Cable tunnel crack identification method, device, equipment, medium and program product

Abstract

The invention relates to a cable tunnel crack identification method, device and equipment, a medium and a program product. The method comprises the following steps: acquiring an infrared image of the inner wall of the cable tunnel; inputting the infrared image into an encoder of the crack segmentation model to obtain encoded infrared image features; the encoder is a ResNet-34 (ResNet-34) Inputting the encoded infrared image features into a decoder of a crack segmentation model, and obtaining a crack prediction map through the decoder; and determining a crack area of the inner wall of the cable tunnel based on the crack prediction map. By adopting the method, the infrared image can be sequentially input into the encoder and the decoder of the crack segmentation model for crack prediction to obtain the crack prediction map, and then the crack region of the inner wall of the cable tunnel is determined based on the crack prediction map, so that the efficiency and the accuracy of cable tunnel infrared image crack identification are improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, device, medium, and program product for identifying cracks in cable tunnels. Background Technology

[0002] With the acceleration of urbanization, cable tunnels, as an important urban infrastructure, directly affect the stability and safety of important systems such as power and communication.

[0003] Traditional cable tunnel inspection methods typically rely on manual inspections and image analysis, which suffer from low efficiency, high labor costs, and insufficient accuracy. In recent years, infrared imaging technology has gradually become an important tool for cable tunnel inspection, effectively detecting anomalies such as cracks. However, traditional crack identification methods still suffer from low accuracy and slow processing speed.

[0004] Therefore, how to utilize advanced neural networks and image processing technologies to improve the efficiency and accuracy of crack identification in infrared images of cable tunnels has become an urgent technical problem to be solved. Summary of the Invention

[0005] Therefore, it is necessary to provide a method, apparatus, equipment, medium, and program product for identifying cracks in cable tunnels that can improve the efficiency and accuracy of identifying crack areas in cable tunnels, in order to address the above-mentioned technical problems.

[0006] In a first aspect, this application provides a method for identifying cracks in cable tunnels, including:

[0007] Acquire infrared images of the inner wall of the cable tunnel;

[0008] The infrared image is input into the encoder of the crack segmentation model to obtain the encoded infrared image features; the encoder is ResNet-34.

[0009] The encoded infrared image features are input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder.

[0010] Based on the crack prediction map, the crack area in the inner wall of the cable tunnel was determined.

[0011] In one embodiment, the encoder that inputs the infrared image into the crack segmentation model to obtain encoded infrared image features includes:

[0012] The infrared image is convolved by the first convolutional layer in the encoder to obtain the first feature map;

[0013] The first feature map is convolved twice by the residual block in the encoder to obtain the encoded infrared image features.

[0014] In one embodiment, the method further includes:

[0015] Batch normalization is performed on the encoded infrared image features to obtain batch normalized infrared image features.

[0016] The output feature map is obtained based on the batch-normalized infrared image features and the first feature map;

[0017] The encoded infrared image features are input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder, including:

[0018] The output feature map is input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder.

[0019] In one embodiment, the above-mentioned inputting the output feature map into the decoder of the crack segmentation model, and obtaining the crack prediction map through the decoder, includes:

[0020] The output feature map is downsampled to obtain the downsampled feature map.

[0021] The downsampled feature map and the output feature map are concatenated to obtain the concatenated feature map.

[0022] The stitched feature map is input into the activation layer of the decoder for activation processing to obtain the crack prediction map.

[0023] In one embodiment, the above-mentioned determination of crack regions in the inner wall of the cable tunnel based on crack prediction maps includes:

[0024] The crack prediction map is binarized to obtain the binarized crack prediction map;

[0025] The crack prediction map after binarization is processed by connected component labeling to obtain the crack region on the inner wall of the cable tunnel.

[0026] In one embodiment, the method further includes:

[0027] The infrared image is preprocessed to obtain the preprocessed infrared image; the preprocessing includes noise reduction, contrast reduction, normalization, annotation and data augmentation.

[0028] The encoder of the crack segmentation model inputs the infrared image, including:

[0029] The preprocessed infrared image is input into the encoder of the crack segmentation model.

