A mine ecological environment monitoring method and system based on image processing

Abstract

The present application relates to the field of image processing technology, and in particular to a method and system for monitoring the ecological environment of a mine based on image processing, the method comprising: collecting monitoring data of each monitoring point in the monitoring mine at a historical moment; calculating the environmental complexity of each monitoring point and the local Moran index of multiple preset scales; calculating the selection score of each preset scale; selecting N preset scales according to the selection score, and counting the proportion of the number of monitoring points whose local Moran index is greater than the index threshold in the N preset scales to draw a proportion curve; determining the inflection point of the proportion curve, and using the preset scale corresponding to the inflection point as the target scale to construct a convolutional neural network, which can output the monitoring results of each monitoring point. Through the technical solution of the present application, the ecological environment monitoring results at various locations in the mine can be accurately obtained.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a method and system for monitoring the ecological environment of a mine based on image processing. Background Art

[0002] The mining process will have a certain impact on the ecological environment of the mine. As the concept of green development continues to gain popularity, how to conduct real-time monitoring of the ecological environment of the mine during the mining process has become an important task in the field of mining.

[0003] At present, a patent application document with publication number CN116189010A discloses a mine ecology identification and early warning method and system based on satellite maps, wherein the method includes: obtaining satellite map data of the mine, integrating satellite map data at different times and angles after preprocessing, generating a mine satellite map time series sequence for feature extraction, matching the extracted features with the meteorological data of the area where the mine is located, obtaining fusion features, and evaluating the current mine ecological environment; constructing a mine ecological environment early warning model, using the fusion features to make predictions based on the evaluation results of the current mine ecological environment, and outputting the analysis and prediction results of the mine ecological environment; comparing the analysis and prediction results of the mine ecological environment with preset thresholds, and generating early warning prompts based on the comparison results.

[0004] The above method obtains fusion features by matching the mine satellite image time series and meteorological data, and inputs the fusion features into the mine ecological environment early warning model to obtain the analysis and prediction results of the mine ecological environment; however, mining is a complex and lengthy process, and the mine terrain is complex. It is impossible to accurately obtain the ecological environment monitoring results at various locations in the mine based on the fusion features alone, resulting in inaccurate mine ecological environment monitoring results. Summary of the invention

[0005] In order to solve the technical problem of inaccurate mine ecological environment monitoring results, the present application provides a mine ecological environment monitoring method and system based on image processing, which can accurately obtain the ecological environment monitoring results at various locations in the mine.

[0006] In a first aspect, the present application provides a method for monitoring the ecological environment of a mine based on image processing, the monitoring method comprising: setting a plurality of monitoring points in a monitored mine, collecting monitoring data of each monitoring point at a historical moment; calculating the environmental complexity of each monitoring point and a plurality of local Moran indexes of preset scales, wherein the environmental complexity is positively correlated with the mining duration and the information entropy of the monitoring data at a historical moment; calculating the selection score of each preset scale, and the preset scale The selection score for: , Including all monitoring points, For monitoring point The complexity of the environment, is the sum of the environmental complexity of all monitoring points, For monitoring point Preset Scale local Moran's index; N preset scales are selected in descending order of the selected scores, and the proportion of the number of monitoring points whose local Moran's index is greater than the index threshold in the N preset scales is counted, and a proportion curve is drawn with N as the horizontal coordinate and the proportion of the number of monitoring points corresponding to N as the vertical coordinate; the inflection point of the proportion curve is determined, and the preset scale corresponding to the inflection point is used as the target scale to construct a convolutional neural network, and the convolutional neural network is trained based on the monitoring data of historical moments, the input of the convolutional neural network is the real-time monitoring data of each monitoring point at any time, and the output is the monitoring result of each monitoring point, and the monitoring result is normal or abnormal.

