Defect detection method for aramid fiber conveying belt and related equipment

Abstract

The invention provides an aramid fiber conveyor belt defect detection method and related equipment, and the method comprises the steps: obtaining a front module depth image and a rear module depth image of a target aramid fiber conveyor belt, and carrying out the detection of the defect of the target aramid fiber conveyor belt based on the absolute parallax between the matched front module depth image and the matched rear module depth image; according to the method, the target aramid conveying belt is subjected to absolute parallax detection, the layered image of the target aramid conveying belt is determined according to the absolute parallax, surface damage and skeleton damage of the conveying belt can be effectively distinguished through the layered image, and then the target aramid conveying belt is detected by adopting the nondestructive detection model, so that the detection accuracy and detection efficiency of the aramid conveying belt are improved. The problem of safety production accidents caused by the fact that abnormity of the aramid fiber conveying belt cannot be found in time is avoided.

Description

Technical Field

[0001] The present application relates to the field of technical transportation equipment detection, and in particular to an aramid conveyor belt defect detection method and related equipment. Background Art

[0002] Aramid conveyor belts play a vital role in mine transportation, and their performance is directly related to the operating efficiency and safety of the entire production system. However, aramid conveyor belts may have various defects during use, such as breakage, wear, joint elongation, etc. If these defects cannot be detected and repaired in time, they will cause further damage to the conveyor belt and even cause serious production accidents. Therefore, effective defect detection of aramid conveyor belts is crucial.

[0003] Based on the above situation, the conveyor belt defect detection method used in the prior art is inefficient and its accuracy is difficult to guarantee. Summary of the invention

[0004] In view of this, the purpose of this application is to propose an aramid conveyor belt defect detection method and related equipment to solve the above-mentioned technical problems.

[0005] Based on the above purpose, the first aspect of the present application provides a method for detecting defects in an aramid conveyor belt, comprising:

[0006] Acquire a front module depth image and a rear module depth image of a target aramid conveyor belt;

[0007] Performing feature extraction on the front module depth image by using a corner detection algorithm to obtain a plurality of first feature points, and performing feature extraction on the rear module depth image by using a corner detection algorithm to obtain a plurality of second feature points;

[0008] Determine a binary bit string corresponding to each first feature point in the front module depth image, and determine a binary bit string corresponding to each second feature point in the rear module depth image;

[0009] Based on the binary bit strings corresponding to each first feature point and the binary bit strings corresponding to each first feature point, determining the matching relationship between each first feature point and the second feature point, and matching the front module depth image and the rear module depth image according to the binary bit strings corresponding to each first feature point to obtain a matched front module depth image and a matched rear module depth image;

[0010] Determine the absolute parallax between the matched front module depth image and the matched rear module depth image, and determine the layered image of the target aramid conveyor belt according to the absolute parallax;

[0011] The layered image is input into a pre-trained non-destructive testing model, and the non-destructive testing model is used to determine whether the target aramid conveyor belt has damage defects.

[0012] Based on the same inventive concept, the second aspect of the present application provides an aramid conveyor belt defect detection device, comprising:

[0013] An acquisition module is configured to acquire a front module depth image and a rear module depth image of a target aramid conveyor belt;

[0014] A feature extraction module is configured to perform feature extraction on the front module depth image by using a corner detection algorithm to obtain a plurality of first feature points, and perform feature extraction on the rear module depth image by using a corner detection algorithm to obtain a plurality of second feature points;

[0015] a binary bit string determination module configured to determine a binary bit string corresponding to each first feature point in the front module depth image, and to determine a binary bit string corresponding to each second feature point in the rear module depth image;

[0016] A matching module, configured to determine a matching relationship between each first feature point and a second feature point based on the binary bit strings corresponding to each first feature point and the binary bit strings corresponding to each first feature point, and to match the front module depth image and the rear module depth image according to the binary bit strings corresponding to each first feature point to obtain a matched front module depth image and a matched rear module depth image;

[0017] A layering module is configured to determine an absolute parallax between a matched front module depth image and a matched rear module depth image, and determine a layered image of the target aramid conveyor belt according to the absolute parallax;

[0018] The damage defect determination module is configured to input the layered image into a pre-trained non-destructive testing model, and determine whether the target aramid conveyor belt has a damage defect through the non-destructive testing model.

[0019] Based on the same inventive concept, the third aspect of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable by the processor, wherein the processor implements the method described in the first aspect above when executing the computer program.

[0020] Based on the same inventive concept, the fourth aspect of the present application provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute the method described in the first aspect above.

[0021] From the above, it can be seen that the aramid conveyor belt defect detection method and related equipment provided by the present application obtain the front module depth image and the rear module depth image of the target aramid conveyor belt, based on determining the absolute parallax between the matched front module depth image and the matched rear module depth image, and determining the layered image of the target aramid conveyor belt according to the absolute parallax. The surface damage and skeleton damage of the conveyor belt can be effectively distinguished through the layered image, and then the target aramid conveyor belt is inspected using a non-destructive testing model, thereby improving the accuracy and efficiency of aramid conveyor belt detection and avoiding the problem of production safety accidents caused by the failure to detect aramid conveyor belt abnormalities in a timely manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the present application or related technologies, the drawings required for use in the embodiments or related technical descriptions are briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 This is a flow chart of a method for detecting defects in an aramid conveyor belt according to an embodiment of the present application;

[0024] Figure 2A This is a schematic diagram of the composition of a binocular linear x-ray aramid conveyor belt defect detection system according to an embodiment of the present application;

[0025] Figure 2B This is a schematic diagram of the binocular x-ray layered display principle of an embodiment of the present application;

[0026] Figure 2C A schematic diagram of a depth image of a front module according to an embodiment of the present application;

[0027] Figure 2D A schematic diagram of a depth image of a rear module according to an embodiment of the present application;

[0028] Figure 2E A schematic diagram of a partial convolution structure of an embodiment of the present application;

[0029] Figure 2F A schematic diagram of the neural network (Faster) structure of an embodiment of the present application;

[0030] Figure 2G A schematic diagram of the structure of a neural network (C2f_Faster) according to an embodiment of the present application;

[0031] Figure 2H This is a schematic diagram of the structure of a cross-view partial sample alignment (CPSA) module according to an embodiment of the present application;

[0032] Fig.2IIt is a schematic diagram of the structure of spatial pyramid pooling (SimSPPFCSPC) of an embodiment of the present application;

[0033] Figure 2J A schematic diagram of the objectives and background of an embodiment of the present application;

[0034] Figure 2K This is a schematic diagram of the nondestructive testing model structure of an embodiment of the present application;

[0035] Figure 2L This is a schematic diagram of the x-ray binocular stereoscopic vision principle of an embodiment of the present application;

[0036] Figure 3 This is a structural block diagram of an aramid conveyor belt defect detection device according to an embodiment of the present application;

[0037] Figure 4 A schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0038] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0039] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present application should be the usual meanings understood by people with ordinary skills in the field to which the present application belongs. The "first", "second" and similar words used in the embodiments of the present application do not represent any order, quantity or importance, but are only used to distinguish different components. "Including" or "comprising" and similar words mean that the elements or objects appearing in front of the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0040] It is understandable that before using the technical solutions of each embodiment of the present application, the type, scope of use, usage scenarios, etc. of the personal information involved will be informed to the user in an appropriate manner, and the user's authorization will be obtained.

