Belt damage detection method and device for underground coal mine belt conveyor

Abstract

The invention belongs to the technical field of coal mine detection, and discloses a belt damage detection method and device for an underground coal mine belt conveyor. The method comprises the following steps: acquiring an original belt damage image of the underground coal mine belt conveyor; denoising the original belt damage image to obtain a denoised belt damage image; performing feature extraction on the denoised belt damage image, and obtaining a thermodynamic feature map, an offset feature map and a size feature map according to the extracted features; determining a damage category and a relative position based on the thermodynamic feature map, determining a damage center position based on the offset feature map, and identifying and determining a damage size based on the size feature map; and obtaining a detection result based on the damage category and the relative position, the damage center position and the damage size. Through the above mode, the image is denoised and then the damage detection is carried out in combination with different feature maps, so that the detection precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of coal mine detection, and in particular to a belt damage detection method and device for a belt conveyor in an underground coal mine. Background Art

[0002] The working environment of underground belt conveyors for coal transportation is harsh, which means that the coal blocks being transported are likely to be mixed with sharp-edged objects such as gangue, anchor rods, and anchor cables. There is a drop of several meters between the belt conveyor's drop port and the transfer point between belts. When the raw coal falls from the drop port to the surface of the conveyor belt, there is a large momentum. If the loaded materials are mixed with sharp foreign objects such as coal gangue or I-beams, it is very likely to pierce the conveyor belt. The belt conveyor is a key transportation equipment in the coal mine production process. Its performance determines whether the coal can be transported safely and efficiently, and it is of vital importance to ensure its normal operation.

[0003] At present, belt damage detection for belt conveyors usually adopts the method of detecting belt damage based on manual visual inspection and detecting belt damage based on data collected by various sensors. Manual visual inspection is usually performed by experienced workers along the transportation direction of the belt conveyor to detect key positions such as the head, tail, conveyor belt, etc. This method has major drawbacks: it is difficult to detect belt puncture, belt tearing, excessive wear and other damage in time, which can easily cause a large number of conveyor belts to be scrapped prematurely. Secondly, the belt conveyor belt damage detection method based on various sensors is affected by the sensors and the complex underground environment, and the detection accuracy and stability are general. The current detection method for belt damage is difficult to detect belt damage in time and has poor detection accuracy.

[0004] The above contents are only used to assist in understanding the technical solution of the present invention and do not constitute an admission that the above contents are prior art. Summary of the invention

[0005] The main purpose of the present invention is to provide a belt damage detection method and device for a belt conveyor in an underground coal mine, aiming to solve the problem that the current detection method for belt damage is difficult to detect belt damage in time and has poor detection accuracy.

[0006] To achieve the above object, the present invention provides a belt damage detection method for a belt conveyor in an underground coal mine, and the belt damage detection method for a belt conveyor in an underground coal mine comprises the following steps:

[0007] Obtain the original belt damage image of the belt conveyor in the coal mine;

[0008] De-noising the original belt damage image to obtain a denoised belt damage image;

[0009] Extracting features from the denoised belt damage image, and obtaining a thermal feature map, an offset feature map, and a size feature map according to the extracted features;

[0010] Determine the damage type and relative position based on the thermal characteristic map, determine the damage center position based on the offset characteristic map, and identify and determine the damage size based on the size characteristic map;

[0011] The detection result is obtained based on the damage type and relative position, the damage center position and the damage size.

[0012] In some embodiments, denoising the original belt damage image includes:

[0013] The original belt damage image is converted from the spatial domain to the frequency domain by Fourier transform, so that the disordered noise points in the space are converted into the spectrum lines in the high frequency range in the frequency domain;

[0014] After converting the two-dimensional image from the spatial domain to the frequency domain, all the spectral lines in the demarcated high-frequency region are filtered out through a low-pass filter;

[0015] The image after filtering out the high-frequency spectral lines is converted back to the spatial domain to complete image denoising.

