing a trace fracture graph obtained after segmentation based on the optimal angle threshold and cropping based on the fracture threshold, and calculating the difference between the angles between any two trace segments in the graph and the horizontal axis; S63: grouping two trace segments whose calculated difference is less than the optimal classified angle value into the same group, and so on, traversing all trace segments, and determining the final number of trace groupings after the classification is completed.

7. The intelligent measurement method for the surface trace spacing of hard rock pillars in underground mines according to claim 5, characterized in that: In step 6 S61, the classification error Error is first calculated according to formula (1), and then the optimal classification angle value classified angle is determined according to the principle of minimizing the classification error Error; In the formula, is the angle of the mth trace segment in the ith trace group; averθ i is the average angle of all trace segments in the i-th trace group, and I is the maximum number of trace groups.

8. The intelligent measurement method for the surface trace spacing of hard rock pillars in underground mines according to claim 6 is characterized in that: In step seven, during the intra-group spacing measurement, the scan line is made perpendicular to most of the trace segments in the same group, and during the total spacing measurement, the scan line is made perpendicular to most of the trace segments in all traces.

9. The intelligent measurement method for the surface trace spacing of hard rock pillars in underground mines according to claim 3 is characterized in that: In step 2 S23, the angle is set to 45 degrees, 90 degrees, 180 degrees or 270 degrees.

Images