[0030] Secondly, this application also provides a crack identification device for cable tunnels, comprising:

[0031] The acquisition module is used to acquire infrared images of the inner wall of the cable tunnel;

[0032] The encoding module is used to input the infrared image into the encoder of the crack segmentation model to obtain the encoded infrared image features; the encoder is ResNet-34.

[0033] The decoding module is used to input the encoded infrared image features into the decoder of the crack segmentation model, and obtain the crack prediction map through the decoder;

[0034] The determination module is used to identify crack areas in the inner wall of cable tunnels based on crack prediction maps.

[0035] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0036] Acquire infrared images of the inner wall of the cable tunnel;

[0037] The infrared image is input into the encoder of the crack segmentation model to obtain the encoded infrared image features; the encoder is ResNet-34.

[0038] The encoded infrared image features are input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder.

[0039] Based on the crack prediction map, the crack area in the inner wall of the cable tunnel was determined.

[0040] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0041] Acquire infrared images of the inner wall of the cable tunnel;

[0042] The infrared image is input into the encoder of the crack segmentation model to obtain the encoded infrared image features; the encoder is ResNet-34.

[0043] The encoded infrared image features are input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder.

[0044] Based on the crack prediction map, the crack area in the inner wall of the cable tunnel was determined.

[0045] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0046] Acquire infrared images of the inner wall of the cable tunnel;

[0047] The infrared image is input into the encoder of the crack segmentation model to obtain the encoded infrared image features; the encoder is ResNet-34.

[0048] The encoded infrared image features are input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder.

[0049] Based on the crack prediction map, the crack area in the inner wall of the cable tunnel was determined.

[0050] The above-mentioned method, apparatus, equipment, medium, and program products for identifying cracks in cable tunnels. The method includes: acquiring infrared images of the inner wall of the cable tunnel; inputting the infrared images into an encoder of a crack segmentation model to obtain encoded infrared image features; the encoder is ResNet-34; inputting the encoded infrared image features into a decoder of the crack segmentation model to obtain a crack prediction map; and determining the crack region in the inner wall of the cable tunnel based on the crack prediction map. This method improves the efficiency and accuracy of crack identification in cable tunnels by sequentially inputting infrared images into the encoder and decoder of the crack segmentation model for crack prediction, obtaining a crack prediction map, and then determining the crack region in the inner wall of the cable tunnel based on the crack prediction map. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a diagram illustrating the application environment of a crack identification method for cable tunnels in one embodiment.

[0053] Figure 2 This is a flowchart illustrating a method for identifying cracks in a cable tunnel in one embodiment.

[0054] Figure 3 This is a flowchart illustrating a method for identifying cracks in a cable tunnel in another embodiment;

[0055] Figure 4 This is a flowchart illustrating a method for identifying cracks in a cable tunnel in another embodiment;

[0056] Figure 5 This is a flowchart illustrating a method for identifying cracks in a cable tunnel in another embodiment;

[0057] Figure 6 This is a flowchart illustrating a method for identifying cracks in a cable tunnel in another embodiment;

[0058] Figure 7 This is a flowchart illustrating a method for identifying cracks in a cable tunnel in another embodiment;

[0059] Figure 8 This is a structural block diagram of a crack detection device for a cable tunnel in one embodiment. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0061] With the acceleration of urbanization, cable tunnels, as an important urban infrastructure, directly affect the stability and safety of important systems such as power and communication.

[0062] Traditional cable tunnel inspection methods typically rely on manual inspections and image analysis, which suffer from low efficiency, high labor costs, and insufficient accuracy. In recent years, infrared imaging technology has gradually become an important tool for cable tunnel inspection, effectively detecting anomalies such as cracks. However, traditional crack identification methods still suffer from low accuracy and slow processing speed.

[0063] Therefore, how to utilize advanced neural networks and image processing technologies to improve the efficiency and accuracy of crack identification in infrared images of cable tunnels has become an urgent technical problem to be solved. This application aims to address this problem.