[0007] First, the monitored mine is divided into multiple monitoring areas, and the center point of each monitoring area is used as the monitoring point; the environmental complexity of each monitoring point is calculated based on the mining time and the information entropy of the monitoring data. The greater the environmental complexity, the greater the difficulty of the ecological environment monitoring results of the corresponding monitoring point; at the same time, the local Moran index of each monitoring point in multiple preset scales is calculated. If the monitoring point At the preset scale The larger the absolute value of the local Moran index is, the better the monitoring point is. At the preset scale There is a strong correlation within the preset scale Internal monitoring point When conducting ecological environment monitoring, the monitoring data of adjacent monitoring points can provide effective reference information and improve the The accuracy of the monitoring results is improved; further, according to the complexity of the environment, the local Moran's index of each monitoring point at the preset scale is weighted and summed to obtain the selection score of the preset scale, and N preset scales are selected in descending order of the selection score, and the proportion of the number of monitoring points whose local Moran's index is greater than the index threshold in the N preset scales is counted, and a proportion curve is drawn with N as the horizontal coordinate and the proportion of the number of monitoring points corresponding to N as the vertical coordinate; the inflection point of the proportion curve is determined, and the preset scale corresponding to the inflection point is used as the target scale to construct a convolutional neural network, so that the convolutional neural network can extract the information within the target scale of each monitoring point, thereby ensuring that the convolutional neural network can accurately output the monitoring results of each monitoring point.

[0008] Preferably, the monitoring point The complexity of the environment for:

[0009] , For the mining time, is the number of historical moments, is the information entropy; the mining duration is the number of moments in the mining process in all historical moments, and the information entropy is the arrangement entropy or sample entropy of a time series composed of monitoring data at multiple historical moments.

[0010] Environmental complexity can accurately quantify the difficulty of obtaining ecological environment monitoring results at the corresponding monitoring point. The more frequent the mining activities, the greater the impact of human factors on the corresponding monitoring point, and the greater the environmental complexity of the monitoring point; the greater the information entropy, the more chaotic the changes in the monitoring data of the corresponding monitoring point, and the greater the environmental complexity of the monitoring point.

[0011] Preferably, the monitoring point Preset Scale The local Moran index for:

[0012] , For monitoring point The standardized value of For preset scale The number of inner adjacent monitoring points, For monitoring point and adjacent monitoring points The spatial weight of Adjacent monitoring point The standardized value of is the sum of the spatial weights of all adjacent monitoring points. To adjacent monitoring point The distance is negatively correlated.

[0013] A local Moran index of a preset scale is used to measure the correlation between the monitoring data of a monitoring point and other monitoring points in the local area of ​​the preset scale; For example, if the preset scale The local Moran index The larger the value, the larger the preset size. Adjacent monitoring points and monitoring points within There is a strong correlation between the monitoring points When conducting ecological environment monitoring, the monitoring data of adjacent monitoring points can provide effective reference information, that is, Monitoring points obtained within The higher the accuracy of the ecological environment monitoring results, the more accurate the preset scale can be quantified through the local Moran index. Monitoring points obtained within The accuracy of ecological environment monitoring results.

[0014] Preferably, counting the proportion of the number of monitoring points whose local Moran's index is greater than the index threshold in the N preset scales includes: for any monitoring point, in response to the local Moran's index of any preset scale in the N preset scales being greater than the index threshold, marking the monitoring point; the proportion of the number of monitoring points is the proportion of the number of marked monitoring points in all monitoring points.

[0015] Preferably, the convolutional neural network includes an encoder and a decoder, the encoder includes multiple convolutional layers, and taking the preset scale corresponding to the inflection point as the target scale to construct the convolutional neural network includes: obtaining the receptive field of each convolutional layer in the encoder, and taking the convolutional layer whose receptive field is equal to the target scale as the target convolutional layer; taking the output results of each target convolutional layer as the input of the decoder to output a monitoring result map, and the monitoring result map includes the monitoring results of each monitoring point.

[0016] The convolutional layers with different receptive fields in the encoder are screened at the target scale to determine the input of the decoder, so that the decoder can learn the neighborhood information of any monitoring point within the target scale, ensuring that each monitoring point can obtain accurate ecological environment monitoring results.