[0016] In some embodiments, converting the image after filtering out the high-frequency spectral lines back to the spatial domain to complete image denoising includes:

[0017] Obtaining structural information and characteristic information of the image from the image after filtering out the high-frequency region spectral lines;

[0018] The width and height of the image after filtering out the high-frequency region spectrum lines are compressed to one-half respectively, and the number of channels is expanded to four times, and the structural information, the feature information and the noise are separated to obtain a reference image;

[0019] Acquire low-resolution features in the original belt damage image and high-frequency features of the reference image according to the feature information, and fuse the low-resolution features and the high-frequency features of the reference image;

[0020] Image reconstruction is performed in the spatial domain based on the fused features and the structural information to complete image denoising.

[0021] In some embodiments, the extracting features from the denoised belt damage image includes:

[0022] Acquire a hidden layer feature map of the denoised belt damage image through a backbone network;

[0023] Determine the output tensor of the two-dimensional average pooling based on the number of channels, width and height of the preset input tensor of the backbone network;

[0024] Dividing the hidden layer feature map into a plurality of feature image blocks using the output tensor;

[0025] Spatial attention is used to compress each feature image block, and the compressed feature image blocks are fused for feature extraction.

[0026] In some embodiments, determining the damage category and relative position based on the thermal characteristic map includes:

[0027] Divide the thermal characteristic map into a plurality of grids according to a preset step size and in combination with the width, height and number of channels of the thermal characteristic map, assign a value in the range of [0,1] to each grid, wherein the value range represents the probability of the damaged object falling within the grid, and the preset step size is positively correlated with the density of the divided grids;

[0028] When any grid is detected to contain a damaged object, a corresponding key point is generated in the grid;

[0029] The damage type and relative position are determined based on the locations of several key points.

[0030] In some embodiments, determining the lesion center position based on the offset characteristic map includes:

[0031] Setting the preset step length as the target step length;

[0032] Determining a damage reference position based on the offset feature map according to the target step size;

[0033] The relative position is adjusted based on the deviation between the damage reference position and the relative position to obtain the damage center position.

[0034] In some embodiments, the determining the damage size based on the size feature map includes:

[0035] Determine the size type corresponding to the size characteristic diagram;

[0036] Outputting the damage size by using a convolutional network and combining the size type, wherein the damage size includes the width and height of the damage;

[0037] If 3D object detection is performed on the damage, the output damage size includes the width, height and depth dimensions of the damage.

[0038] In addition, to achieve the above-mentioned purpose, the present invention also proposes a belt damage detection device for a belt conveyor in an underground coal mine, the belt damage detection device for a belt conveyor in an underground coal mine comprising:

[0039] An acquisition module is used to obtain the original belt damage image of the belt conveyor in the coal mine;

[0040] A denoising module, used for denoising the original belt damage image to obtain a denoised belt damage image;

[0041] A construction module is used to extract features from the denoised belt damage image, and obtain a thermal feature map, an offset feature map, and a size feature map according to the extracted features;

[0042] A processing module, used to determine the damage category and relative position based on the thermal characteristic map, determine the damage center position based on the offset characteristic map, and identify and determine the damage size based on the size characteristic map;

[0043] The detection module is used to obtain the detection result based on the damage type and relative position, the damage center position and the damage size.

[0044] In some embodiments, the denoising module is used to convert the original belt damage image from the spatial domain to the frequency domain through Fourier transform, so that disordered noise points in the space are converted into spectrum lines in the high-frequency range in the frequency domain;

[0045] After converting the two-dimensional image from the spatial domain to the frequency domain, all the spectral lines in the demarcated high-frequency region are filtered out through a low-pass filter;

[0046] The image after filtering out the high-frequency spectral lines is converted back to the spatial domain to complete image denoising.

[0047] In some embodiments, the denoising module is used to obtain structural information and characteristic information of the image from the image after filtering out the high-frequency region spectral lines;

[0048] The width and height of the image after filtering out the high-frequency region spectrum lines are compressed to one-half respectively, and the number of channels is expanded to four times, and the structural information, the feature information and the noise are separated to obtain a reference image;

[0049] Acquire low-resolution features in the original belt damage image and high-frequency features of the reference image according to the feature information, and fuse the low-resolution features and the high-frequency features of the reference image;

[0050] Image reconstruction is performed in the spatial domain based on the fused features and the structural information to complete image denoising.