[0041] For example, in response to receiving an active request from a user, a prompt message is sent to the user to clearly remind the user that the operation requested to be performed will require obtaining and using the user's personal information. Thus, the user can independently choose whether to provide personal information to the electronic device, application, server, storage medium or other software or hardware that performs the operation of the technical solution of the present application according to the prompt message.

[0042] As an optional but non-limiting implementation, in response to receiving the user's active request, the prompt information may be sent to the user in the form of a pop-up window, in which the prompt information may be presented in text form. In addition, the pop-up window may also carry a selection control for the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0043] It is understandable that the above notification and the process of obtaining user authorization are merely illustrative and do not constitute a limitation on the implementation method of this application. Other methods that meet relevant laws and regulations may also be applied to the implementation method of this application.

[0044] Mining conveyor belts play a vital role in mining transportation, and their performance is directly related to the operating efficiency and safety of the entire production system. The strength of aramid fiber is 5 times that of steel of the same quality, but its density is only 1 / 5 of that of steel. In order to achieve energy saving and consumption reduction, aramid rope conveyor belts gradually replace steel rope conveyor belts, and aramid becomes the main conveyor belt for high-power transportation in mining. With the widespread application of aramid conveyor belts for mining, their damage problem has become increasingly prominent. Common types of damage include breakage, wear, joint elongation, etc. If these damages cannot be detected and repaired in time, they will lead to further damage to the conveyor belt and even cause serious production accidents.

[0045] In the field of mining conveyor belt damage detection, the existing strong conveyor belt detection methods are mainly aimed at steel rope core conveyor belts. In the early days, they mainly relied on manual inspection, but this method is inefficient and the accuracy is difficult to guarantee. Magnetic field-based conveyor belt tear detection is a common method. By magnetizing the tested component, the defect leakage magnetic signal is obtained and analyzed and processed. Based on the characteristics of ferromagnetic materials and the principle of the geomagnetic field, the change of magnetic memory signal is used to detect defects. Machine vision detection technology has also been applied in this field. Through visual detection and image processing of longitudinal tearing of conveyor belts, conveyor belt faults can be effectively identified.

[0046] However, there are some problems with the above conveyor belt detection methods. The conveyor belt damage detection method based on magnetic field requires the steel wire rope core as the magnetized material, and the non-destructive detection of the conveyor belt is achieved by detecting the integrity of the steel wire rope. This method is ineffective for aramid conveyor belts. The conveyor belt damage detection method based on visible light or infrared light requires the use of a camera for image acquisition. Since the impact of harsh working conditions on the image obtained by the camera cannot be well solved, the damage detection method based on visible light and infrared light cannot be well applied.

[0047] An embodiment of the present application provides a method for detecting defects in an aramid conveyor belt, which obtains a front module depth image and a rear module depth image of a target aramid conveyor belt, determines the absolute parallax between the matched front module depth image and the matched rear module depth image, and determines a layered image of the target aramid conveyor belt based on the absolute parallax. The surface damage and skeleton damage of the conveyor belt can be effectively distinguished through the layered image, and then a nondestructive testing model is used to detect the target aramid conveyor belt, thereby improving the accuracy and efficiency of aramid conveyor belt detection and avoiding the problem of production safety accidents caused by failure to detect aramid conveyor belt abnormalities in a timely manner.

[0048] like Figure 1 As shown, the method of this embodiment includes:

[0049] Step 101, obtaining a front module depth image and a rear module depth image of a target aramid conveyor belt.

[0050] In this step, a binocular linear X-ray camera aramid conveyor belt defect detection system is designed to control the ray emission power to obtain clear X-ray images of the aramid conveyor belt.

[0051] The binocular linear x-ray camera aramid conveyor belt defect detection system is composed of a digital adjustable X-ray generator, an intrinsically safe high-speed X-ray detector, a flameproof and intrinsically safe controller, a mining intrinsically safe switch, Ethernet, a flameproof and intrinsically safe DC power supply, an intrinsically safe acquisition memory and a computer. The block diagram is shown below. Figure 2A shown.

[0052] The system uses two sets of photoelectric detection modules to collect x-ray images of aramid conveyor belts, and collects all-round, high-precision x-ray images of conveyor belts. The two sets of modules work together to capture the depth information of the conveyor belt from different angles, and use layered imaging technology to display the collected images in layers, which can effectively distinguish between surface damage and skeleton damage of the conveyor belt. Using deep learning models to detect damaged targets can improve detection accuracy and real-time performance, effectively avoid major production safety accidents caused by failure to detect aramid conveyor belt anomalies in a timely manner, and is of vital importance to ensuring production safety and improving production efficiency.

[0053] Since the density of the aramid core skeleton material is similar to that of the upper and lower rubber layers, the traditional monocular X-ray technology cannot effectively distinguish between skeleton damage and rubber layer damage. However, the use of binocular X-ray imaging technology can effectively distinguish between the conveyor belt skeleton and the upper and lower covering rubber layers through the principle of parallax, and can more accurately locate the damage position, effectively distinguish between skeleton damage and surface damage, and make a more accurate assessment of the health of the conveyor belt.

[0054] X-ray binocular imaging is based on the principle of binocular stereo vision. The X-ray source emits a wide X-ray beam to irradiate the object, and then two sets of linear array detectors separated by a distance receive the X-ray signals that penetrate the conveyor belt from different angles, thereby obtaining X-ray images of the conveyor belt at two different viewing angles, such as Figure 2B shown.

[0055] Based on the above description, the present application uses two sets of photoelectric detection modules arranged front and back to scan the running conveyor belt, and uses the signal acquisition module to collect images to generate two pairs of conveyor belt depth images in the running state (i.e., the front module depth image and the rear module depth image), wherein the front module depth image is as follows: Figure 2C As shown, the depth image of the rear module is as follows Figure 2D shown.

[0056] Step 102: extract features from the front module depth image using a corner detection algorithm to obtain a plurality of first feature points, and extract features from the rear module depth image using a corner detection algorithm to obtain a plurality of second feature points.

[0057] In this step, the Harris algorithm is used to extract the feature points of the two conveyor belt depth images. The algorithm detects corner points by calculating the gradient change of the local area of ​​the image, which can effectively extract the feature points in the image.

[0058] Step 103: determining a binary bit string corresponding to each first feature point in the front module depth image, and determining a binary bit string corresponding to each second feature point in the rear module depth image.

[0059] In this step, a binary bit string is a sequence of numbers consisting of 0s and 1s. In this context, each feature point is assigned a unique binary bit string, which is usually achieved through some kind of feature descriptor algorithm. A feature descriptor is a method that encodes the image area around the feature point into a numerical vector (in this case, binary form) so that feature points in different images can be compared and matched.

