Method and device for identifying ground surface cracks in mining area and computer program product

Abstract

The invention provides a mining area earth surface crack identification method and device and a computer program product, and the method comprises the steps: obtaining a mining area earth surface crack image of a target mining area earth surface, and carrying out the cutting processing of the mining area earth surface crack image, and obtaining a plurality of earth surface sub-images, the target mining area earth surface being the mining area earth surface which is currently subjected to crack identification; sequentially inputting the earth surface sub-images into an earth surface crack recognition model to recognize cracks in the earth surface sub-images by using the earth surface crack recognition model, and outputting sub-crack recognition results corresponding to the earth surface sub-images; and splicing the sub-crack identification results corresponding to the ground surface sub-images to obtain a crack identification result of the ground surface of the target mining area. According to the method, the problem of low mining area surface crack identification accuracy in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of surface crack identification in mining areas, and in particular to a method for identifying surface cracks in mining areas, an identification device, a computer-readable storage medium and a computer program product. Background Art

[0002] At present, the monitoring of ground cracks in mining areas mainly relies on manual on-site observations, and the location of cracks is measured through facilities such as total stations and global positioning systems (GPS). This method is time-consuming and laborious, and the data is discrete and limited. In addition, surveys show that the width of surface cracks is often at the sub-meter level, which is difficult to observe through existing optical satellite images. UAV photogrammetry technology is flexible and low-cost, and can quickly obtain high-resolution images with centimeter-level accuracy, providing an ideal data source for dynamic monitoring of ground cracks in mining areas.

[0003] Extracting ground cracks from drone images involves image processing technology. At present, the mainstream crack processing methods include manual interpretation, threshold segmentation, edge detection, and machine learning-based methods. Although the traditional manual interpretation method is relatively accurate, it has subjective errors, high costs, and long processing time, and is not suitable for large-scale data processing. The threshold segmentation method distinguishes objects from backgrounds by setting pixel thresholds of images, but it is highly sensitive to image noise and has difficulty in processing changes in spectral and color features, and it is difficult to select a universally applicable threshold. The edge detection method uses the intensity changes of features such as color, light, shadow, and texture to extract the edges of objects. The Canny algorithm is a commonly used edge detection method that can distinguish edges in noisy environments and has good positioning capabilities for weak edges, but its generalization performance is poor. Machine learning methods based on feature engineering have been widely used in crack extraction. Although these techniques are more accurate than traditional image processing methods, they rely heavily on manually extracted features, and it is difficult to design universal crack features that are applicable to all coal mine sites. Summary of the invention

[0004] The main purpose of the present application is to provide a method, an identification device, a computer-readable storage medium and a computer program product for identifying surface cracks in a mining area, so as to at least solve the problem of low accuracy of crack identification on the surface of a mining area in the prior art.

[0005] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for identifying surface cracks in a mining area is provided, comprising: obtaining a surface crack image of a target mining area surface, and cropping the surface crack image of the mining area to obtain a plurality of surface sub-images, wherein the target mining area surface is the mining area surface currently undergoing crack identification; sequentially inputting the surface sub-images into a surface crack identification model, so as to use the surface crack identification model to identify cracks in the surface sub-images, and outputting sub-crack identification results corresponding to each of the surface sub-images, wherein the surface crack identification model is composed of an encoder, a decoder and a bottleneck. The decoder comprises at least a plurality of BasicBlock modules, a plurality of CBAM modules and an ASPP module. The decoder comprises at least the BasicBlock module and the CBAM module. The BasicBlock module is used to enhance the ability to extract crack features. The CBAM module is used to capture the correlation between the crack features in different dimensions. The ASPP module is used to extract information of the crack features at multiple scales. The sub-crack identification results corresponding to each of the surface sub-images are spliced ​​to obtain the crack identification results on the surface of the target mining area.

[0006] Optionally, the surface crack image of the mining area is cropped to obtain multiple surface sub-images, including: processing the surface crack image of the mining area to generate a whole orthophoto map within the surveying and mapping range; cropping the whole orthophoto map according to a set size to obtain multiple surface sub-images.

[0007] Optionally, before the surface sub-images are sequentially input into a surface crack recognition model to use the surface crack recognition model to identify cracks in the surface sub-images, and before the sub-crack recognition results corresponding to each surface sub-image are outputted, the method further includes: acquiring all historical surface images of the mining area; annotating each of the historical surface images at the pixel level, and generating a label image corresponding to each of the historical surface images; cropping the first surface image and the label image to obtain a second surface image and a label surface image, respectively, the first surface image being the annotated historical surface image, the second surface image being the cropped image of the first surface image, and the label surface image being the cropped image of the label image; screening the second surface image according to the label surface image to remove images in the second surface image where no surface cracks exist, to obtain a surface crack image data set.

[0008] Optionally, before the surface sub-images are sequentially input into a surface crack recognition model to use the surface crack recognition model to identify cracks in the surface sub-images, and before the sub-crack recognition results corresponding to each surface sub-image are outputted, the method further includes: constructing an initial surface crack recognition model based on a U-net network in combination with the BasicBlock module, the CBAM module and the ASPP module, wherein the initial surface crack recognition model is a model to be trained; and a training step of sequentially inputting input data of multiple groups of current training data into the initial surface crack recognition model for training, wherein the current training data is any group of training data in the surface crack image data set, and each group of the training data includes the input data and output data, the input data being a sample surface crack sub-image, and the output data being a sample crack recognition result corresponding to the sample surface crack sub-image; a testing step, testing the trained initial surface crack recognition model according to all the sample crack recognition results to obtain model accuracy; when the model accuracy does not meet the accuracy requirements, adjusting all parameters in the trained initial surface crack recognition model, and repeating the training step and the testing step at least once until the model accuracy meets the accuracy requirements to obtain the surface crack recognition model; when the model accuracy meets the accuracy requirements, determining the trained initial surface crack recognition model as the surface crack recognition model.

[0009] Optionally, the surface sub-images are sequentially input into a surface crack recognition model so as to utilize the surface crack recognition model to identify cracks in the surface sub-images, and sub-crack recognition results corresponding to each surface sub-image are output, including: subjecting the surface sub-images to feature extraction processing by the BasicBlock module and enhancement processing by the CBAM module in multiple encoding stages in the encoder to obtain enhanced feature maps in each encoding stage, the output of each encoding stage being used as input for the next encoding stage, and each encoding stage including a BasicBlock module and a CBAM module; decoding all the enhanced feature maps in the decoder in reverse acquisition order to obtain target decoded feature maps, and the reverse acquisition order is the opposite order of the acquisition order of each enhanced feature map in the encoding stage; inputting the target decoded feature map into the bottleneck layer, and outputting the sub-crack recognition results.

[0010] Optionally, the surface sub-image is subjected to feature extraction processing by the BasicBlock module of multiple encoding stages in the encoder and enhancement processing by the CBAM module to obtain an enhanced feature map of each encoding stage, including: processing the surface sub-image by the BasicBlock module of the first encoding stage to obtain a first encoding feature map, and inputting the first encoding feature map into the CBAM module for feature enhancement to obtain a first enhanced feature map, the first encoding stage being the first encoding stage in the encoder; processing the first enhanced feature map by the BasicBlock module of the second encoding stage to obtain a second encoding feature map, and inputting the second encoding feature map into the CBAM module for feature enhancement to obtain a first enhanced feature map. The block is feature enhanced to obtain a second enhanced feature map, and the second encoding stage is the second encoding stage of the encoder; the second enhanced feature map is processed by the BasicBlock module of the third encoding stage to obtain a third encoding feature map, and the third encoding feature map is input into the CBAM module for feature enhancement to obtain a third enhanced feature map, and the third encoding stage is the third encoding stage of the encoder; the third enhanced feature map is processed by the BasicBlock module of the fourth encoding stage to obtain a fourth encoding feature map, and the fourth encoding feature map is input into the CBAM module for feature enhancement to obtain a fourth enhanced feature map, and the fourth encoding stage is the fourth encoding stage of the encoder.

[0011] Optionally, all the enhanced feature maps are decoded in the decoder in the reverse acquisition order to obtain a target decoding feature map, including: inputting the fourth enhanced feature map into the ASPP module for multi-scale feature extraction, and sequentially performing convolution operations, acceleration operations and activation operations to obtain a first decoding feature map; sequentially performing the convolution operations, the acceleration operations and the activation operations on the third enhanced feature map to obtain a second decoding feature map; fusing the first decoding feature map and the second decoding feature map to obtain a first fused feature map; sequentially performing enhancement processing on the first fused feature map through the deconvolution layer, the BasicBlock module and the CBAM module to obtain a third decoding feature map; sequentially performing the convolution operations on the second enhanced feature map The fourth decoding feature map is subjected to the operation, the acceleration operation and the activation operation to obtain a fourth decoding feature map; the fourth decoding feature map and the third decoding feature map are fused to obtain a second fused feature map; the second fused feature map is subjected to the deconvolution layer, the BasicBlock module and the CBAM module for enhancement processing in sequence to obtain a fifth decoding feature map; the first enhanced feature map is subjected to the convolution operation, the acceleration operation and the activation operation in sequence to obtain a sixth decoding feature map; the fifth decoding feature map and the sixth decoding feature map are fused to obtain a third fused feature map; the third fused feature map is subjected to the deconvolution layer, the BasicBlock module and the CBAM module for enhancement processing in sequence to obtain the target decoding feature map.