[0064] Having described the background technology of the cable tunnel crack identification method provided in the embodiments of this application, the following is a brief description of the implementation environment involved in the cable tunnel crack identification method provided in the embodiments of this application. The cable tunnel crack identification method provided in the embodiments of this application can be applied to, for example... Figure 1The computer device shown includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a method for identifying cracks in cable tunnels. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0065] Those skilled in the art will understand that Figure 1 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. A specific terminal may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0066] Having described the application scenarios of the cable tunnel crack identification method provided in the embodiments of this application above, the following focuses on the cable tunnel crack identification method described in this application.

[0067] In one embodiment, such as Figure 2 As shown, a method for identifying cracks in cable tunnels is provided, which can be applied to... Figure 1 Taking a computer device as an example, the explanation includes the following steps:

[0068] S201. Acquire infrared images of the inner wall of the cable tunnel.

[0069] In this embodiment, an infrared thermal imager can be used to acquire infrared images of the inner wall of the cable tunnel. The image resolution is set to W×H, the thermal sensitivity of the image is set to ≤0.05℃, and the acquired infrared image is recorded as... ,in, These are pixel coordinates.

[0070] S202. Input the infrared image into the encoder of the crack segmentation model to obtain the encoded infrared image features; the encoder is ResNet-34.

[0071] In this embodiment, after obtaining the infrared image, the infrared image can be input into the encoder of the crack segmentation model to obtain the encoded infrared image features; the encoder is ResNet-34.

[0072] Optionally, when training the crack segmentation model, the cross-entropy loss function of equation (1) can be used to measure the difference between the network output and the true label. For binary crack detection tasks, the cross-entropy loss can measure the classification error for each pixel:

[0073] (1);

[0074] in, This represents the total number of samples. It is the true label of the i-th pixel. The value can be 0 or 1 (0 represents the background, and 1 represents the crack area). It is the model's predicted probability for the i-th pixel, i.e. .

[0075] In addition, using a labeled infrared image crack dataset, the neural network is trained using backpropagation and optimization algorithms through the above steps to obtain the final crack segmentation model.

[0076] Furthermore, the performance of the detection results was evaluated using precision, recall, overall accuracy, and similarity coefficient. True positives (TP) represent the number of correctly segmented crack pixels, and false positives (FP) represent incorrect predictions where pixels belonging to non-crack categories were identified as crack pixels; true negatives (TN) represent the number of correctly segmented non-crack pixels, and false negatives (FN) represent incorrect predictions where pixels belonging to crack categories were identified as non-crack pixels. Precision represents the percentage of correct samples among the positive samples in the crack detection results, defined by equation (2):

[0077] (2);

[0078] The recall rate mainly reflects the number of pixels correctly predicted in the positive samples out of all samples, and is defined by equation (3):

[0079] (3);

[0080] Among them, the overall accuracy represents the proportion of correct predictions in all data, which is a comprehensive evaluation index. The calculated value is generally too high, close to 99%, because the negative samples (i.e., background) are much larger than the positive samples (cracks), and is defined as Equation (4):

[0081] (4);

[0082] The similarity coefficient is a metric for segmentation performance, calculated as twice the intersection of the predicted and actual values ​​plus the sum of their intersections and unions. If DSC equals 1, it indicates that the cracks have been extracted 100% completely. Conversely, if DSC equals 0, the road segmentation result is completely incorrect, defined by equation (5):

[0083] (5);

[0084] S203. Input the encoded infrared image features into the decoder of the crack segmentation model, and obtain the crack prediction map through the decoder.

[0085] In this embodiment, after obtaining the encoded infrared image features, the encoded infrared image features are input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder.

[0086] S204. Based on the crack prediction map, determine the crack area in the inner wall of the cable tunnel.

[0087] In this embodiment, after obtaining the crack prediction map, the crack area of ​​the inner wall of the cable tunnel can be determined based on the crack prediction map.

[0088] In this embodiment, by sequentially inputting infrared images into the encoder and decoder of the crack segmentation model for crack prediction, a crack prediction map is obtained. Then, based on the crack prediction map, the crack area of ​​the inner wall of the cable tunnel is determined, which improves the efficiency and accuracy of crack identification in infrared images of cable tunnels.

[0089] In one embodiment, in Figure 2 Based on the illustrated embodiment, the process of encoding infrared images can be described, such as... Figure 3 As shown, the above-mentioned S202 "inputting the infrared image into the encoder of the crack segmentation model to obtain the encoded infrared image features" includes:

[0090] S301. The infrared image is convolved through the first convolutional layer in the encoder to obtain the first feature map.