[0060] For each feature point in the front module depth image and the back module depth image, the corresponding feature descriptor algorithm is used to determine its corresponding binary bit string. These binary bit strings can then be used in the feature point matching process, that is, finding feature point pairs with similar binary bit strings between the two images to restore the structural information of the object for 3D reconstruction.

[0061] Step 104, based on the binary bit strings corresponding to each first feature point and the binary bit strings corresponding to each first feature point, determine the matching relationship between each first feature point and the second feature point, and match the front-module depth image and the rear-module depth image according to the binary bit strings corresponding to each first feature point to obtain a matched front-module depth image and a matched rear-module depth image.

[0062] In this step, based on the binary bit strings corresponding to each feature point, the matching relationship between the feature points in the front module depth image and the feature points in the rear module depth image is determined by comparing the similarity of these bit strings (such as using Hamming distance to measure the difference between binary strings).

[0063] The matching process usually involves finding the feature point pair with the smallest Hamming distance (i.e., the most similar). In order to improve the accuracy and robustness of the matching, some additional matching strategies may be adopted, such as ratio test or cross-check.

[0064] Once the matching relationship between feature points is determined, these matching point pairs can be used to "align" or "match" two depth images. Specifically, by calculating the geometric transformation between matching point pairs (such as affine transformation, perspective transformation or homography matrix), one image (such as the rear module depth image) can be transformed to the same coordinate system as another image (such as the front module depth image), thereby achieving image matching.

[0065] After matching, a matched front module depth image and a matched rear module depth image can be obtained. The two images should now be aligned to some extent, so that the corresponding feature points (or at least most of the feature points) are in the same position or a close position in the image.

[0066] Step 105 , determining an absolute parallax between the matched front module depth image and the matched rear module depth image, and determining a layered image of the target aramid conveyor belt according to the absolute parallax.

[0067] In this step, absolute disparity refers to the difference in depth or distance between two matching points (images from different viewpoints or times). In stereo vision, disparity is the key to calculating depth information. By comparing the positions of matching points (in the image coordinate system) and their corresponding depth values, the absolute disparity between these points can be calculated.

[0068] A larger parallax generally means a closer distance, while a smaller parallax may indicate a greater distance or a flatter surface.

[0069] Once the depth information and disparity map are available, the data can be further analyzed to identify delamination in the aramid conveyor belt. Delamination may appear as discontinuities in the depth image, unusual depth values, or significant changes in the disparity map.

[0070] By setting appropriate thresholds or applying image processing techniques (such as edge detection, region segmentation, etc.), delamination areas on the conveyor belt can be automatically identified and marked.

[0071] Ultimately, these analysis results can be used to generate a delamination image that clearly shows which parts of the aramid conveyor belt have delamination problems, enabling accurate inspection and evaluation of the aramid conveyor belt.

[0072] Step 106: input the layered image into a pre-trained non-destructive testing model, and determine whether the target aramid conveyor belt has damage defects through the non-destructive testing model.

[0073] In this step, the layered images are input into the nondestructive testing model. The model automatically analyzes each layer of the image, extracts the feature information (such as texture, shape, color, etc.), and compares it with the damage features learned during the training process.

[0074] Through complex calculation and reasoning processes, the model can determine whether the target aramid conveyor belt has damage defects, and may further indicate the type, location and severity of the damage.

[0075] This detection method is not only fast and efficient, but also can avoid secondary damage that may be caused by physical contact or destruction in traditional detection methods. It is of great significance to ensure the safe operation of conveyor belts.

[0076] In some embodiments, in step 106, the nondestructive testing model is trained by the following process:

[0077] Step A1, obtaining training image samples, balance factor α, focus parameter θ and adaptive threshold β.

[0078] Step A2: construct an initial model.

[0079] Step A3: input the training image sample into the initial model, and output the prediction result P through the initial model.

[0080] Step A4, constructing ATFL(p) based on the α, the θ, the β and the p by the following formula:

[0081] ATFL(p)=-α(1-p) θ log[β+(1-β)·p]

[0082] Where ATFL(p) represents the adaptive threshold focal loss function.

[0083] Step A5, training and adjusting the initial model according to the minimization result obtained in the process of minimizing the ATFL(p) to obtain a nondestructive testing model.

[0084] In the above scheme, the initial model is a target detection model (FCSA-YOLOv10 model). The initial model introduces the Faster structure for the improvement of the C2f structure, uses the Convolutional Block Attention Module (CBAM) to replace the Multi-Head Self-Attention (MHSA) in the Pyramid Split Attention (PSA) structure, and uses the Spatial Pyramid Pooling (SimSPPFCSPC) module to replace the original Fast Spatial Pyramid Pooling (SPPF) structure.

[0085] The Faster structure is introduced to improve the C2f structure. That is, the C2f_Faster module is an improved structure based on the c2f structure. It introduces modules such as partial convolution, multi-layer perceptron, and spatial mixing, which can improve the performance of the module in feature extraction, information processing, model expression, and adaptability, thereby helping to improve the accuracy and reliability of the entire model in the task of damage detection of mining aramid conveyor belts.

[0086] Among them, partial convolution (PConv) can extract and learn features of the convolution part in a targeted manner by dividing the feature map into the convolution part and the unprocessed part, thereby reducing the redundant information and computational complexity caused by the convolution operation on the entire feature map.

[0087] It receives parameters during initialization, including the dimension dim of the input feature, n_div used to determine the dimension of the convolution part, and the forward propagation method forward ('slicing' or 'split_cat'). The dimension dim_conv of the convolution part and the dimension dim_untouched of the untouched part are calculated based on n_div. Its structure is as follows Figure 2E shown.

[0088] The implementation of partial convolution involves the segmentation and processing of feature maps. The input feature map x is split into two parts x1 and x2 according to the dimension. The size of the split dimensions are dim_conv and dim_untouched respectively. The formula is:

[0089] x1, x2=split(x, [dim_conv, dim_untouched].

[0090] When the number of input and output channels is the same, PConv has a smaller amount of computation than standard convolution (Conv). At the same time, PConv has a smaller amount of memory access. The calculation formula for memory access is:

[0091]

[0092] Where h is the height of the input, w is the width of the input, and c p is the number of channels of spatial features. It can be seen that the memory access of PConv is 1 / 4 of that of conventional Conv.

[0093] By combining PConv with conventional Conv, and combining the batch normalization layer BN and the linear rectification function ReLu, the Faster structure is designed to improve the operation speed while ensuring feature diversity. Figure 2F shown.

[0094] The Faster structure is integrated with the original C2f structure to construct the C2f_Faster structure. The characteristics of the Faster structure are used to ensure that the C2f_Faster structure can improve the model detection speed while ensuring the detection accuracy remains unchanged. Figure 2G shown.