[0012] According to another aspect of the present application, a device for identifying surface cracks in a mining area is provided, the device comprising: a first acquisition unit, used to acquire a surface crack image of a target mining area, and to crop the surface crack image to obtain a plurality of surface sub-images, wherein the target mining area surface is the surface of the mining area currently undergoing crack identification; an input-output unit, used to sequentially input the surface sub-images into a surface crack identification model, so as to identify cracks in the surface sub-images using the surface crack identification model, and output sub-crack identification results corresponding to each surface sub-image, wherein the surface crack identification model is composed of an encoder, a decoder, and a decoder. and a bottleneck layer, the decoder at least includes a plurality of BasicBlock modules, a plurality of CBAM modules and an ASPP module, the decoder at least includes the BasicBlock module and the CBAM module, the BasicBlock module is used to enhance the ability to extract crack features, the CBAM module is used to capture the correlation between the crack features in different dimensions, and the ASPP module is used to extract information of the crack features at multiple scales; a splicing unit is used to splice the sub-crack identification results corresponding to each of the surface sub-images to obtain the crack identification results on the surface of the target mining area.

[0013] According to another aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute any one of the methods described.

[0014] According to another aspect of the present application, a computer program product is provided, comprising computer instructions, wherein when the computer instructions are executed by a processor, any one of the methods described above is implemented.

[0015] By applying the technical solution of the present application, in a method for identifying surface cracks in a mining area, firstly, a surface crack image of the target mining area surface is obtained, and the surface crack image of the mining area is cropped to obtain a plurality of surface sub-images, wherein the target mining area surface is the mining area surface currently undergoing crack identification; then, the surface sub-images are sequentially input into a surface crack identification model, so as to identify cracks in the surface sub-images using the surface crack identification model, and output sub-crack identification results corresponding to each surface sub-image, wherein the surface crack identification model is composed of an encoder, a decoder and a bottleneck layer, wherein the decoder includes at least a plurality of BasicBlock modules, a plurality of CBAM modules and an ASPP module, and the decoder includes at least a BasicBlock module and a CBAM module, wherein the BasicBlock module is used to enhance the ability to extract crack features, the CBAM module is used to capture the correlation between crack features in different dimensions, and the ASPP module is used to extract multi-scale crack feature information; finally, the sub-crack identification results corresponding to each surface sub-image are spliced ​​to obtain the crack identification results of the target mining area surface. This application constructs a surface crack recognition model by combining the BasicBlock module, the CBAM module and the ASPP module, which can effectively extract and fuse feature information at multiple scales, thereby improving the accuracy of surface crack recognition in mining areas. The surface crack image of the mining area is cropped to obtain multiple surface sub-images, and each surface sub-image is input into the surface crack recognition model to output the mining area surface crack recognition result of each surface sub-image, that is, the sub-crack recognition result, and the mining area surface crack recognition result of each surface sub-image is spliced ​​to obtain a complete crack recognition result on the surface of the target mining area. This application solves the problem of low accuracy of crack recognition on the surface of mining areas in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A hardware structure block diagram of a mobile terminal for executing a method for identifying surface cracks in a mining area provided in an embodiment of the present application is shown;

[0017] Figure 2 A schematic flow chart of a method for identifying surface cracks in a mining area provided in accordance with an embodiment of the present application is shown;

[0018] Figure 3 A schematic diagram of crack delineation results of the coal mine ground using drone imaging provided in accordance with an embodiment of the present application is shown;

[0019] Figure 4 A schematic diagram of the structure of a surface crack identification model provided according to an embodiment of the present application is shown;

[0020] Figure 5 A schematic diagram of the structure of a BasicBlock module provided according to an embodiment of the present application is shown;

[0021] Figure 6 A schematic diagram of the structure of a CBAM module provided according to an embodiment of the present application is shown;

[0022] Figure 7 A schematic diagram of the structure of an ASPP module provided according to an embodiment of the present application is shown;

[0023] Figure 8 A structural block diagram of a device for identifying surface cracks in a mining area provided according to an embodiment of the present application is shown.

[0024] The above drawings include the following reference numerals:

[0025] 102, processor; 104, memory; 106, transmission device; 108, input and output devices. DETAILED DESCRIPTION

[0026] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0027] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only 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 ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0028] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0029] As introduced in the background technology, machine learning methods based on feature engineering in the prior art have been widely used in crack extraction. Although these technologies are more accurate than traditional image processing methods, they rely heavily on manually extracted features, and it is difficult to design universal crack features applicable to all coal mine sites, which leads to low crack identification accuracy. In order to solve the problem of low accuracy of crack identification on the surface of mining areas in the prior art, the embodiments of the present application provide a method for identifying surface cracks in mining areas, an identification device, a computer-readable storage medium and a computer program product.

[0030] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0031] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 FIG. 1 is a hardware structure block diagram of a mobile terminal of a method for identifying surface cracks in a mining area according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0032] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the display method of device information in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, the above method is implemented. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof. The transmission device 106 is used to receive or send data via a network. The above-mentioned specific examples of the network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0033] In this embodiment, a method for identifying surface cracks in a mining area that runs on a mobile terminal, a computer terminal or a similar computing device is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0034] Figure 2 FIG. 1 is a flow chart of a method for identifying surface cracks in a mining area according to an embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0035] Step S201, obtaining a surface crack image of a target mining area, and cropping the surface crack image to obtain a plurality of surface sub-images, wherein the target mining area surface is the mining area surface currently undergoing crack identification.

[0036] Specifically, the high-resolution visible light sensor carried by the drone is used to collect the surface crack images of the mining area in real time and pre-process them. Due to the limitation of computer hardware, the surface crack images of the mining area need to be cropped into multiple 256×256 pixel crack images, i.e., the surface sub-images mentioned above. Of course, other pixel sizes can also be used, which are specifically determined by the size that the surface crack recognition model can recognize.

[0037] Step S202, the above-mentioned surface sub-images are sequentially input into the surface crack recognition model, so as to use the above-mentioned surface crack recognition model to identify the cracks in the above-mentioned surface sub-images, and output the sub-crack recognition results corresponding to each of the above-mentioned surface sub-images. The above-mentioned surface crack recognition model is composed of an encoder, a decoder and a bottleneck layer. The above-mentioned decoder at least includes multiple BasicBlock modules, multiple CBAM modules and an ASPP module. The above-mentioned decoder at least includes the above-mentioned BasicBlock module and the above-mentioned CBAM module. The above-mentioned BasicBlock module is used to enhance the extraction capability of crack features, the above-mentioned CBAM module is used to capture the correlation between the above-mentioned crack features in different dimensions, and the above-mentioned ASPP module is used to extract information of the above-mentioned crack features at multiple scales.

[0038] Specifically, the above-mentioned surface crack recognition model is constructed by using the U-Net network in combination with the BasicBlock module, the CBAM module and the ASPP module. The surface crack recognition model includes a U-Net network encoder, a U-Net network decoder and a bottleneck layer. The U-Net network encoder is used to extract and compress the feature information in the input image and convert it into a high-level feature representation; the U-Net network decoder is used to use the high-level feature information extracted by the encoder to gradually restore the spatial dimension of the image through upsampling to generate a segmentation result. The U-Net network encoder replaces the original U-Net convolutional layer with a BasicBlock module, and each BasicBlock module consists of two convolutional layers. In order to avoid the surface crack recognition model from being stuck due to the large size of the input image, the above-mentioned surface sub-images are sequentially input into the surface crack recognition model to obtain the corresponding sub-crack recognition results.

[0039] Step S203, splicing the above-mentioned sub-crack identification results corresponding to each of the above-mentioned surface sub-images to obtain the crack identification results of the surface of the above-mentioned target mining area.

[0040] Specifically, the above-mentioned sub-crack identification results corresponding to each of the above-mentioned surface sub-images are spliced ​​together to obtain the complete crack identification results of the surface of the above-mentioned target mining area. Figure 3 As shown, Figure 3 Results of crack delineation on the ground of coal mines using drone imaging, including (a) cracks on the ground of coal mines, (b) cracks in mining areas, (c) cracks on concrete pavement, and (d) cracks in forests.