[0017] Preferably, the decoder includes multiple upsampling layers, and the output results of each target convolution layer are used as the input of the decoder to output a monitoring result graph, including: in response to the size of the output result of any upsampling layer being the same as the size of the output result of the target convolution layer, the output results of the target convolution layer and the upsampling layer are added and used as the input of the next adjacent upsampling layer.

[0018] Preferably, the training of the convolutional neural network based on the monitoring data at historical moments includes: obtaining the label information of each monitoring point at a historical moment, the label information being normal or abnormal; inputting the monitoring data of each monitoring point at a historical moment into the convolutional neural network, outputting the monitoring results of each monitoring point at a historical moment, calculating the cross entropy loss of the label information and the monitoring results, and updating the convolutional neural network using the gradient descent method; iteratively updating the convolutional neural network until the cross entropy loss is less than the preset loss, completing the training.

[0019] Preferably, the cross entropy loss for:

[0020] , Including all monitoring points, For monitoring point Monitoring results at historical moments, For monitoring point Label information at historical moments, express and The cross entropy function value between For monitoring point The average local Moran's index in the target scale, is the adjustment coefficient.

[0021] Since the local Moran's index of each monitoring point is different at the target scale, the larger the local Moran's index of a monitoring point at the target scale, the greater the correlation between the monitoring point and the adjacent monitoring points in the target scale, and the easier it is to obtain accurate monitoring results for the monitoring point. Therefore, when calculating the cross entropy loss, more attention is paid to the monitoring points with smaller local Moran's index at the target scale to ensure that the monitoring points with smaller local Moran's index at the target scale can obtain accurate monitoring results.

[0022] In the second aspect of the present application, there is also provided a mine ecological environment monitoring system based on image processing, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a mine ecological environment monitoring method based on image processing according to the first aspect of the present application is implemented.

[0023] The technical solution of this application has the following beneficial technical effects:

[0024] First, the monitored mine is divided into multiple monitoring areas, and the center point of each monitoring area is used as the monitoring point; the environmental complexity of each monitoring point is calculated based on the mining time and the information entropy of the monitoring data. The greater the environmental complexity, the greater the difficulty of the ecological environment monitoring results of the corresponding monitoring point; at the same time, the local Moran index of each monitoring point in multiple preset scales is calculated. If the monitoring point At the preset scale The larger the absolute value of the local Moran index is, the better the monitoring point is. At the preset scale There is a strong correlation within the preset scale Internal monitoring point When conducting ecological environment monitoring, the monitoring data of adjacent monitoring points can provide effective reference information and improve the The accuracy of the monitoring results is improved; further, according to the complexity of the environment, the local Moran's index of each monitoring point at the preset scale is weighted and summed to obtain the selection score of the preset scale, and N preset scales are selected in descending order of the selection score, and the proportion of the number of monitoring points whose local Moran's index is greater than the index threshold in the N preset scales is counted, and a proportion curve is drawn with N as the horizontal coordinate and the proportion of the number of monitoring points corresponding to N as the vertical coordinate; the inflection point of the proportion curve is determined, and the preset scale corresponding to the inflection point is used as the target scale to construct a convolutional neural network, so that the convolutional neural network can extract the information within the target scale of each monitoring point, thereby ensuring that the convolutional neural network can accurately output the monitoring results of each monitoring point. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present application will become easy to understand. In the accompanying drawings, several embodiments of the present application are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0026] Figure 1 is a flow chart of a method for monitoring a mine ecological environment based on image processing according to an embodiment of the present application;

[0027] Figure 2 is a schematic diagram of multiple monitoring points according to an embodiment of the present application;

[0028] Figure 3 It is a structural block diagram of a mine ecological environment monitoring system based on image processing according to an embodiment of the present application. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0030] It should be understood that when the terms "first", "second", etc. are used in the claims, specification and drawings of the present application, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the specification and claims of the present application indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their collections.