[0091] In this embodiment, the crack segmentation model can be U-Net, and the encoder part of U-Net can be ResNet-34.

[0092] In this embodiment, after the infrared image is acquired, the infrared image is input to the first convolutional layer of the encoder (ResNet-34). The first convolutional layer performs a convolution operation on the infrared image to obtain a first feature map.

[0093] S302. Perform two convolution operations on the first feature map using the residual block in the encoder to obtain the encoded infrared image features.

[0094] In this embodiment, after obtaining the first feature map, the first feature map is input into the residual block in the encoder for two convolution operations to obtain the encoded infrared image features.

[0095] In this embodiment, the infrared image is input into the encoder of the crack segmentation model for encoding processing to obtain the encoded infrared image features, which provides a basis for subsequently determining the crack area in the inner wall of the cable tunnel in the crack prediction map.

[0096] In one embodiment, in Figure 3 Based on the illustrated embodiments, as Figure 4 As shown, the above method also includes:

[0097] S303. Perform batch normalization on the encoded infrared image features to obtain batch normalized infrared image features.

[0098] In this embodiment, after obtaining the encoded infrared image, the encoded infrared image features are batch normalized to obtain batch normalized infrared image features.

[0099] Optionally, batch normalization is used to accelerate the training process and improve the model's generalization ability, and its formula is Equation (6):

[0100] (6);

[0101] Where x is the input feature matrix, E[x] and Var[x] are the mean and variance of the input feature matrix, respectively, ϵ is a small number to prevent the denominator from being zero, and γ and β are learnable scaling and offset parameters.

[0102] S304. Based on the batch-normalized infrared image features and the first feature map, the output feature map is obtained.

[0103] In this embodiment, after obtaining the batch-normalized infrared image features, the first feature map and the batch-normalized infrared image features are summed to obtain the output feature map. Optionally, the following formula (7) provides a method for obtaining the output feature map:

[0104] (7);

[0105] in, and These represent two convolution operations, BN represents batch normalization, ReLU is the activation function, X is the first feature map, and Y is the output feature map.

[0106] The aforementioned S302, "inputting the encoded infrared image features into the decoder of the crack segmentation model, and obtaining the crack prediction map through the decoder," includes:

[0107] S302. Input the output feature map into the decoder of the crack segmentation model, and obtain the crack prediction map through the decoder.

[0108] In this embodiment, the output feature map obtained above can be input into the decoder of the crack segmentation model to obtain the crack prediction map.

[0109] Optionally, the following provides a method for obtaining the crack prediction map using a decoder; see [link to relevant documentation]. Figure 5 That is, S302 above, "inputting the output feature map into the decoder of the crack segmentation model, and obtaining the crack prediction map through the decoder", includes:

[0110] S3021. The output feature map is downsampled to obtain the downsampled feature map.

[0111] In this embodiment, after obtaining the output feature map, the output feature map can be downsampled to obtain the downsampled feature map.

[0112] S3022. The downsampled feature map and the output feature map are concatenated to obtain a concatenated feature map.

[0113] In this embodiment, after obtaining the downsampled feature map, the downsampled feature map and the output feature map can be concatenated to obtain a concatenated feature map.

[0114] Optionally, the process of obtaining the stitched feature map is described in the following formula (8):

[0115] (8);

[0116] Where F is the concatenated feature map, which serves as the input for subsequent convolution operations to further process and fuse feature information from the encoder and decoder; U is the feature map after downsampling; and E is the output feature map.

[0117] S3023. Input the spliced ​​feature map into the activation layer of the decoder for activation processing to obtain the crack prediction map.

[0118] In this embodiment, after obtaining the stitched feature map, the stitched feature map can be input into the activation layer of the decoder for activation processing to obtain the crack prediction map.

[0119] Optionally, after obtaining the stitched feature map, the Sigmoid activation function (9) can be used in the last layer of the decoder to output the segmentation mask of the crack prediction map. The value of each pixel represents the probability that it belongs to the crack region.