[0041] In this embodiment, firstly, a surface crack image of the target mining area is obtained, and the surface crack image of the mining area is cropped to obtain a plurality of surface sub-images, wherein the target mining area surface is the mining area surface currently undergoing crack identification; then, the surface sub-images are sequentially input into a surface crack identification model, so as to identify cracks in the surface sub-images using the surface crack identification model, and output sub-crack identification results corresponding to each surface sub-image. The surface crack identification model is composed of an encoder, a decoder and a bottleneck layer, wherein the decoder includes at least a plurality of BasicBlock modules, a plurality of CBAM modules and an ASPP module, and the decoder includes at least a BasicBlock module and a CBAM module, wherein the BasicBlock module is used to enhance the ability to extract crack features, the CBAM module is used to capture the correlation between crack features in different dimensions, and the ASPP module is used to extract multi-scale crack feature information; finally, the sub-crack identification results corresponding to each surface sub-image are spliced ​​to obtain the crack identification results of the target mining area surface. This application constructs a surface crack recognition model by combining the BasicBlock module, the CBAM module and the ASPP module, which can effectively extract and fuse feature information at multiple scales, thereby improving the accuracy of surface crack recognition in mining areas. The surface crack image of the mining area is cropped to obtain multiple surface sub-images, and each surface sub-image is input into the surface crack recognition model to output the mining area surface crack recognition result of each surface sub-image, that is, the sub-crack recognition result, and the mining area surface crack recognition result of each surface sub-image is spliced ​​to obtain a complete crack recognition result on the surface of the target mining area. This application solves the problem of low accuracy of crack recognition on the surface of mining areas in the prior art.

[0042] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the implementation process of the method for identifying surface cracks in mining areas of the present application will be described in detail below in combination with specific embodiments.

[0043] In order for the surface crack identification model to accurately identify cracks, in an optional implementation, the above step S201 includes:

[0044] Step S2011, processing the surface crack image of the mining area to generate an orthophoto map of the entire surveying range;

[0045] Step S2012, cropping the entire orthophoto image according to a set size to obtain a plurality of surface sub-images.

[0046] In the above embodiment, the surface crack image of the above-mentioned mining area is preprocessed, and the entire orthophoto map within the range of drone mapping is generated using software such as PIX4D; the entire orthophoto map is cropped into a set size (such as: 256×256 pixels) that can be recognized by the surface crack recognition model to obtain multiple surface sub-images mentioned above.

[0047] In order to improve the efficiency of crack identification, in an optional implementation manner, before the above step S202, the method further includes:

[0048] Step S301, obtaining all historical surface images of the above-mentioned mining area;

[0049] Step S302, performing pixel-level annotation on each of the above historical surface images, and generating a label image corresponding to each of the above historical surface images;

[0050] Step S303, cropping the first surface image and the label image to obtain a second surface image and a label surface image respectively, wherein the first surface image is the annotated historical surface image, the second surface image is the cropped image of the first surface image, and the label surface image is the cropped image of the label image;

[0051] Step S304: screening the second surface image according to the labeled surface image to remove images without surface cracks in the second surface image, thereby obtaining a surface crack image dataset.

[0052] In the above embodiment, it is necessary to obtain a training data set before training the surface crack recognition model. Therefore, based on the surface crack data source, combined with visual interpretation, the Labelme annotation tool is used to annotate all historical surface images at the pixel level to generate label images corresponding to each historical surface image; due to the limitation of computer hardware, the annotated historical surface images and their corresponding label images need to be cropped to 256×256 pixel surface images (i.e., the above-mentioned second surface image); the file names of the cropped surface images and the label images can be kept consistent to correspond one to one, so as to facilitate subsequent model training; in addition, all image data are overlapped and cropped, and the overlap rate can be set to 25% to expand the data set. Since remote sensing image recognition is different from natural image recognition, it is more convenient to draw labels for objects on large-scale uncut images. There are a large number of plot images without surface cracks in the cropped surface images. In order to avoid loss of precision after cropping, the cropped surface images can be screened according to the labels, and only the plot images with surface cracks and the corresponding label images can be retained to obtain the surface crack image data set. In order to improve the robustness and generalization of the designed model, that is, to adapt to the identification of surface cracks in mining areas under complex environments, data enhancement operations are performed on the mining area surface cracks dataset, including adjusting the image contrast, rotation angle, and adding Gaussian noise.

[0053] In order to improve the accuracy of identifying surface cracks in a mining area, in an optional implementation manner, before the above step S202, the method further includes:

[0054] Step S401, constructing an initial surface crack identification model based on a U-net network in combination with the BasicBlock module, the CBAM module and the ASPP module, wherein the initial surface crack identification model is a model to be trained;

[0055] Step S402, a training step, sequentially inputting input data of multiple sets of current training data into the initial surface crack recognition model for training, wherein the current training data is any set of training data in the surface crack image data set, and each set of the training data includes the input data and output data, wherein the input data is a sample surface crack sub-image, and the output data is a sample crack recognition result corresponding to the sample surface crack sub-image;

[0056] Step S403, a testing step, testing the trained initial surface crack recognition model according to all the sample crack recognition results to obtain the model accuracy;

[0057] Step S404, when the accuracy of the above-mentioned model does not meet the accuracy requirement, all parameters in the above-mentioned initial surface crack identification model after training are adjusted, and the above-mentioned training step and the above-mentioned testing step are repeated at least once until the accuracy of the above-mentioned model meets the above-mentioned accuracy requirement, thereby obtaining the above-mentioned surface crack identification model;

[0058] Step S405: When the accuracy of the model meets the accuracy requirement, the trained initial ground crack identification model is determined as the ground crack identification model.

[0059] In the above embodiment, the improved U-Net network consists of three parts: encoder, decoder and bottleneck layer; encoder, in each encoding stage, BasicBlock module is used for feature extraction, and CBAM module is inserted before downsampling for feature enhancement; the bottom layer uses ASPP module for multi-scale feature extraction and enhanced feature representation capability. Decoder, in each decoding stage, upsampling and BasicBlock modules are used to restore spatial resolution, and CBAM module is inserted after upsampling for feature enhancement; through jump connection, feature maps of decoding stage and encoding stage are fused to retain detail information; finally, pixel classification is completed through bottleneck layer 1×1 convolution. First, an initial surface crack recognition model to be trained is preliminarily constructed, and multiple groups of current training data input data are sequentially input into the above initial surface crack recognition model for training. After the current training is completed, all the above sample crack recognition results and the crack recognition results output by the training process are compared and judged to obtain the model accuracy. If the model accuracy does not meet the accuracy requirements, all parameters in the model are adjusted and retrained until the model accuracy meets the accuracy requirements, and the final model is determined as the above surface crack recognition model. Thereby improving the accuracy of surface crack identification in mining areas and reducing labor costs.

[0060] In order to improve the feature extraction capability and semantic understanding capability of the model, in an optional implementation manner, the above step S202 includes:

[0061] Step S2021, subjecting the surface sub-image to feature extraction processing by the BasicBlock module and enhancement processing by the CBAM module in the encoder at multiple encoding stages to obtain an enhanced feature map of each encoding stage, wherein the output of each encoding stage is used as the input of the next encoding stage, and each encoding stage includes one BasicBlock module and one CBAM module;

[0062] Step S2022, decoding all the enhanced feature maps in the decoder in a reverse acquisition order to obtain a target decoded feature map, wherein the reverse acquisition order is an order opposite to the acquisition order of each of the enhanced feature maps in the encoding stage;

[0063] Step S2022, input the target decoding feature map into the bottleneck layer, and output the sub-crack identification result.

[0064] In the above embodiment, if Figure 4As shown in the figure, the surface sub-image is input into the surface crack recognition model through the input layer. In the encoder, it undergoes feature processing in four encoding stages. Each encoding stage is processed by the residual block structure, CBAM module and downsampling block for feature enhancement. Then, it is processed by the ASPP module to realize multi-scale feature extraction. After that, it enters the decoder and undergoes feature processing in four decoding stages. Each decoding stage is processed by the residual block structure, upsampling block and CBAM module in turn to restore the spatial resolution of the image and enhance the features. Finally, the image enters the bottleneck layer to perform a 1×1 convolution operation to output the crack recognition result. By combining convolutions with different void rates and global features, multi-scale information is effectively captured, enabling the network to better handle targets of different scales, and improving the model's feature extraction and semantic understanding capabilities.

[0065] It should also be noted that Figure 4 The following is a schematic diagram of the structure of the BasicBlock module. The specific working principle includes: after the input feature map passes through two convolutional layers and the ReLu function, it is added to the original input feature to obtain the final output feature map. A residual block can be expressed as: l+1 =x l +f(x l , w l ), the residual block is divided into two parts: the direct mapping part and the residual part, x l is the direct mapping part; f(x l , w l ) is the residual part, which consists of two convolution operations, corresponding to Figure 5 The left side of the figure contains the convolution part. The shortcut connection between the input feature map and the output feature map transfers the crack information extracted by the previous layer of the network to the next layer, which avoids information loss to a greater extent and effectively prevents the network degradation caused by increasing the number of neural network layers.