[0031] According to the first aspect of the present application, the present application provides a method for monitoring the ecological environment of a mine based on image processing. Figure 1 1 is a flow chart of a method for monitoring the ecological environment of a mine based on image processing according to an embodiment of the present application. Figure 1 As shown, the mine ecological environment monitoring method based on image processing includes steps S101 to S105, which are described in detail below.

[0032] S101, multiple monitoring points are set up in the monitoring mine, and monitoring data of each monitoring point at a historical moment is collected.

[0033] In one embodiment, the monitored mine occupies a large area. In order to accurately monitor the ecological environment of each location in the mine, it is necessary to set up multiple monitoring points in the monitored mine. One monitoring point corresponds to one monitoring area, and monitoring data of each monitoring point at a historical moment is collected. The monitoring data is any one of vegetation coverage rate, plant diversity index, soil nutrient content or water quality parameters.

[0034] Specifically, setting a plurality of monitoring points in the monitored mine includes: obtaining a plan of the monitored mine, dividing the plan into a plurality of monitoring areas of preset sizes, wherein the center point of each monitoring area corresponds to the plurality of monitoring points. The monitoring areas are all square areas of preset sizes. Figure 2 , is a schematic diagram of multiple monitoring points according to an embodiment of the present application.

[0035] In this way, multiple monitoring points are set up in the monitored mine, and monitoring data of each monitoring point at historical moments are collected. That is to say, at a historical moment, one monitoring point corresponds to one monitoring data, and the monitoring data includes multiple dimensions such as vegetation coverage rate, plant diversity index, soil nutrient content and water quality parameters.

[0036] S102, calculating the environmental complexity of each monitoring point and a plurality of local Moran indexes of preset scales, wherein the environmental complexity is positively correlated with the mining duration and the information entropy of the monitoring data at historical moments.

[0037] In one embodiment, the environmental complexity is used to reflect the difficulty of obtaining the ecological environment monitoring results of the corresponding monitoring point. In the process of mining, the duration of mining activities at each monitoring point is different. The longer the mining duration, the greater the impact of the mining activities on the ecological environment within the monitoring point, and the greater the environmental complexity of the monitoring point; at the same time, for a monitoring point, the monitoring data of multiple historical moments constitute the monitoring data time series of the monitoring point. The greater the information entropy of the monitoring data time series, the more the monitoring data of the monitoring point tends to show irregular changes, the more difficult it is to obtain the ecological environment monitoring results of the monitoring point, and the greater the environmental complexity of the monitoring point.

[0038] Specifically, the monitoring points The complexity of the environment for: , For the mining time, is the number of historical moments, is the information entropy; the mining duration is the number of moments in the mining process in all historical moments, and the information entropy is the arrangement entropy or sample entropy of a time series composed of monitoring data at multiple historical moments.

[0039] In this way, the environmental complexity of each monitoring point is calculated. The environmental complexity is used to accurately quantify the difficulty of obtaining the ecological environment monitoring results of the corresponding monitoring point. The more frequent the mining activities, the greater the impact of human factors on the corresponding monitoring point, and the greater the environmental complexity of the monitoring point; the greater the information entropy, the more chaotic the changes in the monitoring data of the corresponding monitoring point, and the greater the environmental complexity of the monitoring point.

[0040] In one embodiment, a local Moran index of a preset scale is used to measure the correlation between the monitoring data of a monitoring point and other monitoring points in the local area of ​​the preset scale; wherein the preset scale includes 3×3, 5×5, 7×7, 9×9, 11×11, 13×13 and 15×15, etc., and the local area of ​​the preset scale is a rectangular area centered on the monitoring point and having a size of the preset scale.

[0041] Monitoring Points Preset Scale The local Moran index for: , For monitoring point The standardized value of For preset scale The number of inner adjacent monitoring points, For monitoring point and adjacent monitoring points The spatial weight of Adjacent monitoring point The standardized value of is the sum of the spatial weights of all adjacent monitoring points; where the standardized value is the monitoring point Monitoring data minus preset scale The mean of the monitoring data of the adjacent monitoring points is divided by the standard deviation, and the spatial weight is To adjacent monitoring point The distance is negatively correlated.