[0095] In addition, the convolutional block attention module (CBAM) is used to replace the multi-head self-attention (MHSA) in the PSA structure to establish a new CPSA structure.

[0096] CBAM combines channel attention and spatial attention to improve the performance of convolutional neural networks. The channel attention module calculates the importance of each channel to distinguish the features between different channels. The spatial attention module calculates the importance of each pixel in space to better capture the spatial structure in the image. This paper uses deep separable convolution to replace the original calculation operation to form a DW-CBAM structure. In the conveyor belt x-ray depth image target detection, the DW-CBAM structure can simultaneously focus on the shape, texture and other features of the target, which is more conducive to extracting the effective features of the target, and at the same time to improve the efficiency of the attention module operation.

[0097] The MHSA structure in the PSA module is replaced by the DW-CBAM attention mechanism to establish a CPSA module, whose structure is as follows: Figure 2H shown.

[0098] The SPPCSPC module is the spatial pyramid pooling (SPP) structure used in YOLOv7. This structure performs better than the SPP structure, but the number of parameters and calculations has increased a lot, affecting the model detection speed. Based on the SPPCSPC structure, combined with the SimSPPF structure proposed in YOLOv6, this application proposes the SimSPPFCSPC structure as follows Fig.2I As shown in the figure, the CBR module consists of ConV, BN and ReLu. The SimSPPFCSPC structure is based on SPPCSPC and combines the SimSPP structure to replace the 9×9 and 13×13 pooling cores with 5×5 pooling cores, and replace the CBS module with the CBR module, which greatly reduces the amount of calculation. The detection speed is improved while keeping the receptive field unchanged.

[0099] In the X-ray grayscale image of the aramid conveyor belt, the main component is the background, and the defect image occupies only a small part. For the training process, learning background features is easier than learning defect target features. The background can be regarded as a simple sample, and the defect target can be regarded as a difficult sample. However, even if the background is well learned, losses will still occur during training. In fact, the background samples that dominate the X-ray grayscale image dominate the direction of the gradient update, causing the defect target information to be submerged. To address this problem, this application introduces an adaptive threshold focus loss function (ATFL).

[0100] In X-ray grayscale image target detection, there is usually an imbalance between the target and the background, that is, the number of targets is relatively small and the background occupies most of the image area, such as Figure 2J As shown in Figure 3, this makes the model prone to being biased towards the background during training, resulting in a decrease in the target detection performance.

[0101] The adaptive threshold focal loss function can strengthen the loss related to the defect target and reduce the loss related to the background, prompting the model to allocate more attention to the defect target features, thereby alleviating the imbalance between the defect target and the background. By automatically adjusting the threshold, the model can pay more attention to those difficult-to-detect targets and improve the detection accuracy of the target. Specifically, for targets in X-ray grayscale images, they often appear as areas with higher grayscale or have specific shape features in the image. The adaptive threshold focal loss function can adjust the weight of the loss function based on these characteristics, so that the model pays more attention to learning the characteristics of these targets, thereby improving the detection ability of the targets.

[0102] In addition, the adaptive threshold focal loss function can also improve the robustness of the model and reduce the sensitivity of the model to noise and interference. In X-ray grayscale images, there may be interference factors such as noise and artifacts, which will affect the detection performance of the model. The adaptive threshold focal loss function can make the model pay more attention to the real target by adjusting the threshold, rather than being misled by noise and interference, thereby improving the stability and reliability of the model.

[0103] The adaptive threshold focus loss function is used to optimize the imbalance problem between the background and the target of the aramid conveyor belt image, and finally a nondestructive testing model is obtained, whose structure is as follows: Figure 2K shown.

[0104] In some embodiments, in step 103, determining the binary bit string corresponding to each first feature point in the front module depth image includes:

[0105] Step B1, performing the following operations on each first feature point in the front module depth image:

[0106] Step B11, obtaining the target center pixel point p corresponding to the first feature point 1 The gray value I(p 1 ).

[0107] Step B12, using the pixel point q in the preset center neighborhood N 1 The gray value I(q 1 ).

[0108] Step B13, based on the I(p 1 ) and I(q 1 ) is determined by the following formula: 1 ), I(q 1 )]:

[0109]

[0110] Among them, δ[I(p1), I(q 1 )] represents the first relative brightness relationship.

[0111] Step B14, according to the δ[I(p 1 ), I(q 1 )] Census(p 1 ):

[0112]

[0113] Among them, Census(p 1 ) represents the binary bit string corresponding to the first feature point.

[0114] In the above scheme, the non-parametric local (Census) transformation is used for the front view in the front module depth image of the conveyor belt, for each target center pixel p 1 , establish a matrix window and consider each pixel q within the central neighborhood N 1 By comparing p 1 and q 1 Grayscale value, generate a binary string to describe p 1 The arrangement of pixel intensities in the neighborhood. This process can capture the relative ordering information of local pixel intensities in the conveyor belt image and provide basic features for subsequent stereo matching. The Census algorithm reflects the relative brightness relationship between the central pixel and the pixels in its neighborhood (i.e., the first relative brightness relationship), and the mapping relationship defined by its grayscale value is:

[0115]

[0116] By comparing the binary bit string finally output (i.e., the binary bit string corresponding to the first feature point), it can be expressed as:

[0117]

[0118] In the formula, I(p 1 ) is the gray value of the center pixel in the matrix window; I(q 1 ) is the point q in the neighborhood N of the center point 1 The gray value of . It is a bitwise concatenation operation.

[0119] In some embodiments, in step 103, determining the binary bit string corresponding to each second feature point in the rear module depth image includes:

[0120] Step C1, performing the following operations on each second feature point in the rear module depth image:

[0121] Step C11, obtaining the target center pixel point corresponding to the second feature point, 2 The gray value I(p 2 ).

[0122] Step C12, using the pixel point q in the preset center neighborhood N 2 The gray value I(q 2 ).

[0123] Step C13, based on the I(p 2 ) and I(q 2 ) is determined by the following formula: 2 ), I(q 2 )]:

[0124]

[0125] Among them, δ[I(p 2 ), I(q 2 )] represents the second relative brightness relationship.

[0126] Step C14, according to the δ[I(p 2 ), I(q 2 )] Census(p 2 ):

[0127]

[0128] Among them, Census(p 2 ) represents the binary bit corresponding to the second feature point.

[0129] In the above scheme, the non-parametric local (Census) transformation is used for the front view in the rear module depth image of the conveyor belt, for each target center pixel p 2 , establish a matrix window and consider each pixel q within the central neighborhood N 2 By comparing p 2 and q 2 Grayscale value, generate a binary string to describe p 2 The arrangement of pixel intensities in the neighborhood. This process can capture the relative ordering information of local pixel intensities in the conveyor belt image and provide basic features for subsequent stereo matching. The Census algorithm reflects the relative brightness relationship between the central pixel and the pixels in its neighborhood (i.e., the second relative brightness relationship), and the mapping relationship defined by its grayscale value is:

[0130]

[0131] By comparing the binary bit string finally output (i.e. the binary bit string corresponding to the second feature point), it can be expressed as:

[0132]

[0133] In the formula, I(p 2 ) is the gray value of the center pixel in the matrix window; I(q 2 ) is the point q in the neighborhood N of the center point 2 The gray value of . It is a bitwise concatenation operation.