[0120] (9);

[0121] in The Sigmoid activation function is represented by the formula: . It is the input feature map of the last layer of the decoder, which is the feature map obtained after the depthwise separable convolution operation.

[0122] In this embodiment, the infrared image is input into the encoder of the crack segmentation model for encoding processing to obtain the encoded infrared image features, which provides a basis for subsequently determining the crack area in the inner wall of the cable tunnel in the crack prediction map.

[0123] In one embodiment, in Figure 2 Based on the illustrated embodiment, the process of determining the crack region in the inner wall of the cable tunnel can be described, such as... Figure 6 As shown, the above-mentioned S204 "Determining the crack area of ​​the inner wall of the cable tunnel based on the crack prediction map" includes:

[0124] S401. Perform binarization processing on the crack prediction map to obtain the binarized crack prediction map.

[0125] In this embodiment, after obtaining the crack prediction map, the crack prediction map is binarized to obtain a binarized crack prediction map. Optionally, after obtaining the crack prediction map, dynamic threshold binarization is performed on the crack prediction map to obtain a binarized crack prediction map, so as to distinguish the crack region in the image from the background region, as shown in the above formula (10):

[0126] (10);

[0127] in, , , for The mean and standard deviation, β=0.5.

[0128] S402. Perform connected component labeling on the binarized crack prediction map to obtain the crack region on the inner wall of the cable tunnel.

[0129] In this embodiment, after obtaining the binarized crack prediction map, the binarized crack prediction map can be processed by connected component labeling to obtain the crack region of the inner wall of the cable tunnel.

[0130] Optionally, given a binarized image In this context, a pixel value of 0 represents the background, and a pixel value of 1 represents the foreground (crack). Create a marker image of the same size. This is used to store the label of the connected region to which each pixel belongs. Initially, all pixel values ​​are 0.

[0131] Furthermore, from the labeled images The image is scanned starting from the top left corner, checking pixel by pixel. If the current pixel... Then it is necessary to determine which connected component the pixel belongs to. Using 4-neighborhood, when scanning the current pixel, check the pixels above and to the left of it, that is... and If the pixels above or to the left are also marked as a connected region, use equation (11) to set the current pixel... It should also be marked as the area's label:

[0132] (11);

[0133] Here, `min` indicates that if at least one of the two neighboring pixels is already labeled, the pixel with the smallest label is selected. If neither the pixels above nor to the left are labeled as connected components (i.e., both are background or unprocessed), a new label is assigned to the current pixel. , where l is a new unique label representing a new connected region.

[0134] Finally, the resulting labeled map In this process, each crack region will be assigned a unique label. Connected component labels have different parameters when labeling crack segmentation results, such as area, length, and width. Assume n is the nth connected component, where n = 1, 2, 3… The average of the horizontal coordinates of the i-th row of the nth connected component is… The average value of the j-th column vertical coordinates of the n-th connected component is Therefore, the center coordinates of the nth connected component are defined as in formula (12):

[0135] (12);

[0136] Wherein, the center point of the nth connected component The average value to the edge of the region is Equation (13):

[0137] (13);

[0138] in, This represents the sum of the edge pixels of the nth connected component. Then it represents the coordinates of the k-th edge pixel in the n-th connected component, so the standard deviation of the distance in the n-th connected component is defined by equation (14):

[0139] (14);

[0140] It should be noted that the connected component labeling algorithm and distance standard deviation are used to identify and eliminate irrelevant interference regions. A certain threshold is set, when... At this time, the corresponding connected components are deleted to eliminate irrelevant interference regions. Morphological operations are used to smooth the crack boundaries, remove noise points, fill small cavities, and further optimize the shape of the crack region.

[0141] In this embodiment, the crack prediction map is binarized to obtain a binarized crack prediction map. Then, the binarized crack prediction map is subjected to connected component labeling to obtain the crack region of the inner wall of the cable tunnel. This method can efficiently obtain the crack region of the inner wall of the cable tunnel.

[0142] In one embodiment, in Figure 2 Based on the illustrated embodiments, as Figure 7 As shown, the above method also includes:

[0143] S205. Preprocess the infrared image to obtain a preprocessed infrared image; the preprocessing includes noise reduction, contrast reduction, normalization, annotation, and data augmentation.