[0051] The present invention obtains the original belt damage image of the belt conveyor in the coal mine; denoises the original belt damage image to obtain the denoised belt damage image; extracts features from the denoised belt damage image, and obtains a thermal feature map, an offset feature map, and a size feature map based on the extracted features; determines the damage category and relative position based on the thermal feature map, determines the damage center position based on the offset feature map, and determines the damage size based on the size feature map; and obtains the detection result based on the damage category and relative position, the damage center position, and the damage size. In the above manner, damage detection is performed by denoising the image and then combining different feature maps to improve the detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a flow chart of a first embodiment of a belt damage detection method for a belt conveyor in an underground coal mine according to the present invention;

[0053] Figure 2 It is a structural block diagram of the first embodiment of the belt damage detection device for an underground belt conveyor in a coal mine according to the present invention.

[0054] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0055] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0056] The embodiment of the present invention provides a belt damage detection method for a belt conveyor in a coal mine. Figure 1 , Figure 1 The present invention is a flow chart of a first embodiment of a belt damage detection method for a belt conveyor in an underground coal mine.

[0057] In this embodiment, the belt damage detection method of the coal mine underground belt conveyor includes the following steps:

[0058] Step S10: Acquire the original belt damage image of the belt conveyor in the coal mine.

[0059] In this embodiment, the executor of this embodiment is a belt damage detection device for an underground belt conveyor in a coal mine, wherein the belt damage detection device for an underground belt conveyor in a coal mine has functions such as data processing, data communication and program running. The belt damage detection device for an underground belt conveyor in a coal mine can be a computer terminal device or other network device, and of course it can also be other devices with similar functions, and this embodiment does not impose any restrictions on this.

[0060] It should be noted that the current belt damage detection for belt conveyors usually adopts the method of detecting belt damage based on manual visual inspection and detecting belt damage based on data collected by various sensors. Manual visual inspection is usually performed by experienced workers along the transportation direction of the belt conveyor to detect key positions such as the head, tail, conveyor belt, etc. This method has major drawbacks: it is difficult to detect belt puncture, belt tearing, excessive wear and other damage in time, which can easily cause a large number of conveyor belts to be scrapped prematurely. Secondly, the belt conveyor belt damage detection method based on various sensors is affected by the sensors and the complex underground environment, and the detection accuracy and stability are general. The current detection method for belt damage is difficult to detect belt damage in time and has poor detection accuracy.

[0061] In order to solve the above technical problems, in this embodiment, the original belt damage image of the belt conveyor in the coal mine is obtained; the original belt damage image is denoised to obtain a denoised belt damage image; the denoised belt damage image is feature extracted, and a thermal feature map, an offset feature map and a size feature map are obtained according to the extracted features; the damage category and relative position are determined based on the thermal feature map, the damage center position is determined based on the offset feature map, and the damage size is determined based on the size feature map; the detection result is obtained based on the damage category and relative position, the damage center position and the damage size. In the above manner, image denoising is used and then different feature maps are combined for damage detection, thereby improving the detection accuracy.

[0062] In the specific implementation, in this embodiment, it is necessary to first collect the original belt damage image of the underground belt conveyor in the coal mine. Specifically, the original belt damage image of the underground belt conveyor in the coal mine can be obtained in real time through an image acquisition device such as a camera.

[0063] Step S20: denoising the original belt damage image to obtain a denoised belt damage image.

[0064] After obtaining the original belt damage image, in order to improve the subsequent detection accuracy, the original belt damage image needs to be denoised to obtain a denoised belt damage image.

[0065] The denoising process in this embodiment specifically includes first converting the original belt damage image from the spatial domain to the frequency domain through Fourier transform. It should be noted that the original belt damage image is essentially a two-dimensional image, and the image noise appears as disordered points in the spatial domain. After converting from the spatial domain to the frequency domain through Fourier transform, the disordered noise points in the space can be converted into spectral lines in the high-frequency range of the frequency domain. These designated high-frequency area spectral lines are all filtered out through a low-pass filter, and finally the image after filtering out the high-frequency area spectral lines is converted back to the spatial domain to complete image denoising.