[0042] Among them, the spatial weight Satisfies the relationship: , For monitoring point and adjacent monitoring points distance.

[0043] Understandably, monitoring points Preset Scale The local Moran index The value range is ,when When Monitoring data and preset scale The monitoring data of adjacent monitoring points are positively correlated. When Monitoring data and preset scale The monitoring data of adjacent monitoring points are negatively correlated, and the local Moran index The larger the absolute value of At the preset scale The stronger the correlation within When Monitoring data and preset scale There is no correlation between the monitoring data of adjacent monitoring points.

[0044] S103, calculating the selection score of each preset scale.

[0045] In one embodiment, one monitoring point corresponds to one environmental complexity and a plurality of local Moran indexes of preset sizes.

[0046] For monitoring points For example, if the preset scale The local Moran index The larger the value, the larger the preset size. Adjacent monitoring points and monitoring points within There is a strong correlation between the monitoring points When conducting ecological environment monitoring, the monitoring data of adjacent monitoring points can provide effective reference information, that is, the local Moran index Can be reflected in the preset scale Monitoring points obtained within Accuracy of ecological environment monitoring results, local Moran index The larger the preset scale Monitoring points obtained within The higher the accuracy of ecological environment monitoring results.

[0047] Therefore, when calculating the selection score of a preset scale, the larger the average local Moran index of all monitoring points under the preset scale, the larger the selection score of the preset scale. In addition, due to the different environmental complexity of each monitoring point, the environmental complexity of different monitoring points also needs to be considered when calculating the average local Moran index. The selection score for: , Including all monitoring points, For monitoring point The complexity of the environment, is the sum of the environmental complexity of all monitoring points, For monitoring point Preset Scale The local Moran index.

[0048] In the above calculation formula, The larger the value, the higher the monitoring point. The greater the environmental complexity, the more points should be selected when calculating the selected score. Assign greater weight; Indicates monitoring point At the preset scale The local Moran index, whose absolute value The larger the value, the smaller the preset scale. Monitoring points obtained within The higher the accuracy of ecological environment monitoring results, the higher the preset scale The larger the selection score.

[0049] In this way, the selection score of each preset scale is obtained. The larger the selection score, the more accurate the monitoring result obtained when monitoring the mine ecological environment under the preset scale.

[0050] S104, selecting N preset scales in descending order of the selected scores, counting the proportion of the number of monitoring points whose local Moran's index is greater than the index threshold in the N preset scales, and drawing a proportion curve with N as the horizontal axis and the proportion of the number of monitoring points corresponding to N as the vertical axis.

[0051] In one embodiment, the value of N is a positive integer. After selecting N preset scales in order of the selected scores from large to small, the proportion of monitoring points whose local Moran's index is greater than the index threshold is counted. The larger the proportion, the more monitoring points that can obtain accurate prediction results, where the value of the index threshold is 0.5.

[0052] Specifically, counting the proportion of the number of monitoring points whose local Moran's index is greater than the index threshold in the N preset scales includes: for any monitoring point, in response to the local Moran's index of any preset scale in the N preset scales being greater than the index threshold, marking the monitoring point; the proportion of the number of monitoring points is the proportion of the number of marked monitoring points in all monitoring points.

[0053] Adjust the value of N, and calculate the proportion of the number of monitoring points corresponding to different N to draw a proportion curve. The horizontal axis of the proportion curve is the different values ​​of N, and the vertical axis is the proportion of the number of monitoring points corresponding to N. Therefore, the proportion curve can reflect the change of the proportion of the number of monitoring points with N. As N increases, more preset scales are considered, which makes the proportion of the number of monitoring points gradually increase, that is, more monitoring points can obtain accurate monitoring results. However, as N increases, the amount of calculation will also increase. In order to balance the amount of calculation and the accuracy of the monitoring results, it is necessary to determine the target scale according to the inflection point of the proportion curve.