[0144] In this embodiment, after obtaining the infrared image, the infrared image can be preprocessed to obtain a preprocessed infrared image.

[0145] Optionally, Gaussian filtering can be applied to the infrared image using equation (15) to eliminate noise:

[0146] (15);

[0147] in, It is a Gaussian kernel with a kernel size of . , Indicates the pixel position of the filtered image. The value, Represents the pixel position in the original image The pixel value.

[0148] In addition, to improve the contrast between the crack and the background, an adaptive contrast enhancement technique is used as shown in equation (16):

[0149] (16);

[0150] in, and For local mean and standard deviation, Prevent division by zero.

[0151] Alternatively, the image can be normalized to [0,1] using formula (17):

[0152] (17)

[0153] Furthermore, the dataset was created using the LabelMe annotation tool on the infrared images. Manual annotation was performed to obtain the true labels of the cracks. ,in Represents the crack pixels. Construct the original dataset. Initially, N=5000.

[0154] Optionally, data augmentation can be performed by expanding the data volume to N=20000 through geometric and photometric transformations to improve the robustness of the model.

[0155] The aforementioned S202 "encoder that inputs infrared images to the crack segmentation model" includes:

[0156] S202. Input the preprocessed infrared image into the encoder of the crack segmentation model.

[0157] In this embodiment, after obtaining the preprocessed infrared image, the preprocessed infrared image can be input into the encoder of the crack segmentation model.

[0158] In this embodiment, the infrared image is input into the encoder of the crack segmentation model for encoding processing to obtain the encoded infrared image features, which provides a basis for subsequently determining the crack area in the inner wall of the cable tunnel in the crack prediction map.

[0159] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0160] Based on the same inventive concept, this application also provides a cable tunnel crack identification device for implementing the above-described cable tunnel crack identification method. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations of one or more cable tunnel crack identification device embodiments provided below can be found in the limitations of the cable tunnel crack identification method described above, and will not be repeated here.

[0161] In one exemplary embodiment, such as Figure 8 As shown, a crack identification device for cable tunnels is provided, comprising: a data acquisition module 10, an encoding module 11, a decoding module 12, and a determination module 13, wherein:

[0162] The acquisition module 10 is used to acquire infrared images of the inner wall of the cable tunnel.

[0163] Encoding module 11 is used to input the infrared image into the encoder of the crack segmentation model to obtain the encoded infrared image features; the encoder is ResNet-34.

[0164] The decoding module 12 is used to input the encoded infrared image features into the decoder of the crack segmentation model, and obtain the crack prediction map through the decoder.

[0165] Module 13 is used to determine the crack area in the inner wall of the cable tunnel based on the crack prediction map.

[0166] In an exemplary embodiment, the encoding module 11 includes: a first convolutional unit and a second convolutional unit, wherein:

[0167] The first convolutional unit is specifically used to perform convolution processing on the infrared image through the first convolutional layer in the encoder to obtain the first feature map;

[0168] The second convolutional unit is specifically used to perform two convolution operations on the first feature map through the residual block in the encoder to obtain the encoded infrared image features.

[0169] In an exemplary embodiment, the encoding module 11 further includes: a normalization processing unit and an acquisition unit, wherein:

[0170] The normalization processing unit is specifically used to perform batch normalization processing on the encoded infrared image features to obtain batch-normalized infrared image features.

[0171] The acquisition unit is specifically used to obtain the output feature map based on the batch-normalized infrared image features and the first feature map.

[0172] In an exemplary embodiment, the decoding module 12 includes: a downsampling processing unit, a splicing processing unit, and an activation processing unit, wherein:

[0173] The downsampling processing unit performs downsampling processing on the output feature map to obtain the downsampled feature map.

[0174] The splicing processing unit specifically splices the downsampled feature map and the output feature map to obtain a spliced ​​feature map.

[0175] The activation processing unit is specifically used to input the stitched feature map into the activation layer of the decoder for activation processing to obtain the crack prediction map.

[0176] In an exemplary embodiment, the determination module 13 includes: a binarization processing unit and a tagging unit, wherein:

[0177] The binarization processing unit is specifically used to perform binarization processing on the crack prediction map to obtain the binarized crack prediction map.