[0066] In some embodiments, when restoring back to the spatial domain, in this embodiment, it is necessary to first obtain the structural information and feature information of the image from the image after filtering out the high-frequency region spectral lines. The structural information includes the size of the image, and the feature information includes low-resolution features. The pixel density corresponding to the low-resolution features is low, and there are fewer pixels per unit area. The division of high and low resolution can be set based on actual image processing requirements, and this is not limited in this embodiment. For the image after filtering out the high-frequency region spectral lines, this embodiment also adopts image compression for processing, for example, the width and height are compressed to one-half respectively, and the number of channels is expanded to four times. After compression, a reference image can be obtained, and then high-frequency features are further extracted from the reference image. The high-frequency features are the rapidly changing parts of the image, corresponding to the edges, details, textures and noise of the image. The noise has been removed from the high-frequency features obtained from the denoised reference image, and what remains are the edges, details and textures of the image. The low-resolution features are then fused with the high-frequency features to obtain the denoised image.

[0067] Step S30: extracting features from the denoised belt damage image, and obtaining a thermal feature map, an offset feature map, and a size feature map based on the extracted features.

[0068] After obtaining the denoised belt damage image, in this embodiment, the denoised belt damage image can be subjected to feature extraction. The specific process of feature extraction is to obtain the hidden layer feature map of the denoised belt damage image through the backbone network. The backbone network in this embodiment is a deep convolutional neural network, which includes 101 layers, specifically a network architecture of ResNet-101. The hidden layer feature map can be obtained from the denoised belt damage image through the backbone network. Then, the output tensor of the two-dimensional average pooling is determined based on the number of channels, width, and height of the preset input tensor of the backbone network. For example, assuming that the preset input tensor of the given backbone network is T b ∈R C×H×W , where W and H represent the number of channels, width, and height of the input tensor, respectively. The output tensor C of the two-dimensional average pooling is avg Should have dimensions W×H, C avg =AvgPool(T b ), where AvgPool(·) is a two-dimensional pooling operator. Then, the output tensor is used to divide the hidden layer feature map into several feature image blocks, for example, using C avg The hidden layer feature map is divided into N blocks, and the span of adjacent feature maps is d span =C / N. avg It can be regarded as the channel index of the channel-spatial fusion attention module. avgThe first N values ​​are used as indicators of spatial attention, and the distance between each two divided indicators must be greater than d span . Use spatial attention to compress each feature image block, and fuse the compressed feature image blocks to extract features. For the i-th feature map Fi in N parts, apply spatial attention to compress the feature map, such as Spa i = SpaAtt(F i ), where i∈[0,N]. SpaAtt(·) represents the compression operation of the spatial channel, which consists of average pooling and maximum pooling. Finally, the thermal feature map, offset feature map, and size feature map can be obtained based on the extracted features.

[0069] Step S40: determining the damage category and relative position based on the thermal characteristic map, determining the damage center position based on the offset characteristic map, and identifying and determining the damage size based on the size characteristic map.

[0070] After obtaining the thermal characteristic map, in this embodiment, the thermal characteristic map is divided into several grids according to the preset step size and combined with the width, height and number of channels of the thermal characteristic map. Assuming that the width, height and number of channels of the thermal characteristic map are W, H, and C respectively, the thermal characteristic map is divided into W / R×H / R×C grid maps, where R represents the preset step size, and each grid is assigned a value in the range of [0,1]. The range represents the probability that the damaged object falls within the grid. The preset step size is positively correlated with the density of the divided grids. The larger the step size, the greater the grid density. Then, when it is detected that any grid contains a damaged object, a corresponding key point is generated in the grid. The key point is also the key point corresponding to the damage, and the location of several key points is also the location of the damage. In addition, different types of damage appear at different locations, and the corresponding damage type can also be determined based on this.

[0071] Furthermore, the preset step size is set as the target step size, for example, R=4, and then the damage reference position is determined based on the offset feature map according to the target step size. There is a certain deviation between the target position output by the thermal feature map and the actual position of the target to be measured. Through the offset feature map, the deviation between the target position output by the above thermal feature map and the actual position is corrected, so that the damage center position can be obtained. The offset feature map uses a convolutional network to regress the deviation correction value.