[0134] In some embodiments, in step 104, determining the matching relationship between each first feature point and the second feature point based on the binary bit string corresponding to each first feature point and the binary bit string corresponding to each first feature point includes:

[0135] Step D1, performing the following operations on the binary bit string x corresponding to each first feature point and the binary bit string y corresponding to each second feature point:

[0136] Step D11, determining Dice based on the x and the y by the following formula:

[0137]

[0138] Among them, Dice represents the similarity index.

[0139] Step D12: Select a second feature point with the highest similarity index from each Dice corresponding to the first feature point as a matching point for the first feature point.

[0140] In the above scheme, when performing binocular vision stereo matching, it is necessary to calculate the binary bit strings output after the Census transformation of the corresponding pixel points in the two views, and then proceed to the next step based on these two bit strings, calculate the Hamming distance, and you can get the similarity of the two points.

[0141] The binocular stereo matching (D-Census) algorithm uses the Dice coefficient to replace the Hamming distance used in the Census transformation. The Dice coefficient is an indicator used to measure the similarity between two sets.

[0142] In the stereo matching algorithm, the Census transformation results of the two views are regarded as two sets, and the similarity is evaluated by calculating the intersection and union between them. The Dice coefficient is defined as twice the intersection divided by the sum of the number of elements in the two sets.

[0143] Specifically, we first calculate the intersection of the two Census transform results (i.e., the number of bits marked as 1 in both) through a bitwise AND operation, and then calculate twice the intersection and the sum of the number of elements in the two sets.

[0144] The Dice coefficient is the ratio of twice the intersection to the sum of the number of elements in the two sets. This coefficient not only focuses on the number of matching sites, but also considers the overall similarity of the two sets, and can provide a more comprehensive similarity measurement method. This enables the Dice coefficient to provide more robust matching performance when dealing with images with complex textures, lighting changes, or a certain degree of occlusion. The Dice coefficient can be expressed as follows:

[0145]

[0146] For each target pixel p, find the pixel in the disparity search range d. min and d maxThe matching point with the highest similarity within the pixel is taken as the matching cost L(p, d) of the two pixels. The formula is as follows:

[0147] L(p, d) = Dice[census 1 (p), census 2 (pd)],d∈[d min , d max ].

[0148] In some embodiments, in step 105, determining the absolute disparity between the matched front module depth image and the matched rear module depth image includes:

[0149] Step E1, obtaining the centroid coordinates U corresponding to the matched front module depth image 1 , and the centroid coordinates U corresponding to the depth image of the matched module 2 .

[0150] Step E2, based on the U 1 and the U 2 D is obtained by the following formula:

[0151] U 1 -U 2 =D

[0152] Here, D represents the absolute disparity.

[0153] In the above solution, the centroid coordinates of the two depth images (ie, the matched front module depth image and the matched rear module depth image) are determined respectively.

[0154] The centroid coordinates here can be understood as the weighted average position of the depth values ​​of all pixels in the image, which reflects the center position of the depth information in the image.

[0155] For example, the centroid coordinates of the front module depth image are obtained by calculating the weighted average of the depth values ​​of all pixels (or valid pixels) in the front module depth image and their corresponding coordinates. The weight can be the depth value of the pixel itself, or a weight based on some other standard. For the centroid coordinates of the rear module depth image, similarly, the centroid coordinates of the rear module depth image are also obtained by calculating the weighted average of the depth values ​​of all pixels in the image and their coordinates.

[0156] The two centroid coordinates represent the “center” position of the two depth images in terms of depth information.

[0157] After obtaining the centroid coordinates of the two depth images, the centroid coordinates of the front module depth image and the centroid coordinates of the rear module depth image are used to calculate the coordinate difference. This coordinate difference may include the difference on the X axis (horizontal direction) and the difference on the Y axis (vertical direction), and sometimes may also include the difference on the Z axis (depth direction), although it is less common to directly calculate the Z axis difference in the depth image, because the depth image itself already reflects the depth information, thereby obtaining the absolute disparity, which provides important reference information for subsequent image processing or analysis.

[0158] Among them, the absolute disparity reflects the relative offset of the depth information of the two depth images.

[0159] In some embodiments, in step 105, determining the layered image of the target aramid conveyor belt according to the absolute parallax includes:

[0160] Step F1, obtaining the movement distance d of the target object point of the target aramid conveyor belt, the coordinate x1 of the target object point in the first dimension direction in the three-dimensional space in the front module depth image, the coordinate x2 of the target object point in the first dimension direction in the three-dimensional space in the rear module depth image, the coordinate y of the target object point in the second dimension direction in the three-dimensional space, and the coordinate Z of the target object point in the third dimension direction in the three-dimensional space.

[0161] Step F2, determining Zp based on said d, said z and D by the following formula:

[0162]

[0163] Wherein, D represents the absolute parallax, and Zp represents the coordinates of the target object point of the target aramid conveyor belt in the third dimension direction.

[0164] Step F3, determining Yp based on y by the following formula:

[0165] Yp=y

[0166] Wherein, Yp represents the coordinate of the target object point of the target aramid conveyor belt in the second dimension direction.

[0167] Step F4, determining Xp based on x1, d and D by the following formula: 1 :

[0168]

[0169] Among them, Xp 1 Represents a first coordinate of a target object point of the target aramid conveyor belt in the first dimensional direction.

[0170] Step E5, determining Xp based on x2, d and D by the following formula: 2 :

[0171]

[0172] Among them, Xp 2 The second coordinate of the target object point of the target aramid conveyor belt in the first dimension direction.

[0173] Step F6, based on the Xp 1 , the Xp 2 , the Yp and the Zp determine the layered image of the target aramid conveyor belt.

[0174] In the above scheme, the binocular x-ray imaging parallax principle is based on the binocular stereo vision principle, which is similar to the binocular stereo parallax principle in visible light. In binocular x-ray imaging, two x-rays at different angles are used to illuminate the object to obtain images at different viewing angles, such as Figure 2L As shown. Taking the distance between the centers of the two cameras as the baseline, when a point P in space is imaged at points P1 and P2 in the two cameras, the position of the ray source is defined as the initial point, point P is imaged at position P1 in the front photoelectric module, and the coordinate on the x-axis is x1, and point P is imaged at position P2 in the rear photoelectric module, and the coordinate on the x-axis is x2.