[0066] Figure 6 The schematic diagram of the CBAM module is shown in Figure 2. Its specific working principle includes: Assume that F = C × H × W is the input feature map, C refers to the number of channels of the image, H represents the number of pixels in the vertical dimension of the image, and W represents the number of pixels in the horizontal dimension of the image. The CBAM module calculates the one-dimensional channel attention feature map M in turn. c ∈C×1×1 and a two-dimensional spatial attention feature map M S ∈1×H×W, and finally output the weighted features with channels and spaces. Channel attention feature F′=M C (F)×F, spatial attention feature F″=M S (F′)×F′, where F′ represents the input feature map after the one-dimensional channel attention operation, and F″ is the final refined output. The structure of the above channel attention module is as follows Figure 6As shown in a; the input feature map F is subjected to global maximum pooling and global average pooling respectively to obtain the spatial attention feature map M S ∈C×1×1 feature map, which are the global average pooling features And the global maximum pooling feature After two layers of fully connected neural networks, the number of neurons in the first layer is (C is the number of image channels, r is the compression rate). ReLu is the activation function, the number of neurons in the second layer is equal to the number of channels C, the output features of the fully connected neural network are summed, and the channel attention feature map M is generated through the sigmoid activation function. C , the channel attention feature image formula is: Among them, σ represents the sigmoid function, W0 and W1 are the weights of the channel attention module, which are the input layer and output layer weights in the multi-layer perceptron feedforward neural network. The structure of the above spatial attention module is as follows Figure 6 As shown in b in the figure; the spatial attention is input into the feature map F′=C×H×W for average pooling and maximum pooling to obtain the global average pooling feature F′ avg And the global maximum pooling feature F′ max Then, for F′ avg and F′ max The two feature maps are concatenated and compressed to H×W×1 after a 7×7 convolution operation. The spatial attention feature map M is generated by the sigmoid activation function. S Finally, the output feature map of the spatial attention module is multiplied by the input feature map to obtain the final generated feature map. The calculation formula of the feature map of the spatial attention module is: S (F′)=σ{f 7×7 [(F′ avg +F′ max )]}where σ represents the sigmoid function, f 7×7 represents a convolution operation with a filter size of 7×7.

[0067] Figure 7The ASPP module is a structural diagram of the ASPP module. Its specific working principle includes: the above ASPP module is essentially composed of a 1×1 convolution + pooling pyramid + ASPP Pooling. The expansion factor of each layer of the pooling pyramid can be customized, thereby realizing free multi-scale feature extraction. When the ASPP module processes the input feature map, the specific steps include: 1) using 3×3 convolution kernels with different dilation rates (6, 12 and 18) to perform dilated convolution operations. Multi-scale dilated convolution includes: using a 3×3 convolution kernel with a dilation rate of 6 to perform dilated convolution operations to expand the receptive field without increasing the resolution of the feature map; using a 3×3 convolution kernel with a dilation rate of 12 to perform dilated convolution operations to further expand the receptive field; using a 3×3 convolution kernel with a dilation rate of 18 to perform dilated convolution operations to maximize the capture of long-distance context information; 2) applying 1×1 convolution at the same time to reduce the number of channels and computational complexity while keeping the spatial size unchanged; 3) obtaining global features through global average pooling, performing global average pooling on the input feature map, and generating a global feature vector of size 1×1×C. The global feature vector is convolved using 1×1 convolution to reduce the number of channels, and the result is restored to the size of the original feature map H×W×C through bilinear interpolation; Feature fusion: All feature maps generated by the ASPP module in the above steps of processing the input feature map are connected in the channel dimension to form a fused feature map with a size of H×W×(4C), that is, it contains five feature maps, each with C channels; 1×1 convolution adjusts the output: The number of channels of the connected feature map is adjusted back to C through a 1×1 convolution, and BN and ReLU activation operations are performed. The size of the final output feature map is H×W×C, the same as the number of channels of the input feature map; 1×1 convolution is applied to the fused feature map to adjust the number of output channels and reduce the computational complexity; Use activation functions (such as ReLU) for nonlinear transformation to enhance the feature expression ability.

[0068] In order to fully extract the global feature information and local feature information of the surface sub-image to enhance the accuracy of crack identification, in an optional implementation, the above step S2021 includes:

[0069] Step S20211, processing the surface sub-image through the BasicBlock module of the first encoding stage to obtain a first encoding feature map, and inputting the first encoding feature map into the CBAM module for feature enhancement to obtain a first enhanced feature map, wherein the first encoding stage is the first encoding stage in the encoder;

[0070] Step S20212, processing the first enhanced feature map through the BasicBlock module of the second encoding stage to obtain a second encoded feature map, and inputting the second encoded feature map into the CBAM module for feature enhancement to obtain a second enhanced feature map, wherein the second encoding stage is the second encoding stage of the encoder;

[0071] Step S20213, processing the second enhanced feature map through the BasicBlock module of the third encoding stage to obtain a third encoded feature map, and inputting the third encoded feature map into the CBAM module for feature enhancement to obtain a third enhanced feature map, wherein the third encoding stage is the third encoding stage of the encoder;

[0072] Step S20214, processing the above-mentioned third enhanced feature map through the above-mentioned BasicBlock module of the fourth encoding stage to obtain a fourth encoded feature map, and inputting the above-mentioned fourth encoded feature map into the above-mentioned CBAM module for feature enhancement to obtain a fourth enhanced feature map. The above-mentioned fourth encoding stage is the fourth encoding stage of the above-mentioned encoder.

[0073] In the above embodiment, Figure 3 This is the structural diagram of the surface crack identification model. Figure 3 The surface sub-image is divided into sub-patches by the BasicBlock module of the first residual block structure from top to bottom on the left side of the figure, and the first coded feature map F1 is quasi-enhanced by the CBAM module of the first coding stage to obtain the first enhanced feature map E1; the first enhanced feature map E1 is quasi-enhanced by the CBAM module of the second coding stage (i.e., Figure 3 The second encoding feature map F2 is processed by the BasicBlock module of the second residual block structure from top to bottom on the left side of the figure, and the second encoding feature map F2 is processed by the CBAM module of the second encoding stage to obtain the second enhanced feature map E2 with a size of The second enhanced feature map E2 passes through the third encoding stage (i.e. Figure 3 The BasicBlock module of the third residual block structure from top to bottom on the left side of the figure obtains the third encoded feature map F3, and the third encoded feature map F3 obtains the third enhanced feature map E3 through the CBAM module of the third encoding stage, with a size of The third enhanced feature map E3 passes through the fourth encoding stage (i.e. Figure 3 The fourth encoding feature map F4 is obtained by the BasicBlock module of the fourth residual block structure from top to bottom on the left side of the figure. The fourth encoding feature map F4 is obtained by the CBAM module of the fourth encoding stage to obtain the fourth enhanced feature map E4, with a size of At this point, the encoding is completed and the high-level feature information in the above-mentioned surface sub-image is extracted.

[0074] In order to fuse the multi-scale information in the image and effectively retain the feature detail information, in an optional implementation manner, the above step S2022 includes:

[0075] Step S20220, inputting the fourth enhanced feature map into the ASPP module for multi-scale feature extraction, and sequentially performing convolution operation, acceleration operation and activation operation to obtain a first decoding feature map;

[0076] Step S20221, sequentially performing the convolution operation, the acceleration operation, and the activation operation on the third enhanced feature map to obtain a second decoding feature map;

[0077] Step S20222, fusing the first decoding feature map and the second decoding feature map to obtain a first fused feature map;

[0078] Step S20223, performing enhancement processing on the first fusion feature map in sequence through the deconvolution layer, the BasicBlock module and the CBAM module to obtain a third decoding feature map;

[0079] Step S20224, sequentially performing the convolution operation, the acceleration operation, and the activation operation on the second enhanced feature map to obtain a fourth decoding feature map;

[0080] Step S20225, fusing the fourth decoding feature map and the third decoding feature map to obtain a second fused feature map;

[0081] Step S20226, performing enhancement processing on the second fused feature map through the deconvolution layer, the BasicBlock module and the CBAM module in sequence to obtain a fifth decoding feature map;

[0082] Step S20227, sequentially performing the convolution operation, the acceleration operation, and the activation operation on the first enhanced feature map to obtain a sixth decoding feature map;

[0083] Step S20228, fusing the fifth decoding feature map and the sixth decoding feature map to obtain a third fused feature map;

[0084] Step S20229, the third fusion feature map is enhanced by the deconvolution layer, the BasicBlock module and the CBAM module in sequence to obtain the target decoding feature map.