[0054] S105, determining the inflection point of the proportion curve, taking the preset scale corresponding to the inflection point as the target scale to construct a convolutional neural network, and training the convolutional neural network based on the monitoring data at historical moments, wherein the input of the neural network is the real-time monitoring data of each monitoring point at any moment, and the output is the monitoring result, which is normal or abnormal.

[0055] In one embodiment, the inflection point of the proportion curve is obtained, and all preset scales are screened according to the N value corresponding to the inflection point to obtain the preset scale corresponding to the inflection point, and the preset scale corresponding to the inflection point is used as the target scale, and the number of the target scale is at least one. For example, if the N value corresponding to the inflection point is 3, then 3 preset scales are selected in descending order of the selected scores to obtain the preset scale corresponding to the inflection point.

[0056] In one embodiment, the convolutional neural network includes an encoder and a decoder, the encoder includes multiple convolutional layers, and using the preset scale corresponding to the inflection point as the target scale to construct a convolutional neural network includes: obtaining the receptive field of each convolutional layer in the encoder, and using the convolutional layer whose receptive field is equal to the target scale as the target convolutional layer; using the output results of each target convolutional layer as the input of the decoder to output a monitoring result map, and the monitoring result map includes the monitoring results of each monitoring point.

[0057] Among them, the decoder includes multiple upsampling layers, and the output results of each target convolution layer are used as the input of the decoder to output the monitoring result graph, including: in response to the size of the output result of any upsampling layer being the same as the size of the output result of the target convolution layer, the output results of the target convolution layer and the upsampling layer are added and used as the input of the next adjacent upsampling layer.

[0058] Among them, the receptive field refers to the preset scale range that a value can perceive in the output result of the convolutional layer. In the convolutional neural network, as the convolutional layer deepens, the receptive field will gradually increase, which means that convolutional layers with a larger number of layers can capture information within a larger preset scale range of a monitoring point.

[0059] The encoder may use the feature extraction part of an existing convolutional neural network such as ResNet, DenseNet or VGGNet. For example, the encoder may use VGGNet. If the target scale is 5×5, 13×13 and 17×17, a total of three preset sizes, then the convolution layer conv1_2 with a receptive field of 5×5, the convolution layer conv3_4 with a receptive field of 13×13, and the convolution layer conv4_4 with a receptive field of 17×17 in VGGNet are obtained as target convolution layers, and the output result of the target convolution layer is used as the input of the decoder.

[0060] It can be understood that the input of the convolutional neural network is the real-time monitoring data of each monitoring point at any time, and the output is the monitoring results of each monitoring point; if the monitoring points are regarded as pixels, the process of outputting the monitoring results of each monitoring point using the convolutional neural network can be regarded as the process of classifying each pixel in the image using the semantic segmentation network. Therefore, the convolutional neural network can be regarded as a semantic segmentation network, and the encoder and decoder of the convolutional neural network can adopt existing semantic segmentation networks such as UNet and FCN.

[0061] At this point, the construction of the convolutional neural network is completed. The convolutional layers with different receptive fields in the encoder are screened through the target scale to determine the input of the decoder, so that the decoder can learn the neighborhood information of any monitoring point within the target scale, ensuring that each monitoring point can obtain accurate ecological environment monitoring results.

[0062] In one embodiment, after the construction of the convolutional neural network is completed, in order for the convolutional neural network to output accurate monitoring results, the convolutional neural network needs to be trained. Specifically, the training of the convolutional neural network based on the monitoring data at historical moments includes: obtaining the label information of each monitoring point at a historical moment, and the label information is normal or abnormal; inputting the monitoring data of each monitoring point at a historical moment into the convolutional neural network, outputting the monitoring results of each monitoring point at a historical moment, calculating the cross entropy loss of the label information and the monitoring results, and updating the convolutional neural network using the gradient descent method; iteratively updating the convolutional neural network until the cross entropy loss is less than the preset loss, and the training is completed.