[0175] Although the x-ray imaging model is similar to that of visible light, the formation of parallax is different. A point P (x, y, z) in space, after passing through x-rays, the projection point in the front photoelectric module is P1, and after passing through x-rays, the projection point in the front photoelectric module is P2. The movement distance of point P in the conveyor belt motion plane is d. Taking the x-ray source as the coordinate origin, the central axis of the ray source intersects with the detector at point C, and the right is the positive direction of the coordinate system. Then the coordinates of P1 are (x1, y, z), and the coordinates of P2 are (x2, y, z). According to the parallax that has been calculated, the principle of triangle similarity is used to restore the structural information of the object and display the object in layers. The specific formula is:

[0176]

[0177] Yp=y;

[0178]

[0179]

[0180] It can be seen from the above formula that as the object depth Zp increases, the disparity D will gradually decrease. The different depths of the target object can be determined based on the size of the disparity.

[0181] In some embodiments, as shown in Table 1, the nondestructive testing model (FCSA-YOLOv10 model) of the present application is compared with the detection results of Li, YOLOv4, YOLOv3-LITE, YOLOv7, YOLOv8 and YOLOv10 on the mining aramid conveyor belt data set. The results show that the FCSA-YOLOv10 model of the present application performs well in all indicators, especially in the detection accuracy of small targets such as cracks and wear areas.

[0182] Table 1

[0183]

[0184] The experimental results show that the FCSA-YOLOv10 model of this application performs well on the mining aramid conveyor belt dataset, with detection accuracy (mAP), recall rate (Recall) and precision (Precision) reaching 98.5%, 95.1% and 96.7% respectively. In comparison, the performance of other models is inferior, which verifies the effectiveness of the improved method of this application.

[0185] In order to verify the impact of each module on the model performance, this application conducted an ablation experiment based on the FCSA-YOLOv10 model. The experiment tested the contribution of different modules to the detection performance, including CPSA, C2f_Faster, SimSPPFCSPC modules and ATFL function. All experiments were conducted under the same data set and parameter settings. The ablation experiment results are shown in Table 2:

[0186] Table 2

[0187]

[0188]

[0189] The results of the ablation experiment also verified the effect of each module on the improvement of model performance. The basic performance average precision (mAP) of the YOLOv10 model was 92.5%, the recall rate (Recall) was 89.8%, and the precision rate (Precision) was 91.1%. After the introduction of CPSA alone, the mAP increased to 94.2% and the Recall increased to 91.5%, indicating that CPSA significantly enhanced the model's ability to focus on key features. After the introduction of the SimSPPFCSPC module alone, the mAP reached 93.8% and the Recall reached 91.1%, indicating that the module enhanced the feature fusion capability. After the introduction of ATFL, the mAP and Recall increased to 93.2% and 90.9% respectively, showing its superiority in bounding box regression. After the introduction of C2f_Faster, the mAP and Recall increased to 93.2% and 90.9% respectively, showing its superiority in bounding box regression. After combining the two modules of CPSA+SimSPPFCSPC, the mAP reaches 96.1% and the Recall is improved to 93.4%, indicating that the synergy of these two modules significantly improves the model performance. After combining CPSA and ATFL, the mAP reaches 95.5% and the Recall is 92.7%. After combining SimSPPFCSPC and ATFL, the mAP is 95.1% and the Recall is 92.2%. After combining all the improved modules, the model performance reaches the best, with mAP of 98.5%, Recall of 95.1% and Precision of 96.7%.

[0190] From the above analysis, it can be seen that each improved module plays an important role in improving the performance of the model, especially the introduction of the CPSA module and the ATFL module, which significantly improves the detection accuracy and recall rate of the model. Finally, the FCSA-YOLOv10 model after combining all modules performs best in all indicators, verifying the effectiveness and superiority of this application.

[0191] The experimental results show that the FCSA-YOLOv10 model of this application is superior to traditional detection methods and other advanced models in all indicators. The specific conclusions are as follows:

[0192] (1) CPSA: Focusing on the shape, texture and other features of the target is more conducive to extracting effective features of the target and improving the computational efficiency of the attention module.

[0193] (2)SimSPPFCSPC module: Enhances the feature fusion capability, enabling the model to better handle damage features of different scales.

[0194] (3) ATFL function: It handles the problem of imbalance between image background and target and improves the accuracy of the model.

[0195] The X-ray nondestructive testing technology for aramid conveyor belts for mining, based on the FCSA-YOLOv10 model proposed in this study, significantly improves the accuracy and efficiency of testing by introducing a variety of advanced algorithms and mechanisms. Experimental results show that this technology has high reliability and stability in practical applications and can meet the high-precision and real-time testing requirements of mining sites.

[0196] Through the above scheme, the adaptive threshold focal loss function is used to optimize the imbalance problem between the background and the target of the aramid conveyor belt image, so that the detection accuracy of the obtained non-destructive testing model is significantly improved, and then the front module depth image and the rear module depth image of the target aramid conveyor belt are obtained, based on determining the absolute parallax between the matched front module depth image and the matched rear module depth image, and the layered image of the target aramid conveyor belt is determined according to the absolute parallax. The surface damage and skeleton damage of the conveyor belt can be effectively distinguished through the layered image, and then the non-destructive testing model is used to detect the target aramid conveyor belt, so as to improve the accuracy and efficiency of the aramid conveyor belt detection and avoid the problem of safety production accidents caused by the failure to discover the abnormality of the aramid conveyor belt in time.

[0197] It should be noted that the method of the embodiment of the present application can be performed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present application, and the multiple devices will interact with each other to complete the described method.

[0198] It should be noted that the above describes some embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the above embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0199] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present application also provides an aramid conveyor belt defect detection device.

[0200] refer to Figure 3 , the aramid conveyor belt defect detection device comprises:

[0201] An acquisition module 301 is configured to acquire a front module depth image and a rear module depth image of a target aramid conveyor belt;

[0202] The feature extraction module 302 is configured to perform feature extraction on the front module depth image by using a corner detection algorithm to obtain a plurality of first feature points, and perform feature extraction on the rear module depth image by using a corner detection algorithm to obtain a plurality of second feature points;

[0203] A binary bit string determination module 303 is configured to determine a binary bit string corresponding to each first feature point in the front module depth image, and determine a binary bit string corresponding to each second feature point in the rear module depth image;

[0204] The matching module 304 is configured to determine the matching relationship between each first feature point and the second feature point based on the binary bit string corresponding to each first feature point and the binary bit string corresponding to each first feature point, and match the front module depth image and the rear module depth image according to the binary bit string corresponding to each first feature point to obtain a matched front module depth image and a matched rear module depth image;

[0205] The layering module 305 is configured to determine the absolute parallax between the matched front module depth image and the matched rear module depth image, and determine the layered image of the target aramid conveyor belt according to the absolute parallax;

[0206] The damage defect determination module 306 is configured to input the layered image into a pre-trained non-destructive testing model, and determine whether the target aramid conveyor belt has a damage defect through the non-destructive testing model.