[0085] In the above embodiment, the fourth enhanced feature map E4 is subjected to multi-scale feature extraction through the ASPP module to obtain the ASPP feature map, and the ASPP feature map A is subjected to a 1×1 convolution layer (i.e., the above convolution operation), and then the BN (i.e., the above acceleration operation) and the ReLU operation (i.e., the above activation operation) are performed to obtain the first decoding feature map G1. The third enhanced feature map E3 is subjected to a 1×1 convolution layer, and then the BN and ReLU operations are performed to obtain the second decoding feature map G2, which is then fused with the first decoding feature map G1 to obtain the first fused feature map. After the first fused feature map passes through 1 layer of deconvolution layer (also called upsampling layer) and BasicBlock module, BN and ReLU operations are performed, and feature enhancement is performed through the CBAM module to obtain the third decoding feature map G3; after the second enhanced feature map E2 passes through 1 layer of 1×1 convolution layer, BN and ReLU operations are performed to obtain the fourth decoding feature map G4, and it is fused with the third decoding feature map G3 to obtain the second fused feature map; after the second fused feature map passes through 1 layer of 1 deconvolution layer (or upsampling layer) and BasicBlock module, BN and ReLU operations are performed, and feature enhancement is performed through the CBAM module to obtain the fifth decoding feature map G5; after the first enhanced feature map E1 passes through 1 layer of 1×1 convolution layer, BN and ReLU operations are performed to obtain the sixth decoding feature map G6, and it is fused with the fifth decoding feature map G5 to obtain the third fused feature map. After the third fusion feature map passes through 1 upsampling layer and BasicBlock module, BN and ReLU operations are performed, and feature enhancement is performed through CBAM module to obtain the above target decoding feature map G7. The decoding process is now completed, corresponding to Figure 3 The four residual block structures from bottom to top on the right. One residual block structure corresponds to one decoding stage. In each decoding stage, the upsampling layer and BasicBlock module are used to restore the spatial resolution. After upsampling, the CBAM module is inserted for feature enhancement. Through the skip connection, the feature maps of the decoding stage and the encoding stage are fused to retain the detail information.

[0086] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0087] The embodiment of the present application also provides a device for identifying surface cracks in a mining area. It should be noted that the device for identifying surface cracks in a mining area in the embodiment of the present application can be used to execute the identification method for surface cracks in a mining area provided in the embodiment of the present application. The device is used to implement the above-mentioned embodiments and preferred implementation modes, and those that have been described will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0088] The following is an introduction to the device for identifying surface cracks in a mining area provided in an embodiment of the present application.

[0089] Figure 8 is a structural block diagram of a device for identifying surface cracks in a mining area according to an embodiment of the present application. Figure 8 As shown, the device comprises:

[0090] The first acquisition unit 10 is used to acquire a surface crack image of a target mining area, and to crop the surface crack image to obtain a plurality of surface sub-images. The target mining area surface is the mining area surface currently undergoing crack identification.

[0091] Specifically, the high-resolution visible light sensor carried by the drone is used to collect the surface crack images of the mining area in real time and pre-process them. Due to the limitation of computer hardware, the surface crack images of the mining area need to be cropped into multiple 256×256 pixel crack images, i.e., the surface sub-images mentioned above. Of course, other pixel sizes can also be used, which are specifically determined by the size that the surface crack recognition model can recognize.

[0092] The input-output unit 20 is used to input the above-mentioned surface sub-images into the surface crack recognition model in sequence, so as to use the above-mentioned surface crack recognition model to identify the cracks in the above-mentioned surface sub-images, and output the sub-crack recognition results corresponding to each of the above-mentioned surface sub-images. The above-mentioned surface crack recognition model is composed of an encoder, a decoder and a bottleneck layer. The above-mentioned decoder at least includes multiple BasicBlock modules, multiple CBAM modules and an ASPP module. The above-mentioned decoder at least includes the above-mentioned BasicB1ock module and the above-mentioned CBAM module. The above-mentioned BasicBlock module is used to enhance the ability to extract crack features. The above-mentioned CBAM module is used to capture the correlation between the above-mentioned crack features in different dimensions. The above-mentioned ASPP module is used to extract information of the above-mentioned crack features at multiple scales.

[0093] Specifically, the above-mentioned surface crack recognition model is constructed by using the U-Net network in combination with the BasicBlock module, the CBAM module and the ASPP module. The surface crack recognition model includes a U-Net network encoder, a U-Net network decoder and a bottleneck layer. The U-Net network encoder is used to extract and compress the feature information in the input image and convert it into a high-level feature representation; the U-Net network decoder is used to use the high-level feature information extracted by the encoder to gradually restore the spatial dimension of the image through upsampling to generate a segmentation result. The U-Net network encoder replaces the original U-Net convolutional layer with a BasicBlock module, and each BasicBlock module consists of two convolutional layers. In order to avoid the surface crack recognition model from being stuck due to the large size of the input image, the above-mentioned surface sub-images are sequentially input into the surface crack recognition model to obtain the corresponding sub-crack recognition results.

[0094] The splicing unit 30 is used to splice the above-mentioned sub-crack identification results corresponding to each of the above-mentioned surface sub-images to obtain the crack identification results of the surface of the above-mentioned target mining area.

[0095] Specifically, the above-mentioned sub-crack identification results corresponding to each of the above-mentioned surface sub-images are spliced ​​together to obtain the complete crack identification results of the surface of the above-mentioned target mining area. Figure 3 As shown, Figure 3 Results of crack delineation on the ground of coal mines using drone imaging, including (a) cracks on the ground of coal mines, (b) cracks in mining areas, (c) cracks on concrete pavement, and (d) cracks in forests.

[0096] In this embodiment, the first acquisition unit is used to acquire a surface crack image of the target mining area, and to crop the surface crack image of the mining area to obtain multiple surface sub-images. The target mining area surface is the surface of the mining area currently undergoing crack identification. The input and output unit is used to sequentially input the surface sub-images into the surface crack identification model, so as to identify cracks in the surface sub-images using the surface crack identification model, and output sub-crack identification results corresponding to each surface sub-image. The surface crack identification model is composed of an encoder, a decoder, and a bottleneck layer. The decoder includes at least a plurality of BasicBlock modules, a plurality of CBAM modules, and an ASPP module. The decoder includes at least a BasicBlock module and a CBAM module. The BasicBlock module is used to enhance the ability to extract crack features. The CBAM module is used to capture the correlation between crack features in different dimensions. The ASPP module is used to extract multi-scale crack feature information. The splicing unit is used to splice the sub-crack identification results corresponding to each surface sub-image to obtain the crack identification result of the target mining area surface. This application constructs a surface crack recognition model by combining the BasicBlock module, the CBAM module and the ASPP module, which can effectively extract and fuse feature information at multiple scales, thereby improving the accuracy of surface crack recognition in mining areas. The surface crack image of the mining area is cropped to obtain multiple surface sub-images, and each surface sub-image is input into the surface crack recognition model to output the mining area surface crack recognition result of each surface sub-image, that is, the sub-crack recognition result, and the mining area surface crack recognition result of each surface sub-image is spliced ​​to obtain a complete crack recognition result on the surface of the target mining area. This application solves the problem of low accuracy of crack recognition on the surface of mining areas in the prior art.

[0097] In order for the surface crack identification model to accurately identify cracks, in an optional implementation manner, the first acquisition unit includes:

[0098] A processing module processes the surface crack image of the mining area to generate an orthophoto map of the entire area within the surveying and mapping range;

[0099] The cropping module crops the entire orthophoto image according to a set size to obtain the plurality of surface sub-images.

[0100] In the above embodiment, the surface crack image of the above-mentioned mining area is preprocessed, and the entire orthophoto map within the range of drone mapping is generated using software such as PIX4D; the entire orthophoto map is cropped into a set size (such as: 256×256 pixels) that can be recognized by the surface crack recognition model to obtain multiple surface sub-images mentioned above.

[0101] In order to improve the efficiency of crack identification, in an optional implementation manner, the device further includes:

[0102] A second acquisition unit is used to acquire all historical surface images of the surface of the mining area before sequentially inputting the surface sub-images into a surface crack recognition model to identify cracks in the surface sub-images using the surface crack recognition model and outputting sub-crack recognition results corresponding to each of the surface sub-images;

[0103] A labeling unit, used to perform pixel-level labeling on each of the above historical surface images and generate a label image corresponding to each of the above historical surface images;

[0104] a cropping unit, used for cropping the first surface image and the label image to obtain a second surface image and a label surface image respectively, wherein the first surface image is the annotated historical surface image, the second surface image is the cropped image of the first surface image, and the label surface image is the cropped image of the label image;

[0105] The screening unit is used to screen the second surface image according to the labeled surface image to remove images without surface cracks in the second surface image, so as to obtain a surface crack image data set.

[0106] In the above embodiment, it is necessary to obtain a training data set before training the surface crack recognition model. Therefore, based on the surface crack data source, combined with visual interpretation, the Labelme annotation tool is used to annotate all historical surface images at the pixel level to generate label images corresponding to each historical surface image; due to the limitation of computer hardware, the annotated historical surface images and their corresponding label images need to be cropped to 256×256 pixel surface images (i.e., the above-mentioned second surface image); the file names of the cropped surface images and the label images can be kept consistent to correspond one to one, so as to facilitate subsequent model training; in addition, all image data are overlapped and cropped, and the overlap rate can be set to 25% to expand the data set. Since remote sensing image recognition is different from natural image recognition, it is more convenient to draw labels for objects on large-scale uncut images. There are a large number of plot images without surface cracks in the cropped surface images. In order to avoid loss of precision after cropping, the cropped surface images can be screened according to the labels, and only the plot images with surface cracks and the corresponding label images can be retained to obtain the surface crack image data set. In order to improve the robustness and generalization of the designed model, that is, to adapt to the identification of surface cracks in mining areas under complex environments, data enhancement operations are performed on the mining area surface cracks dataset, including adjusting the image contrast, rotation angle, and adding Gaussian noise.