[0063] Among them, the label information can be manually marked and can reflect the actual monitoring results of each monitoring point at a historical moment; the preset loss is 0.01.

[0064] The cross entropy loss for: , Including all monitoring points, For monitoring point Monitoring results at historical moments, For monitoring point Label information at historical moments, express and The cross entropy loss can be used to constrain the monitoring results output by the convolutional neural network to be consistent with the label information, so that the convolutional neural network can learn the mapping relationship between the monitoring results and the monitoring data, ensuring the accuracy of the monitoring results of each monitoring point.

[0065] In another embodiment, since the local Moran's index of each monitoring point at the target scale is different, the larger the local Moran's index of a monitoring point at the target scale, the greater the correlation between the monitoring point and the adjacent monitoring points in the target scale, and the easier it is for the monitoring point to obtain accurate monitoring results; to ensure that the monitoring point with a smaller local Moran's index at the target scale can obtain accurate monitoring results, therefore, when calculating the cross entropy loss, more attention should be paid to the monitoring point with a smaller local Moran's index at the target scale. Specifically, the cross entropy loss for: , Including all monitoring points, For monitoring point Monitoring results at historical moments, For monitoring point Label information at historical moments, express and The cross entropy function value between For monitoring point The average local Moran's index in the target scale, is the adjustment coefficient.

[0066] Among them, the adjustment coefficient is a minimum value to avoid The value of is 0. In the embodiment of the present application, the adjustment coefficient The value of is 0.01.

[0067] In this way, the training of the convolutional neural network is completed, and the trained convolutional neural network can be used to monitor the ecological environment of the mine; the real-time monitoring data of each monitoring point at any time is collected, and the real-time monitoring data of each monitoring point is input into the trained convolutional neural network, and the monitoring results of each monitoring point at that time are output, and the monitoring results are normal or abnormal.

[0068] The above specific embodiments introduce the technical principles and implementation details of a mine ecological environment monitoring method based on image processing of the present application. First, the monitored mine is divided into multiple monitoring areas, and the center point of each monitoring area is used as the monitoring point; the environmental complexity of each monitoring point is calculated based on the mining time and the information entropy of the monitoring data. The greater the environmental complexity, the greater the difficulty of the ecological environment monitoring result of the corresponding monitoring point; at the same time, the local Moran index of each monitoring point in multiple preset scales is calculated. If the monitoring point At the preset scale The larger the absolute value of the local Moran index is, the better the monitoring point is. At the preset scale There is a strong correlation within the preset scale Internal monitoring point When conducting ecological environment monitoring, the monitoring data of adjacent monitoring points can provide effective reference information and improve the The accuracy of the monitoring results is improved; further, according to the complexity of the environment, the local Moran's index of each monitoring point at the preset scale is weighted and summed to obtain the selection score of the preset scale, and N preset scales are selected in descending order of the selection score, and the proportion of the number of monitoring points whose local Moran's index is greater than the index threshold in the N preset scales is counted, and a proportion curve is drawn with N as the horizontal coordinate and the proportion of the number of monitoring points corresponding to N as the vertical coordinate; the inflection point of the proportion curve is determined, and the preset scale corresponding to the inflection point is used as the target scale to construct a convolutional neural network, so that the convolutional neural network can extract the information within the target scale of each monitoring point, thereby ensuring that the convolutional neural network can accurately output the monitoring results of each monitoring point.

[0069] According to the second aspect of the present application, the present application also provides a mine ecological environment monitoring system based on image processing. Figure 3 1 is a block diagram of a mine ecological environment monitoring system based on image processing according to an embodiment of the present application. Figure 3 As shown, the system 50 includes a processor and a memory, the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a method for monitoring the mining ecological environment based on image processing according to the first aspect of the present application is implemented. The system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, whose settings and functions are known in the art, so they are not described here.

[0070] In the present application, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in this application may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.

[0071] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0072] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the patent application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent application shall be subject to the attached claims.