[0207] In some embodiments, the aramid conveyor belt defect detection device further includes a model training module, and the model training module is specifically configured as follows:

[0208] Obtain training image samples, balance factor α, focus parameter θ and adaptive threshold β;

[0209] Build an initial model;

[0210] Input the training image sample into the initial model, and output the prediction result P through the initial model;

[0211] ATFL(p) is constructed based on the α, the θ, the β and the p by the following formula:

[0212] ATFL(p)=-α(1-p) θ log[β+(1-β)·p]

[0213] Wherein, ATFL(p) represents the adaptive threshold focal loss function;

[0214] The initial model is trained and adjusted according to the minimization result obtained in the process of minimizing the ATFL(p) to obtain a nondestructive testing model.

[0215] In some embodiments, the binary bit string determination module 303 is specifically configured to:

[0216] The following operations are performed on each first feature point in the front module depth image:

[0217] Get the target center pixel point p corresponding to the first feature point 1 The gray value I(p 1 );

[0218] Using the pixel q in the preset center neighborhood N 1 The gray value I(q 1 );

[0219] Based on the I(p 1 ) and I(q 1 ) is determined by the following formula: 1 ), I(q 1 )]:

[0220]

[0221] Among them, δ[I(p 1 ), I(q 1 )] represents a first relative brightness relationship;

[0222] According to the δ[I(p 1 ), I(q 1 )] Census(p 1 ):

[0223]

[0224] Among them, Census(p 1 ) represents a binary bit string corresponding to the first feature point. In some embodiments, the binary bit string determination module 303 is specifically configured to perform the following operations on each second feature point in the rear module depth image:

[0225] Get the target center pixel point p corresponding to the second feature point 2 The gray value I(p 2 );

[0226] Using the pixel q in the preset center neighborhood N 2 The gray value I(q 2 );

[0227] Based on the I(p 2 ) and I(q 2 ) is determined by the following formula: 2 ), I(q 2 )]:

[0228]

[0229] Among them, δ[I(p 2 ), I(q 2 )] represents the second relative brightness relationship;

[0230] According to the δ[I(p 2 ), I(q 2 )] Census(p 2 ):

[0231]

[0232] Among them, Census(p 2 ) represents the binary bit string corresponding to the second feature point.

[0233] In some embodiments, the matching module 304 is specifically configured to:

[0234] The following operations are performed on the binary bit string x corresponding to each first feature point and the binary bit string y corresponding to each second feature point:

[0235] Dice is determined based on the x and the y by the following formula:

[0236]

[0237] Among them, Dice represents the similarity index;

[0238] A second feature point with the highest similarity index is selected from each Dice corresponding to the first feature point as a matching point of the first feature point.

[0239] In some embodiments, the stratification module 305 is specifically configured to:

[0240] Get the center of gravity coordinates U corresponding to the matched front module depth image 1 , and the centroid coordinates U corresponding to the depth image of the matched module 2 ;

[0241] Based on the U 1 and the U 2 D is obtained by the following formula:

[0242] U 1 -U2 =D

[0243] Here, D represents the absolute disparity.

[0244] In some embodiments, the stratification module 305 is specifically configured to:

[0245] Obtaining a movement distance d of a target object point of the target aramid conveyor belt, a coordinate x1 of the target object point in the first dimension direction in the three-dimensional space in the front module depth image, a coordinate x2 of the target object point in the first dimension direction in the three-dimensional space in the rear module depth image, a coordinate y of the target object point in the second dimension direction in the three-dimensional space, and a coordinate z of the target object point in the third dimension direction in the three-dimensional space;

[0246] Based on the d, the z and D, Zp is determined by the following formula:

[0247]

[0248] Wherein, D represents the absolute parallax, and Zp represents the coordinates of the target object point of the target aramid conveyor belt in the third dimension direction;

[0249] Based on y, Yp is determined by the following formula:

[0250] Yp=y

[0251] Wherein, Yp represents the coordinate of the target object point of the target aramid conveyor belt in the second dimension direction;

[0252] Xp is determined based on the x1, the d and the D by the following formula 1 :

[0253]

[0254] Among them, Xp 1 A first coordinate representing a target object point of the target aramid conveyor belt in the first dimension direction;

[0255] Xp is determined based on x2, d and D by the following formula 2 :

[0256] Among them, XΨ 2 A second coordinate representing a target object point of the target aramid conveyor belt in the first dimension direction;

[0257] Based on the Xp 1 , the Xp 2 , the Yp and the Zp determine the layered image of the target aramid conveyor belt.

[0258] For the convenience of description, the above device is described in terms of functions divided into various modules. Of course, when implementing the present application, the functions of each module can be implemented in the same or multiple software and / or hardware.

[0259] The device of the above embodiment is used to implement the corresponding aramid conveyor belt defect detection method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0260] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the aramid conveyor belt defect detection method described in any of the above embodiments is implemented.

[0261] Figure 4 A more specific schematic diagram of the hardware structure of an electronic device provided in this embodiment is shown, and the device may include: a processor 401, a memory 402, an input / output interface 403, a communication interface 404, and a bus 405. The processor 401, the memory 402, the input / output interface 403, and the communication interface 404 are connected to each other in communication within the device through the bus 405.

[0262] The processor 401 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0263] The memory 402 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 402 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program codes are stored in the memory 402 and called and executed by the processor 401.

[0264] The input / output interface 403 is used to connect the input / output module to realize information input and output. The input / output module can be configured in the device as a component (not shown in the figure), or it can be externally connected to the device to provide corresponding functions. The input device may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, etc.

[0265] The communication interface 404 is used to connect a communication module (not shown) to realize communication interaction between the device and other devices. The communication module can realize communication through a wired mode (such as USB, network cable, etc.) or a wireless mode (such as mobile network, WIFI, Bluetooth, etc.).

[0266] The bus 405 comprises a pathway for transmitting information between the various components of the device (eg, the processor 401 , the memory 402 , the input / output interface 403 , and the communication interface 404 ).

[0267] It should be noted that, although the above device only shows the processor 401, the memory 402, the input / output interface 403, the communication interface 404 and the bus 405, in the specific implementation process, the device may also include other components necessary for normal operation. In addition, it can be understood by those skilled in the art that the above device may also only include the components necessary for implementing the embodiments of the present specification, and does not necessarily include all the components shown in the figure.

[0268] The electronic device of the above embodiment is used to implement the corresponding aramid conveyor belt defect detection method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0269] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present application also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the aramid conveyor belt defect detection method described in any of the above embodiments.

[0270] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, 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, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic 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.

[0271] The computer instructions stored in the storage medium of the above embodiment are used to enable the computer to execute the aramid conveyor belt defect detection method described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0272] A person skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present application is limited to these examples. In line with the concept of the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present application as described above, which are not provided in detail for the sake of simplicity.

[0273] In addition, to simplify the description and discussion, and in order not to make the embodiments of the present application difficult to understand, the known power supply / ground connection with the integrated circuit (IC) chip and other components may or may not be shown in the provided drawings. In addition, the device can be shown in the form of a block diagram to avoid making the embodiments of the present application difficult to understand, and this also takes into account the fact that the details of the implementation of these block diagram devices are highly dependent on the platform to be implemented in the embodiments of the present application (that is, these details should be fully within the scope of understanding of those skilled in the art). In the case of elaborating specific details (e.g., circuits) to describe exemplary embodiments of the present application, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details or when these specific details are changed. Therefore, these descriptions should be considered to be illustrative rather than restrictive.