[0107] In order to improve the accuracy of identifying surface cracks in mining areas, in an optional implementation, the device further includes:

[0108] A construction unit is used to construct an initial surface crack recognition model based on a U-net network in combination with the BasicBlock module, the CBAM module and the ASPP module before sequentially inputting the surface sub-images into the surface crack recognition model to identify cracks in the surface sub-images using the surface crack recognition model and outputting sub-crack recognition results corresponding to each of the surface sub-images, wherein the initial surface crack recognition model is a model to be trained;

[0109] A training unit, used to execute the training step, inputting the input data of multiple groups of current training data into the above-mentioned initial surface crack recognition model in sequence for training, wherein the above-mentioned current training data is any group of training data in the above-mentioned surface crack image data set, and each group of the above-mentioned training data includes the above-mentioned input data and output data, wherein the above-mentioned input data is a sample surface crack sub-image, and the above-mentioned output data is a sample crack recognition result corresponding to the above-mentioned sample surface crack sub-image;

[0110] A testing unit, used to execute a testing step, and test the trained initial surface crack recognition model according to all the sample crack recognition results to obtain the model accuracy;

[0111] A repeating unit, used for adjusting all parameters in the trained initial surface crack identification model when the accuracy of the model does not meet the accuracy requirement, and repeating the training step and the testing step at least once until the accuracy of the model meets the accuracy requirement, thereby obtaining the surface crack identification model;

[0112] The determination unit is used to determine the trained initial surface crack identification model as the surface crack identification model when the accuracy of the model meets the accuracy requirement.

[0113] In the above embodiment, the improved U-Net network consists of three parts: encoder, decoder and bottleneck layer; encoder, in each encoding stage, BasicBlock module is used for feature extraction, and CBAM module is inserted before downsampling for feature enhancement; the bottom layer uses ASPP module for multi-scale feature extraction and enhanced feature representation capability. Decoder, in each decoding stage, upsampling and BasicBlock modules are used to restore spatial resolution, and CBAM module is inserted after upsampling for feature enhancement; through jump connection, feature maps of decoding stage and encoding stage are fused to retain detail information; finally, pixel classification is completed through bottleneck layer 1×1 convolution. First, an initial surface crack recognition model to be trained is preliminarily constructed, and multiple groups of current training data input data are sequentially input into the above initial surface crack recognition model for training. After the current training is completed, all the above sample crack recognition results and the crack recognition results output by the training process are compared and judged to obtain the model accuracy. If the model accuracy does not meet the accuracy requirements, all parameters in the model are adjusted and retrained until the model accuracy meets the accuracy requirements, and the final model is determined as the above surface crack recognition model. Thereby improving the accuracy of surface crack identification in mining areas and reducing labor costs.

[0114] In order to improve the feature extraction and semantic understanding capabilities of the model, in an optional implementation, the input and output unit includes:

[0115] A first processing unit is used to perform feature extraction processing on the BasicBlock module and enhancement processing on the CBAM module of the surface sub-image in the encoder after multiple encoding stages, so as to obtain an enhanced feature map of each encoding stage, wherein the output of each encoding stage is used as the input of the next encoding stage, and each encoding stage includes one BasicBlock module and one CBAM module;

[0116] A second processing unit is used to decode all the enhanced feature maps in the decoder in a reverse acquisition order to obtain a target decoded feature map, wherein the reverse acquisition order is an order opposite to the acquisition order of each of the enhanced feature maps in the encoding stage;

[0117] The output unit is used to input the target decoding feature map into the bottleneck layer, and output the sub-crack identification result.

[0118] In the above embodiment, if Figure 4As shown in the figure, the surface sub-image is input into the surface crack recognition model through the input layer. In the encoder, it undergoes feature processing in four encoding stages. Each encoding stage is processed by the residual block structure, CBAM module and downsampling block for feature enhancement. Then, it is processed by the ASPP module to realize multi-scale feature extraction. After that, it enters the decoder and undergoes feature processing in four decoding stages. Each decoding stage is processed by the residual block structure, upsampling block and CBAM module in turn to restore the spatial resolution of the image and enhance the features. Finally, the image enters the bottleneck layer to perform a 1×1 convolution operation to output the crack recognition result. By combining convolutions with different void rates and global features, multi-scale information is effectively captured, enabling the network to better handle targets of different scales, and improving the model's feature extraction and semantic understanding capabilities.

[0119] In order to fully extract the global feature information and local feature information of the surface sub-image to enhance the accuracy of crack identification, in an optional implementation, the first processing unit includes:

[0120] A first processing module processes the surface sub-image through the BasicBlock module of the first encoding stage to obtain a first encoding feature map, and inputs the first encoding feature map to the CBAM module for feature enhancement to obtain a first enhanced feature map, wherein the first encoding stage is the first encoding stage in the encoder;

[0121] A second processing module processes the first enhanced feature map through the BasicBlock module of the second encoding stage to obtain a second encoded feature map, and inputs the second encoded feature map into the CBAM module for feature enhancement to obtain a second enhanced feature map, wherein the second encoding stage is the second encoding stage of the encoder;

[0122] A third processing module processes the second enhanced feature map through the BasicBlock module of the third encoding stage to obtain a third encoded feature map, and inputs the third encoded feature map into the CBAM module for feature enhancement to obtain a third enhanced feature map. The third encoding stage is the third encoding stage of the encoder.

[0123] The fourth processing module processes the third enhanced feature map through the BasicBlock module of the fourth encoding stage to obtain a fourth encoded feature map, and inputs the fourth encoded feature map into the CBAM module for feature enhancement to obtain a fourth enhanced feature map. The fourth encoding stage is the fourth encoding stage of the encoder.

[0124] In the above embodiment, Figure 3 This is the structural diagram of the surface crack identification model. Figure 3 The surface sub-image is divided into sub-patches by the BasicBlock module of the first residual block structure from top to bottom on the left side of the figure, and the first coded feature map F1 is quasi-enhanced by the CBAM module of the first coding stage to obtain the first enhanced feature map E1; the first enhanced feature map E1 is quasi-enhanced by the CBAM module of the second coding stage (i.e., Figure 3 The second encoding feature map F2 is processed by the BasicBlock module of the second residual block structure from top to bottom on the left side of the figure, and the second encoding feature map F2 is processed by the CBAM module of the second encoding stage to obtain the second enhanced feature map E2 with a size of The second enhanced feature map E2 passes through the third encoding stage (i.e. Figure 3 The BasicBlock module of the third residual block structure from top to bottom on the left side of the figure obtains the third encoded feature map F3, and the third encoded feature map F3 obtains the third enhanced feature map E3 through the CBAM module of the third encoding stage. The third enhanced feature map E3 passes through the fourth encoding stage (i.e. Figure 3 The fourth encoding feature map F4 is obtained by the BasicBlock module of the fourth residual block structure from top to bottom on the left side of the figure. The fourth encoding feature map F4 is obtained by the CBAM module of the fourth encoding stage to obtain the fourth enhanced feature map E4, with a size of At this point, the encoding is completed and the high-level feature information in the above-mentioned surface sub-image is extracted.

[0125] In order to fuse the multi-scale information in the image and effectively retain the feature detail information, in an optional implementation manner, the second processing unit includes:

[0126] A fifth processing module, inputting the fourth enhanced feature map into the ASPP module for multi-scale feature extraction, and sequentially performing convolution operation, acceleration operation and activation operation to obtain a first decoding feature map;

[0127] A sixth processing module, performing the convolution operation, the acceleration operation and the activation operation on the third enhanced feature map in sequence to obtain a second decoding feature map;

[0128] A first fusion module, fusing the first decoding feature map and the second decoding feature map to obtain a first fused feature map;

[0129] A seventh processing module, performing enhancement processing on the first fusion feature map by sequentially passing the deconvolution layer, the BasicBlock module and the CBAM module to obtain a third decoding feature map;

[0130] an eighth processing module, performing the convolution operation, the acceleration operation and the activation operation on the second enhanced feature map in sequence to obtain a fourth decoding feature map;

[0131] A second fusion module, fusing the fourth decoding feature map and the third decoding feature map to obtain a second fused feature map;

[0132] A ninth processing module, performing enhancement processing on the second fused feature map through the deconvolution layer, the BasicBlock module and the CBAM module in sequence to obtain a fifth decoding feature map;

[0133] a tenth processing module, performing the convolution operation, the acceleration operation and the activation operation on the first enhanced feature map in sequence to obtain a sixth decoding feature map;

[0134] A third fusion module, fusing the fifth decoding feature map and the sixth decoding feature map to obtain a third fused feature map;

[0135] The eleventh processing module performs enhancement processing on the third fusion feature map by sequentially passing the deconvolution layer, the BasicBlock module and the CBAM module to obtain the target decoding feature map.