[0072] Further, the size type corresponding to the size feature map can be obtained according to the size feature map, and the size type is also the different categories of size features represented by the feature map, such as spatial size, pixel size, physical size, and proportional size, etc., and in this embodiment, it is the width and height size values. It should be noted that the width and height size values ​​are for two-dimensional images. If it is necessary to further obtain the 3D size, it is necessary to output the depth size based on the width and height.

[0073] Step S50: Obtaining a detection result based on the damage type and relative position, the damage center position and the damage size.

[0074] Based on the relative position and center position obtained above, the actual specific position of the damage can be determined, and then the damage type and size can be combined to realize the damage detection of the belt. For example, if the damage type is crack damage, the crack position is the edge of the belt, and the specific length of the crack is X and the width is Y.

[0075] This embodiment obtains an original belt damage image of a belt conveyor in an underground coal mine; denoises the original belt damage image to obtain a denoised belt damage image; extracts features from the denoised belt damage image, and obtains a thermal feature map, an offset feature map, and a size feature map based on the extracted features; determines the damage category and relative position based on the thermal feature map, determines the damage center position based on the offset feature map, and identifies and determines the damage size based on the size feature map; obtains detection results based on the damage category and relative position, the damage center position, and the damage size, and performs damage detection using image denoising in combination with different feature maps, thereby improving detection accuracy.

[0076] Reference Figure 2 , Figure 2 It is a structural block diagram of the first embodiment of the belt damage detection device for an underground belt conveyor in a coal mine according to the present invention.

[0077] like Figure 2 As shown, the belt damage detection device for a belt conveyor in an underground coal mine proposed in an embodiment of the present invention comprises:

[0078] The acquisition module 10 is used to acquire the original belt damage image of the belt conveyor in the coal mine.

[0079] The denoising module 20 is used to denoise the original belt damage image to obtain a denoised belt damage image.

[0080] The construction module 30 is used to extract features from the denoised belt damage image, and obtain a thermal feature map, an offset feature map and a size feature map according to the extracted features.

[0081] The processing module 40 is used to determine the damage category and relative position based on the thermal characteristic map, determine the damage center position based on the offset characteristic map, and identify and determine the damage size based on the size characteristic map.

[0082] The detection module 50 is used to obtain the detection result based on the damage type and relative position, the damage center position and the damage size.

[0083] This embodiment obtains an original belt damage image of a belt conveyor in an underground coal mine; denoises the original belt damage image to obtain a denoised belt damage image; extracts features from the denoised belt damage image, and obtains a thermal feature map, an offset feature map, and a size feature map based on the extracted features; determines the damage category and relative position based on the thermal feature map, determines the damage center position based on the offset feature map, and identifies and determines the damage size based on the size feature map; obtains detection results based on the damage category and relative position, the damage center position, and the damage size, and performs damage detection using image denoising in combination with different feature maps, thereby improving detection accuracy.

[0084] In some embodiments, the denoising module 20 is used to convert the original belt damage image from the spatial domain to the frequency domain through Fourier transform, so that the disordered noise points in the space are converted into spectral lines in the high-frequency range in the frequency domain; after converting the two-dimensional image from the spatial domain to the frequency domain, all the spectral lines in the designated high-frequency area are filtered out through a low-pass filter; and the image after filtering out the spectral lines in the high-frequency area is converted back to the spatial domain to complete image denoising.

[0085] In some embodiments, the denoising module 20 is used to obtain structural information and feature information of the image from the image after filtering out the spectral lines in the high-frequency region; compress the width and height of the image after filtering out the spectral lines in the high-frequency region to one-half respectively, and expand the number of channels to four times, and separate the structural information from the feature information and the noise to obtain a reference image; obtain low-resolution features in the original belt damage image and high-frequency features of the reference image according to the feature information, and fuse the low-resolution features and the high-frequency features of the reference image; reconstruct the image in the spatial domain based on the fused features and the structural information to complete image denoising.

[0086] In some embodiments, the construction module 30 is used to obtain the hidden layer feature map of the denoised belt damage image through the backbone network; determine the output tensor of the two-dimensional average pooling based on the number of channels, width and height of the preset input tensor of the backbone network; use the output tensor to divide the hidden layer feature map into a number of feature image blocks; use spatial attention to compress each feature image block, and fuse the compressed feature image blocks to extract features.