[0274] Although the present application has been described in conjunction with specific embodiments of the present application, many replacements, modifications and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may use the embodiments discussed.

[0275] The embodiments of the present application are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the present application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present application should be included in the protection scope of the present application.

Claims

1. A method for detecting defects in an aramid conveyor belt, characterized in that: include: Acquire a front module depth image and a rear module depth image of a target aramid conveyor belt; Performing feature extraction on the front module depth image by using a corner detection algorithm to obtain a plurality of first feature points, and performing feature extraction on the rear module depth image by using a corner detection algorithm to obtain a plurality of second feature points; Determine a binary bit string corresponding to each first feature point in the front module depth image, and determine a binary bit string corresponding to each second feature point in the rear module depth image; Based on the binary bit strings corresponding to each first feature point and the binary bit strings corresponding to each first feature point, determining the matching relationship between each first feature point and the second feature point, and matching the front module depth image and the rear module depth image according to the binary bit strings corresponding to each first feature point to obtain a matched front module depth image and a matched rear module depth image; Determine the absolute parallax between the matched front module depth image and the matched rear module depth image, and determine the layered image of the target aramid conveyor belt according to the absolute parallax; The layered image is input into a pre-trained non-destructive testing model, and the non-destructive testing model is used to determine whether the target aramid conveyor belt has damage defects.

2. The method according to claim 1, characterized in that The nondestructive testing model is trained through the following process: Obtain training image samples, balance factor α, focus parameter θ and adaptive threshold β; Build an initial model; Input the training image sample into the initial model, and output the prediction result p through the initial model; ATFL(p) is constructed based on the α, the θ, the β and the p by the following formula: ATFL(p)=-α(1-p) θ log[β+(1-β)·p] Wherein, ATFL(p) represents the adaptive threshold focal loss function; The initial model is trained and adjusted according to the minimization result obtained in the process of minimizing the ATFL(p) to obtain a nondestructive testing model.

3. The method according to claim 1, characterized in that The step of determining the binary bit string corresponding to each first feature point in the front module depth image includes: The following operations are performed on each first feature point in the front module depth image: Obtaining the grayscale value I(p1) of the target center pixel point p1 corresponding to the first feature point; Using the gray value I(q1) of the pixel point q1 within the preset central neighborhood N; Based on I(p1) and I(q1), δ[I(p1), I(q1)] is determined by the following formula: Wherein, δ[I(p1), I(q1)] represents the first relative brightness relationship; According to the δ[I(p1), I(q1)], Census(p1) is determined by the following formula: Here, Census(p1) represents the binary bit string corresponding to the first feature point.

4. The method according to claim 1, characterized in that: The determining of the binary bit string corresponding to each second feature point in the rear module depth image includes: The following operations are performed on each second feature point in the rear module depth image: Obtaining a grayscale value I(p2) of a target central pixel point p2 corresponding to the second feature point; Using the gray value I(q2) of the pixel point q2 within the preset central neighborhood N; Based on I(p2) and I(q2), δ[I(p2), I(q2)] is determined by the following formula: Wherein, δ[I(p2), I(q2)] represents the second relative brightness relationship; According to the δ[I(p2), I(q2)], Census(p2) is determined by the following formula: Here, Census(p2) represents the binary bit string corresponding to the second feature point.

5. The method according to claim 1, characterized in that The determining the matching relationship between each first feature point and the second feature point based on the binary bit string corresponding to each first feature point and the binary bit string corresponding to each first feature point includes: The following operations are performed on the binary bit string x corresponding to each first feature point and the binary bit string y corresponding to each second feature point: Dice is determined based on the x and the y by the following formula: Among them, Dice represents the similarity index; A second feature point with the highest similarity index is selected from each Dice corresponding to the first feature point as a matching point of the first feature point.

6. The method according to claim 1, characterized in that The determining of the absolute disparity between the matched front module depth image and the matched rear module depth image comprises: Obtaining the center of gravity coordinates U1 corresponding to the matched front module depth image and the center of gravity coordinates U2 corresponding to the matched rear module depth image; Based on U1 and U2, D is obtained by the following formula: U1-U2=D Here, D represents the absolute disparity.

7. The method according to claim 1, characterized in that The step of determining the layered image of the target aramid conveyor belt according to the absolute parallax comprises: Obtaining a movement distance d of a target object point of the target aramid conveyor belt, a coordinate x1 of the target object point in the first dimension direction in the three-dimensional space in the front module depth image, a coordinate x2 of the target object point in the first dimension direction in the three-dimensional space in the rear module depth image, a coordinate y of the target object point in the second dimension direction in the three-dimensional space, and a coordinate z of the target object point in the third dimension direction in the three-dimensional space; Based on the d, the Z and D, Zp is determined by the following formula: Wherein, D represents the absolute parallax, and Zp represents the coordinates of the target object point of the target aramid conveyor belt in the third dimension direction; Based on y, Yp is determined by the following formula: Yp=y Wherein, Yp represents the coordinate of the target object point of the target aramid conveyor belt in the second dimension direction; Xp1 is determined based on the x1, the d and the D by the following formula: Wherein, Xp1 represents the first coordinate of the target object point of the target aramid conveyor belt in the first dimension direction; Xp2 is determined based on the x2, the d and the D by the following formula: Wherein, Xp2 represents the second coordinate of the target object point of the target aramid conveyor belt in the first dimension direction; A layered image of the target aramid conveyor belt is determined based on the Xp1, the Xp2, the Yp, and the Zp.

8. A defect detection device for an aramid conveyor belt, characterized in that: include: An acquisition module is configured to acquire a front module depth image and a rear module depth image of a target aramid conveyor belt; A feature extraction module is configured to perform feature extraction on the front module depth image by using a corner detection algorithm to obtain a plurality of first feature points, and perform feature extraction on the rear module depth image by using a corner detection algorithm to obtain a plurality of second feature points; a binary bit string determination module configured to determine a binary bit string corresponding to each first feature point in the front module depth image, and to determine a binary bit string corresponding to each second feature point in the rear module depth image; A matching module, configured to determine a matching relationship between each first feature point and a second feature point based on the binary bit strings corresponding to each first feature point and the binary bit strings corresponding to each first feature point, and to match the front module depth image and the rear module depth image according to the binary bit strings corresponding to each first feature point to obtain a matched front module depth image and a matched rear module depth image; A layering module is configured to determine an absolute parallax between a matched front module depth image and a matched rear module depth image, and determine a layered image of the target aramid conveyor belt according to the absolute parallax; The damage defect determination module is configured to input the layered image into a pre-trained non-destructive testing model, and determine whether the target aramid conveyor belt has a damage defect through the non-destructive testing model.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to enable a computer to execute the method according to any one of claims 1 to 7.

Images