[0136] In the above embodiment, the fourth enhanced feature map E4 is subjected to multi-scale feature extraction through the ASPP module to obtain the ASPP feature map, and the ASPP feature map A is subjected to a 1×1 convolution layer (i.e., the above convolution operation), and then the BN (i.e., the above acceleration operation) and the ReLU operation (i.e., the above activation operation) are performed to obtain the first decoding feature map G1. The third enhanced feature map E3 is subjected to a 1×1 convolution layer, and then the BN and ReLU operations are performed to obtain the second decoding feature map G2, which is then fused with the first decoding feature map G1 to obtain the first fused feature map. After the first fused feature map passes through 1 layer of deconvolution layer (or upsampling layer) and BasicBlock module, BN and ReLU operations are performed, and feature enhancement is performed through the CBAM module to obtain the third decoding feature map G3; after the second enhanced feature map E2 passes through 1 layer of 1×1 convolution layer, BN and ReLU operations are performed to obtain the fourth decoding feature map G4, and it is fused with the third decoding feature map G3 to obtain the second fused feature map; after the second fused feature map passes through 1 layer of 1 deconvolution layer (or upsampling layer) and BasicBlock module, BN and ReLU operations are performed, and feature enhancement is performed through the CBAM module to obtain the fifth decoding feature map G5; after the first enhanced feature map E1 passes through 1 layer of 1×1 convolution layer, BN and ReLU operations are performed to obtain the sixth decoding feature map G6, and it is fused with the fifth decoding feature map G5 to obtain the third fused feature map. After the third fusion feature map passes through 1 upsampling layer and BasicBlock module, BN and ReLU operations are performed, and feature enhancement is performed through CBAM module to obtain the above target decoding feature map G7. The decoding process is now completed, corresponding to Figure 3 The four residual block structures from bottom to top on the right. One residual block structure corresponds to one decoding stage. In each decoding stage, the upsampling layer and BasicBlock module are used to restore the spatial resolution. After upsampling, the CBAM module is inserted for feature enhancement. Through the skip connection, the feature maps of the decoding stage and the encoding stage are fused to retain the detail information.

[0137] The above-mentioned device for identifying surface cracks in a mining area includes a processor and a memory. The above-mentioned first acquisition unit, input and output unit, and splicing unit are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions. The above-mentioned modules are all located in the same processor; or, the above-mentioned modules are located in different processors in any combination.

[0138] The processor includes a kernel, which retrieves the corresponding program unit from the memory. One or more kernels can be set, and the problem of low accuracy of crack identification on the surface of the mining area in the prior art can be solved by adjusting kernel parameters.

[0139] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0140] An embodiment of the present invention provides a computer-readable storage medium, which includes a stored program, wherein when the program is run, the device where the computer-readable storage medium is located is controlled to execute the method for identifying surface cracks in a mining area.

[0141] Specifically, the identification methods of surface cracks in mining areas include:

[0142] Step S201, obtaining a surface crack image of a target mining area, and performing cropping processing on the surface crack image of the mining area to obtain a plurality of surface sub-images, wherein the target mining area surface is the mining area surface for which crack identification is currently being performed;

[0143] Step S202, sequentially inputting the surface sub-images into a surface crack recognition model, so as to use the surface crack recognition model to recognize cracks in the surface sub-images, and outputting sub-crack recognition results corresponding to each of the surface sub-images, wherein the surface crack recognition model is composed of an encoder, a decoder and a bottleneck layer, wherein the decoder at least includes a plurality of BasicBlock modules, a plurality of CBAM modules and an ASPP module, and the decoder at least includes the BasicBlock module and the CBAM module, wherein the BasicBlock module is used to enhance the ability to extract crack features, the CBAM module is used to capture the correlation between the crack features in different dimensions, and the ASPP module is used to extract information of the crack features at multiple scales;

[0144] Step S203, splicing the above-mentioned sub-crack identification results corresponding to each of the above-mentioned surface sub-images to obtain the crack identification results of the surface of the above-mentioned target mining area.

[0145] An embodiment of the present invention provides a processor, and the processor is used to run a program, wherein the method for identifying surface cracks in a mining area is executed when the program is run.

[0146] Specifically, the identification methods of surface cracks in mining areas include:

[0147] Step S201, obtaining a surface crack image of a target mining area, and performing cropping processing on the surface crack image of the mining area to obtain a plurality of surface sub-images, wherein the target mining area surface is the mining area surface for which crack identification is currently being performed;

[0148] Step S202, sequentially inputting the surface sub-images into a surface crack recognition model, so as to use the surface crack recognition model to recognize cracks in the surface sub-images, and outputting sub-crack recognition results corresponding to each of the surface sub-images, wherein the surface crack recognition model is composed of an encoder, a decoder and a bottleneck layer, wherein the decoder at least includes a plurality of BasicBlock modules, a plurality of CBAM modules and an ASPP module, and the decoder at least includes the BasicBlock module and the CBAM module, wherein the BasicBlock module is used to enhance the ability to extract crack features, the CBAM module is used to capture the correlation between the crack features in different dimensions, and the ASPP module is used to extract information of the crack features at multiple scales;

[0149] Step S203, splicing the above-mentioned sub-crack identification results corresponding to each of the above-mentioned surface sub-images to obtain the crack identification results of the surface of the above-mentioned target mining area.

[0150] An embodiment of the present invention provides a system for identifying surface cracks in a mining area. The system for identifying surface cracks in a mining area includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, at least the following steps are implemented:

[0151] Step S201, obtaining a surface crack image of a target mining area, and performing cropping processing on the surface crack image of the mining area to obtain a plurality of surface sub-images, wherein the target mining area surface is the mining area surface for which crack identification is currently being performed;

[0152] Step S202, sequentially inputting the surface sub-images into a surface crack recognition model, so as to use the surface crack recognition model to recognize cracks in the surface sub-images, and outputting sub-crack recognition results corresponding to each of the surface sub-images, wherein the surface crack recognition model is composed of an encoder, a decoder and a bottleneck layer, wherein the decoder at least includes a plurality of BasicBlock modules, a plurality of CBAM modules and an ASPP module, and the decoder at least includes the BasicBlock module and the CBAM module, wherein the BasicBlock module is used to enhance the ability to extract crack features, the CBAM module is used to capture the correlation between the crack features in different dimensions, and the ASPP module is used to extract information of the crack features at multiple scales;

[0153] Step S203, splicing the above-mentioned sub-crack identification results corresponding to each of the above-mentioned surface sub-images to obtain the crack identification results of the surface of the above-mentioned target mining area.

[0154] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program for initializing at least the following method steps:

[0155] Step S201, obtaining a surface crack image of a target mining area, and performing cropping processing on the surface crack image of the mining area to obtain a plurality of surface sub-images, wherein the target mining area surface is the mining area surface for which crack identification is currently being performed;

[0156] Step S202, sequentially inputting the surface sub-images into a surface crack recognition model, so as to use the surface crack recognition model to recognize cracks in the surface sub-images, and outputting sub-crack recognition results corresponding to each of the surface sub-images, wherein the surface crack recognition model is composed of an encoder, a decoder and a bottleneck layer, wherein the decoder at least includes a plurality of BasicBlock modules, a plurality of CBAM modules and an ASPP module, and the decoder at least includes the BasicBlock module and the CBAM module, wherein the BasicBlock module is used to enhance the ability to extract crack features, the CBAM module is used to capture the correlation between the crack features in different dimensions, and the ASPP module is used to extract information of the crack features at multiple scales;

[0157] Step S203, splicing the above-mentioned sub-crack identification results corresponding to each of the above-mentioned surface sub-images to obtain the crack identification results of the surface of the above-mentioned target mining area.

[0158] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0159] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0160] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0161] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0162] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0163] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0164] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0165] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0166] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0167] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0168] 1) The method for identifying surface cracks in a mining area of ​​the present application first obtains a surface crack image of the target mining area surface, and crops the surface crack image of the mining area to obtain multiple surface sub-images, wherein the target mining area surface is the mining area surface currently undergoing crack identification; then, the surface sub-images are sequentially input into a surface crack identification model to identify cracks in the surface sub-images using the surface crack identification model, and sub-crack identification results corresponding to each surface sub-image are output. The surface crack identification model is composed of an encoder, a decoder, and a bottleneck layer. The decoder includes at least a plurality of BasicBlock modules, a plurality of CBAM modules, and an ASPP module. The decoder includes at least a BasicBlock module and a CBAM module. The BasicBlock module is used to enhance the ability to extract crack features. The CBAM module is used to capture the correlation between crack features in different dimensions. The ASPP module is used to extract multi-scale crack feature information; finally, the sub-crack identification results corresponding to each surface sub-image are spliced ​​to obtain the crack identification results of the target mining area surface. This application constructs a surface crack recognition model by combining the BasicBlock module, the CBAM module and the ASPP module, which can effectively extract and fuse feature information at multiple scales, thereby improving the accuracy of surface crack recognition in mining areas. The surface crack image of the mining area is cropped to obtain multiple surface sub-images, and each surface sub-image is input into the surface crack recognition model to output the mining area surface crack recognition result of each surface sub-image, that is, the sub-crack recognition result, and the mining area surface crack recognition result of each surface sub-image is spliced ​​to obtain a complete crack recognition result on the surface of the target mining area. This application solves the problem of low accuracy of crack recognition on the surface of mining areas in the prior art.