[0087] In some embodiments, the processing module 40 is used to divide the thermal characteristic map into several grids according to a preset step size and in combination with the width, height and number of channels of the thermal characteristic map, and each grid is assigned a value in the range of [0,1], and the value range represents the probability that the damaged object falls within the grid. The preset step size is positively correlated with the density of the divided grids; when a damaged object is detected in any grid, a corresponding key point is generated in the grid; and the damage category and relative position are determined based on the positions of several key points.

[0088] In some embodiments, the processing module 40 is used to set the preset step length as the target step length; determine the damage reference position based on the offset feature map according to the target step length; and adjust the relative position based on the deviation between the damage reference position and the relative position to obtain the damage center position.

[0089] In some embodiments, the processing module 40 is used to determine the size type corresponding to the size feature map; use a convolutional network and combine the size type to output the damage size, and the damage size includes the width and height dimensions of the damage; if 3D target detection is performed on the damage, the output damage size includes the width, height and depth dimensions of the damage.

[0090] An embodiment of the present application also provides a belt damage detection device for a belt conveyor in an underground coal mine, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus, and the memory is used to store computer programs; the processor is used to implement the above-mentioned belt damage detection method for the belt conveyor in an underground coal mine when executing the program stored in the memory.

[0091] The communication bus mentioned in the above-mentioned belt damage detection device for underground belt conveyor of coal mine can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc.

[0092] The communication interface is used for communication between the belt damage detection device of the above-mentioned coal mine underground belt conveyor and other devices.

[0093] The memory may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0094] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, and discrete hardware components.

[0095] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive Solid State Disk (SSD)), etc.

[0096] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article 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, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0097] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0098] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

[0099] It should be understood that the above is only an example and does not constitute any limitation on the technical solution of the present invention. In specific applications, technicians in this field can make settings as needed, and the present invention does not limit this.

[0100] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of the present invention. In practical applications, technicians in this field can select part or all of them according to actual needs to achieve the purpose of the present embodiment, and no limitation is made here.

[0101] In addition, for technical details not fully described in this embodiment, reference can be made to the belt damage detection method for an underground belt conveyor in a coal mine provided in any embodiment of the present invention, and will not be repeated here.

[0102] In addition, it should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system 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, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0103] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0104] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as a read-only memory (ROM) / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0105] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

[0106] It is understandable that the system provided by the embodiment of the present invention corresponds to the method provided by the embodiment of the present invention, and the explanation, examples and beneficial effects of the relevant contents can refer to the corresponding parts in the above method.

Claims

1. A belt damage detection method for a belt conveyor in an underground coal mine, characterized in that: The belt damage detection method of the underground belt conveyor of the coal mine comprises: Obtain the original belt damage image of the belt conveyor in the coal mine; De-noising the original belt damage image to obtain a denoised belt damage image; Extracting features from the denoised belt damage image, and obtaining a thermal feature map, an offset feature map, and a size feature map according to the extracted features; Determine the damage type and relative position based on the thermal characteristic map, determine the damage center position based on the offset characteristic map, and identify and determine the damage size based on the size characteristic map; The detection result is obtained based on the damage type and relative position, the damage center position and the damage size.

2. The belt damage detection method for a belt conveyor in an underground coal mine as claimed in claim 1, characterized in that: Denoising the original belt damage image, including: The original belt damage image is converted from the spatial domain to the frequency domain by Fourier transform, so that the disordered noise points in the space are converted into the spectrum lines in the high frequency range in the frequency domain; After converting the two-dimensional image from the spatial domain to the frequency domain, all the spectral lines in the demarcated high-frequency region are filtered out through a low-pass filter; The image after filtering out the high-frequency spectral lines is converted back to the spatial domain to complete image denoising.

3. The belt damage detection method for a belt conveyor in an underground coal mine as claimed in claim 2, characterized in that: The image after filtering out the high-frequency spectral lines is converted back to the spatial domain to complete image denoising, including: Obtaining structural information and characteristic information of the image from the image after filtering out the high-frequency region spectral lines; The width and height of the image after filtering out the high-frequency region spectrum lines are compressed to one-half respectively, and the number of channels is expanded to four times, and the structural information, the feature information and the noise are separated to obtain a reference image; Acquire low-resolution features in the original belt damage image and high-frequency features of the reference image according to the feature information, and fuse the low-resolution features and the high-frequency features of the reference image; Image reconstruction is performed in the spatial domain based on the fused features and the structural information to complete image denoising.