[0178] The marking unit is specifically used to perform connected component marking processing on the binarized crack prediction map to obtain the crack region on the inner wall of the cable tunnel.

[0179] In one exemplary embodiment, the above-described apparatus further includes:

[0180] The preprocessing module is used to preprocess the infrared image to obtain the preprocessed infrared image; the preprocessing includes noise reduction, contrast reduction, normalization, annotation and data augmentation.

[0181] The aforementioned encoding module is used to input the preprocessed infrared image into the encoder of the crack segmentation model.

[0182] The modules in the aforementioned cable tunnel crack detection device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0183] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0184] Acquire infrared images of the inner wall of the cable tunnel;

[0185] The infrared image is input into the encoder of the crack segmentation model to obtain the encoded infrared image features; the encoder is ResNet-34.

[0186] The encoded infrared image features are input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder.

[0187] Based on the crack prediction map, the crack area in the inner wall of the cable tunnel was determined.

[0188] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0189] The infrared image is convolved by the first convolutional layer in the encoder to obtain the first feature map;

[0190] The first feature map is convolved twice by the residual block in the encoder to obtain the encoded infrared image features.

[0191] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0192] Batch normalization is performed on the encoded infrared image features to obtain batch normalized infrared image features.

[0193] The output feature map is obtained based on the batch-normalized infrared image features and the first feature map;

[0194] The encoded infrared image features are input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder, including:

[0195] The output feature map is input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder.

[0196] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0197] The output feature map is downsampled to obtain the downsampled feature map.

[0198] The downsampled feature map and the output feature map are concatenated to obtain the concatenated feature map.

[0199] The stitched feature map is input into the activation layer of the decoder for activation processing to obtain the crack prediction map.

[0200] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0201] The crack prediction map is binarized to obtain the binarized crack prediction map;

[0202] The crack prediction map after binarization is processed by connected component labeling to obtain the crack region on the inner wall of the cable tunnel.

[0203] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0204] The infrared image is preprocessed to obtain the preprocessed infrared image; the preprocessing includes noise reduction, contrast reduction, normalization, annotation and data augmentation.

[0205] The encoder of the crack segmentation model inputs the infrared image, including:

[0206] The preprocessed infrared image is input into the encoder of the crack segmentation model.

[0207] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0208] Acquire infrared images of the inner wall of the cable tunnel;

[0209] The infrared image is input into the encoder of the crack segmentation model to obtain the encoded infrared image features; the encoder is ResNet-34.

[0210] The encoded infrared image features are input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder.

[0211] Based on the crack prediction map, the crack area in the inner wall of the cable tunnel was determined.

[0212] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0213] The infrared image is convolved by the first convolutional layer in the encoder to obtain the first feature map;

[0214] The first feature map is convolved twice by the residual block in the encoder to obtain the encoded infrared image features.

[0215] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0216] Batch normalization is performed on the encoded infrared image features to obtain batch normalized infrared image features.

[0217] The output feature map is obtained based on the batch-normalized infrared image features and the first feature map;

[0218] The encoded infrared image features are input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder, including:

[0219] The output feature map is input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder.

[0220] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0221] The output feature map is downsampled to obtain the downsampled feature map.

[0222] The downsampled feature map and the output feature map are concatenated to obtain the concatenated feature map.

[0223] The stitched feature map is input into the activation layer of the decoder for activation processing to obtain the crack prediction map.

[0224] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0225] The crack prediction map is binarized to obtain the binarized crack prediction map;

[0226] The crack prediction map after binarization is processed by connected component labeling to obtain the crack region on the inner wall of the cable tunnel.

[0227] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0228] The infrared image is preprocessed to obtain the preprocessed infrared image; the preprocessing includes noise reduction, contrast reduction, normalization, annotation and data augmentation.

[0229] The encoder of the crack segmentation model inputs the infrared image, including:

[0230] The preprocessed infrared image is input into the encoder of the crack segmentation model.

[0231] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0232] Acquire infrared images of the inner wall of the cable tunnel;

[0233] The infrared image is input into the encoder of the crack segmentation model to obtain the encoded infrared image features; the encoder is ResNet-34.