[0169] 2) The mining area surface crack identification device of the present application, by combining the BasicBlock module, the CBAM module and the ASPP module to construct a surface crack identification model, can effectively extract and fuse feature information at multiple scales, thereby improving the accuracy of mining area surface crack identification. The mining area surface crack image is cropped to obtain multiple surface sub-images, and each surface sub-image is input into the surface crack identification model to output the mining area surface crack identification result of each surface sub-image, that is, the sub-crack identification result, and the mining area surface crack identification result of each surface sub-image is spliced ​​to obtain a complete crack identification result on the target mining area surface. The present application solves the problem of low accuracy of crack identification on the mining area surface in the prior art.

[0170] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for identifying surface cracks in a mining area, characterized in that: include: Acquire a surface crack image of a target mining area, and perform cropping processing on the surface crack image to obtain a plurality of surface sub-images, wherein the target mining area surface is the mining area surface on which crack identification is currently being performed; The surface sub-images are sequentially input into a surface crack recognition model, so as to use the surface crack recognition model to identify cracks in the surface sub-images, and output sub-crack recognition results corresponding to each surface sub-image. The surface crack recognition model is composed of an encoder, a decoder and a bottleneck layer. The decoder includes at least a plurality of BasicBlock modules, a plurality of CBAM modules and an ASPP module. The decoder includes at least the BasicBlock module and the CBAM module. The BasicBlock module is used to enhance the extraction capability of crack features. The CBAM module is used to capture the correlation between the crack features in different dimensions. The ASPP module is used to extract information of the crack features at multiple scales. The sub-crack identification results corresponding to each of the surface sub-images are spliced ​​to obtain the crack identification results of the surface of the target mining area.

2. The method according to claim 1, characterized in that The surface crack image of the mining area is cropped to obtain multiple surface sub-images, including: Processing the surface crack image of the mining area to generate an entire orthophoto map within the surveying and mapping range; The entire orthophoto image is cropped according to a set size to obtain a plurality of surface sub-images.

3. The method according to claim 1, characterized in that The surface sub-images are sequentially input into a surface crack recognition model, so as to use the surface crack recognition model to recognize cracks in the surface sub-images. Before outputting the sub-crack identification results corresponding to each of the surface sub-images, the method further includes: Acquire all historical surface images of the surface of the mining area; Performing pixel-level annotation on each of the historical surface images, and generating a label image corresponding to each of the historical surface images; Cropping the first surface image and the label image to obtain a second surface image and a label surface image respectively, wherein the first surface image is the annotated historical surface image, the second surface image is the image cropped from the first surface image, and the label surface image is the image cropped from the label image; The second surface images are screened according to the labeled surface images to remove images without surface cracks in the second surface images, thereby obtaining a surface crack image dataset.

4. The method according to claim 3, characterized in that The surface sub-images are sequentially input into a surface crack recognition model, so as to use the surface crack recognition model to recognize cracks in the surface sub-images. Before outputting the sub-crack identification results corresponding to each of the surface sub-images, the method further includes: Based on the U-net network, the BasicBlock module, the CBAM module and the ASPP module are combined to construct an initial surface crack identification model, and the initial surface crack identification model is a model to be trained; A training step, sequentially inputting input data of multiple sets of current training data into the initial surface crack recognition model for training, wherein the current training data is any set of training data in the surface crack image data set, and each set of the training data includes the input data and output data, wherein the input data is a sample surface crack sub-image, and the output data is a sample crack recognition result corresponding to the sample surface crack sub-image; A testing step, testing the trained initial surface crack identification model according to all the sample crack identification results to obtain the model accuracy; When the model accuracy does not meet the accuracy requirement, all parameters in the trained initial surface crack identification model are adjusted, and the training step and the testing step are repeated at least once until the model accuracy meets the accuracy requirement, thereby obtaining the surface crack identification model; When the model accuracy meets the accuracy requirement, the trained initial ground crack identification model is determined as the ground crack identification model.

5. The method according to claim 1, characterized in that The surface sub-images are sequentially input into a surface crack recognition model to use the surface crack recognition model to identify cracks in the surface sub-images, and sub-crack recognition results corresponding to each surface sub-image are output, including: The surface sub-image is subjected to feature extraction processing by the BasicBlock module and enhancement processing by the CBAM module in multiple encoding stages in the encoder to obtain an enhanced feature map of each encoding stage, the output of each encoding stage is used as the input of the next encoding stage, and each encoding stage includes a BasicBlock module and a CBAM module; Decoding all the enhanced feature maps in the decoder in a reverse acquisition order to obtain a target decoded feature map, wherein the reverse acquisition order is an order opposite to the acquisition order of each enhanced feature map in the encoding stage; The target decoding feature map is input into the bottleneck layer, and the sub-crack identification result is output.

6. The method according to claim 5, characterized in that The surface sub-image is subjected to feature extraction processing by the BasicBlock module and enhancement processing by the CBAM module in multiple encoding stages in the encoder to obtain an enhanced feature map of each encoding stage, including: Processing the surface sub-image through the BasicBlock module of the first encoding stage to obtain a first encoding feature map, and inputting the first encoding feature map into the CBAM module for feature enhancement to obtain a first enhanced feature map, wherein the first encoding stage is the first encoding stage in the encoder; Processing the first enhanced feature map through the BasicBlock module of the second encoding stage to obtain a second encoded feature map, and inputting the second encoded feature map into the CBAM module for feature enhancement to obtain a second enhanced feature map, wherein the second encoding stage is the second encoding stage of the encoder; Processing the second enhanced feature map through the BasicBlock module of the third encoding stage to obtain a third encoded feature map, and inputting the third encoded feature map into the CBAM module for feature enhancement to obtain a third enhanced feature map, wherein the third encoding stage is the third encoding stage of the encoder; The third enhanced feature map is processed by the BasicBlock module of the fourth encoding stage to obtain a fourth encoded feature map, and the fourth encoded feature map is input into the CBAM module for feature enhancement to obtain a fourth enhanced feature map. The fourth encoding stage is the fourth encoding stage of the encoder.

7. The method according to claim 6, characterized in that Decoding all the enhanced feature maps in the decoder in reverse acquisition order to obtain a target decoded feature map, including: Inputting the fourth enhanced feature map into the ASPP module for multi-scale feature extraction, and sequentially performing convolution operations, acceleration operations, and activation operations to obtain a first decoding feature map; The third enhanced feature map is sequentially subjected to the convolution operation, the acceleration operation, and the activation operation to obtain a second decoding feature map; Fusing the first decoding feature map and the second decoding feature map to obtain a first fused feature map; The first fusion feature map is sequentially subjected to enhancement processing by the deconvolution layer, the BasicBlock module and the CBAM module to obtain a third decoding feature map; The second enhanced feature map is sequentially subjected to the convolution operation, the acceleration operation, and the activation operation to obtain a fourth decoding feature map; Fusing the fourth decoding feature map and the third decoding feature map to obtain a second fused feature map; The second fused feature map is sequentially subjected to enhancement processing by the deconvolution layer, the BasicBlock module, and the CBAM module to obtain a fifth decoding feature map; Perform the convolution operation, the acceleration operation, and the activation operation on the first enhanced feature map in sequence to obtain a sixth decoding feature map; Fusing the fifth decoding feature map and the sixth decoding feature map to obtain a third fused feature map; The third fused feature map is sequentially enhanced by passing through the deconvolution layer, the BasicBlock module and the CBAM module to obtain the target decoding feature map.

8. A device for identifying surface cracks in a mining area, characterized in that: The device comprises: A first acquisition unit is used to acquire a surface crack image of a target mining area, and to crop the surface crack image of the mining area to obtain a plurality of surface sub-images, wherein the target mining area surface is the surface of the mining area where crack identification is currently being performed; An input-output unit, used for sequentially inputting the surface sub-images into a surface crack recognition model, so as to use the surface crack recognition model to recognize cracks in the surface sub-images, and outputting sub-crack recognition results corresponding to each surface sub-image, wherein the surface crack recognition model is composed of an encoder, a decoder and a bottleneck layer, wherein the decoder includes at least a plurality of BasicBlock modules, a plurality of CBAM modules and an ASPP module, and the decoder includes at least the BasicBlock module and the CBAM module, wherein the BasicBlock module is used to enhance the ability to extract crack features, the CBAM module is used to capture the correlation between the crack features in different dimensions, and the ASPP module is used to extract information of the crack features at multiple scales; The splicing unit is used to splice the sub-crack identification results corresponding to each of the surface sub-images to obtain the crack identification results of the surface of the target mining area.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 7.

10. A computer program product comprising computer instructions, characterized in that: When the computer instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images