Claims

1. A mine ecological environment monitoring method based on image processing, characterized in that: The monitoring method comprises: Set up multiple monitoring points in the monitoring mine and collect monitoring data of each monitoring point at historical moments; Calculate the environmental complexity of each monitoring point and the local Moran index of multiple preset scales; Calculate the selected scores of each preset scale, preset scale The selection score for: , Including all monitoring points, For monitoring point The complexity of the environment, is the sum of the environmental complexity of all monitoring points, For monitoring point Preset Scale The local Moran index of in, , For the mining time, is the number of historical moments, is the information entropy; the mining duration is the number of moments in the mining process among all historical moments, and the information entropy is the permutation entropy or sample entropy of the time series composed of monitoring data at multiple historical moments; , For monitoring point The standardized value of For preset scale The number of inner adjacent monitoring points, For monitoring point and adjacent monitoring points The spatial weight of Adjacent monitoring point The standardized value of is the sum of the spatial weights of all adjacent monitoring points. To adjacent monitoring point The distance is negatively correlated; Select N preset scales in descending order of the selected scores, count the proportion of monitoring points whose local Moran index is greater than the index threshold in the N preset scales, and draw a proportion curve with N as the horizontal axis and the proportion of monitoring points corresponding to N as the vertical axis; Determine the inflection point of the proportion curve, use the preset scale corresponding to the inflection point as the target scale to construct a convolutional neural network, and train the convolutional neural network based on the monitoring data at historical moments. The input of the convolutional neural network is the real-time monitoring data of each monitoring point at any time, and the output is the monitoring result of each monitoring point, which is normal or abnormal. In the process of training the convolutional neural network, the monitoring results output by the convolutional neural network are constrained by using the cross entropy loss to ensure the accuracy of the monitoring results; the cross entropy loss for: , For monitoring point Monitoring results at historical moments, For monitoring point Label information at historical moments, express and The cross entropy function value between For monitoring point The average local Moran's index in the target scale, is the adjustment coefficient.

2. The method for monitoring the ecological environment of a mine based on image processing according to claim 1, characterized in that: The percentage of monitoring points whose local Moran index is greater than the index threshold in N preset scales is counted as follows: For any monitoring point, in response to a local Moran index of any preset scale among the N preset scales being greater than an index threshold, marking the monitoring point; The monitoring point quantity ratio is the quantity ratio of the marked monitoring points to all monitoring points.

3. The method for monitoring the mine ecological environment based on image processing according to claim 1, characterized in that: The convolutional neural network includes an encoder and a decoder, the encoder includes a plurality of convolutional layers, and taking the preset scale corresponding to the inflection point as the target scale to construct the convolutional neural network includes: Get the receptive field of each convolutional layer in the encoder, and use the convolutional layer with a receptive field equal to the target scale as the target convolutional layer; The output result of each target convolution layer is used as the input of the decoder to output a monitoring result map, which includes the monitoring results of each monitoring point.

4. The method for monitoring the mine ecological environment based on image processing according to claim 3 is characterized in that: The decoder includes multiple upsampling layers, and the output results of each target convolution layer are used as the input of the decoder to output the monitoring result map including: In response to the size of the output result of any upsampling layer being the same as the size of the output result of the target convolution layer, the output results of the target convolution layer and the upsampling layer are added together and used as the input of the next adjacent upsampling layer.

5. The method for monitoring the mine ecological environment based on image processing according to claim 1, characterized in that: The convolutional neural network training based on monitoring data at historical moments includes: Obtaining label information of each monitoring point at a historical moment, wherein the label information is normal or abnormal; Inputting the monitoring data of each monitoring point at a historical moment into the convolutional neural network, outputting the monitoring results of each monitoring point at a historical moment, calculating the cross entropy loss of the label information and the monitoring results, and updating the convolutional neural network using a gradient descent method; The convolutional neural network is iteratively updated until the cross entropy loss is less than the preset loss, completing the training.

6. A mine ecological environment monitoring system based on image processing, characterized in that: It comprises a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a mine ecological environment monitoring method based on image processing according to any one of claims 1 to 5 is implemented.

Images