[0234] The encoded infrared image features are input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder.

[0235] Based on the crack prediction map, the crack area in the inner wall of the cable tunnel was determined.

[0236] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0237] The infrared image is convolved by the first convolutional layer in the encoder to obtain the first feature map;

[0238] The first feature map is convolved twice by the residual block in the encoder to obtain the encoded infrared image features.

[0239] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0240] Batch normalization is performed on the encoded infrared image features to obtain batch normalized infrared image features.

[0241] The output feature map is obtained based on the batch-normalized infrared image features and the first feature map;

[0242] The encoded infrared image features are input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder, including:

[0243] The output feature map is input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder.

[0244] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0245] The output feature map is downsampled to obtain the downsampled feature map.

[0246] The downsampled feature map and the output feature map are concatenated to obtain the concatenated feature map.

[0247] The stitched feature map is input into the activation layer of the decoder for activation processing to obtain the crack prediction map.

[0248] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0249] The crack prediction map is binarized to obtain the binarized crack prediction map;

[0250] The crack prediction map after binarization is processed by connected component labeling to obtain the crack region on the inner wall of the cable tunnel.

[0251] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0252] The infrared image is preprocessed to obtain the preprocessed infrared image; the preprocessing includes noise reduction, contrast reduction, normalization, annotation and data augmentation.

[0253] The encoder of the crack segmentation model inputs the infrared image, including:

[0254] The preprocessed infrared image is input into the encoder of the crack segmentation model.

[0255] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0256] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0257] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for identifying cracks in cable tunnels, characterized in that, The method includes: Acquire infrared images of the inner wall of the cable tunnel; The infrared image is input into the encoder of the crack segmentation model to obtain the encoded infrared image features; the encoder is ResNet-34. The encoded infrared image features are input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder. Based on the crack prediction map, the crack area in the inner wall of the cable tunnel is determined.

2. The method according to claim 1, characterized in that, The step of inputting the infrared image into the encoder of the crack segmentation model to obtain the encoded infrared image features includes: The infrared image is convolved by the first convolutional layer in the encoder to obtain a first feature map; The first feature map is convolved twice using the residual block in the encoder to obtain the encoded infrared image features.

3. The method according to claim 2, characterized in that, The method further includes: The encoded infrared image features are subjected to batch normalization to obtain batch-normalized infrared image features. Based on the batch-normalized infrared image features and the first feature map, an output feature map is obtained; The step of inputting the encoded infrared image features into the decoder of the crack segmentation model and obtaining the crack prediction map through the decoder includes: The output feature map is input into the decoder of the crack segmentation model, and the crack prediction map is obtained through the decoder.

4. The method according to claim 3, characterized in that, The step of inputting the output feature map into the decoder of the crack segmentation model and obtaining the crack prediction map through the decoder includes: The output feature map is downsampled to obtain a downsampled feature map. The downsampled feature map and the output feature map are concatenated to obtain a concatenated feature map. The stitched feature map is input into the activation layer of the decoder for activation processing to obtain the crack prediction map.

5. The method according to claim 1, characterized in that, The step of determining the crack region in the inner wall of the cable tunnel based on the crack prediction map includes: The crack prediction map is binarized to obtain a binarized crack prediction map. The crack prediction map after binarization is processed by connected component labeling to obtain the crack region on the inner wall of the cable tunnel.

6. The method according to claim 1, characterized in that, The method further includes: The infrared image is preprocessed to obtain a preprocessed infrared image; the preprocessing includes noise reduction, contrast reduction, normalization, annotation, and data augmentation. The encoder that inputs the infrared image into the crack segmentation model includes: The preprocessed infrared image is input into the encoder of the crack segmentation model.

7. A crack detection device for cable tunnels, characterized in that, The device includes: The acquisition module is used to acquire infrared images of the inner wall of the cable tunnel; An encoding module is used to input the infrared image into the encoder of the crack segmentation model to obtain encoded infrared image features; the encoder is ResNet-34. The decoding module is used to input the encoded infrared image features into the decoder of the crack segmentation model, and obtain the crack prediction map through the decoder; A determination module is used to determine the crack area in the inner wall of the cable tunnel based on the crack prediction map.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images