4. The belt damage detection method for a belt conveyor in an underground coal mine as claimed in claim 1, characterized in that: The step of extracting features from the denoised belt damage image comprises: Acquire a hidden layer feature map of the denoised belt damage image through a backbone network; Determine the output tensor of the two-dimensional average pooling based on the number of channels, width and height of the preset input tensor of the backbone network; Dividing the hidden layer feature map into a plurality of feature image blocks using the output tensor; Spatial attention is used to compress each feature image block, and the compressed feature image blocks are fused for feature extraction.

5. The belt damage detection method for a belt conveyor in an underground coal mine as claimed in claim 1, characterized in that: The determining of the damage category and relative position based on the thermal characteristic map includes: Divide the thermal characteristic map into a plurality of grids according to a preset step size and in combination with the width, height and number of channels of the thermal characteristic map, assign a value in the range of [0,1] to each grid, wherein the value range represents the probability of the damaged object falling within the grid, and the preset step size is positively correlated with the density of the divided grids; When any grid is detected to contain a damaged object, a corresponding key point is generated in the grid; The damage type and relative position are determined based on the locations of several key points.

6. The method for detecting belt damage of a belt conveyor in an underground coal mine as claimed in claim 5, characterized in that: The determining of the damage center position based on the offset characteristic map comprises: Setting the preset step length as the target step length; Determining a damage reference position based on the offset feature map according to the target step size; The relative position is adjusted based on the deviation between the damage reference position and the relative position to obtain the damage center position.

7. The method for detecting belt damage of a belt conveyor in an underground coal mine as claimed in claim 6, characterized in that: The step of identifying and determining the damage size based on the size characteristic diagram comprises: Determine the size type corresponding to the size characteristic diagram; Outputting the damage size by using a convolutional network and combining the size type, wherein the damage size includes the width and height of the damage; If 3D object detection is performed on the damage, the output damage size includes the width, height and depth dimensions of the damage.

8. A belt damage detection device for a belt conveyor in an underground coal mine, characterized in that: The belt damage detection device for the underground belt conveyor of a coal mine comprises: An acquisition module is used to obtain the original belt damage image of the belt conveyor in the coal mine; A denoising module, used for denoising the original belt damage image to obtain a denoised belt damage image; A construction module is used to extract features from the denoised belt damage image, and obtain a thermal feature map, an offset feature map, and a size feature map according to the extracted features; A processing module, used to determine the damage category and relative position based on the thermal characteristic map, determine the damage center position based on the offset characteristic map, and identify and determine the damage size based on the size characteristic map; The detection module is used to obtain the detection result based on the damage type and relative position, the damage center position and the damage size.

9. The belt damage detection device for a belt conveyor in an underground coal mine as claimed in claim 8, characterized in that: The denoising module is used to convert the original belt damage image from the spatial domain to the frequency domain through Fourier transform, so as to convert the disordered noise points in the space into the spectrum lines in the high frequency range in the frequency domain; After converting the two-dimensional image from the spatial domain to the frequency domain, all the spectral lines in the demarcated high-frequency region are filtered out through a low-pass filter; The image after filtering out the high-frequency spectral lines is converted back to the spatial domain to complete image denoising.

10. The belt damage detection device for a belt conveyor in an underground coal mine as claimed in claim 9, characterized in that: The denoising module is used to obtain the structural information and characteristic information of the image from the image after filtering out the high-frequency region spectral lines; The width and height of the image after filtering out the high-frequency region spectrum lines are compressed to one-half respectively, and the number of channels is expanded to four times, and the structural information, the feature information and the noise are separated to obtain a reference image; Acquire low-resolution features in the original belt damage image and high-frequency features of the reference image according to the feature information, and fuse the low-resolution features and the high-frequency features of the reference image; Image reconstruction is performed in the spatial domain based on the fused features and the structural information to complete image denoising.

Images