Cable tunnel crack water seepage identification method and device and computer equipment

Abstract

The invention relates to a cable tunnel crack water seepage identification method and device and computer equipment. The method comprises the following steps: performing phase reversal processing on a cable tunnel image to obtain a phase reversal image; processing the cable tunnel image and the inverted image through a fusion sub-model in the recognition model, and determining a biphase fusion feature; performing data enhancement on the cable tunnel image based on the biphase fusion features and an enhancement sub-model in the recognition model to obtain an enhanced image; inputting the enhanced image and the biphase fusion feature into a water leakage segmentation marking sub-model in the recognition model to obtain a tunnel leakage water segmentation marking graph output by the water leakage segmentation marking sub-model; the water seepage area in the cable tunnel is determined based on the tunnel leakage water segmentation mark graph, and the recognition accuracy of the cable tunnel leakage water is improved.

Description

Technical Field

[0001] The present application relates to the field of image recognition technology, and in particular to a method, device and computer equipment for identifying water seepage in cracks in cable tunnels. Background Art

[0002] Cable tunnels are crucial infrastructure for power transmission, and their structural safety is directly linked to the stable operation of power systems. However, due to long-term environmental influences such as groundwater, soil pressure, and temperature fluctuations, cable tunnels are prone to cracks, which in turn lead to water seepage. Water seepage not only accelerates the aging of tunnel structures but can also cause cable short circuits, equipment damage, and even serious safety incidents. Therefore, timely identification and diagnosis of water seepage in cable tunnel cracks is crucial to ensuring the safe operation of power systems.

[0003] In traditional technologies, the identification of water seepage in cable tunnel cracks mainly relies on manual inspections and traditional monitoring methods. Traditional monitoring methods include the use of humidity sensors, temperature sensors and other equipment.

[0004] However, the traditional method for identifying water seepage in cable tunnel cracks has the problem of inaccurate identification. Summary of the Invention

[0005] Based on this, it is necessary to provide a method, device and computer equipment for identifying water seepage in cable tunnel cracks that can improve identification accuracy in response to the above technical problems.

[0006] In a first aspect, the present application provides a method for identifying water seepage in cracks in cable tunnels, comprising:

[0007] Inverting the cable tunnel image to obtain an inverted image;

[0008] Processing the cable tunnel image and the inverted image by using a fusion sub-model in the recognition model to determine a dual-phase fusion feature;

[0009] Based on the two-phase fusion feature and the enhancer model in the recognition model, data enhancement is performed on the cable tunnel image to obtain an enhanced image;

[0010] Inputting the enhanced image and the dual-phase fusion feature into the water leakage segmentation mark sub-model in the recognition model to obtain a tunnel water leakage segmentation mark map output by the water leakage segmentation mark sub-model;

[0011] Determine water seepage areas in cable tunnels based on tunnel water leakage segmentation marker maps.

[0012] In one embodiment, the water leakage segmentation and marking sub-model includes an adaptive module and a segment arbitrary module, and the step of inputting the enhanced image and the two-phase fusion feature into the water leakage segmentation and marking sub-model in the recognition model to obtain a tunnel water leakage segmentation and marking map output by the water leakage segmentation and marking sub-model includes:

[0013] Separating pixel data of the enhanced image according to pixel colors to obtain three-channel features corresponding to the enhanced image;

[0014] Inputting the three-channel features into an adaptive module for processing to obtain first processed data;

[0015] The first processed data, the three-channel features and the two-phase fusion features are stacked by the segment arbitrary module to obtain a tunnel water leakage segmentation mark map output by the water leakage segmentation mark sub-model.

[0016] In one embodiment, inputting the three-channel features into an adaptive module for processing to obtain first processed data includes:

[0017] Inputting the three-channel features into the convolution layer of the adaptive module for feature extraction to obtain single-channel features corresponding to the three-channel features; the single-channel features include red channel data, green channel data, and blue channel data;

[0018] performing a maximum pooling operation on the single-channel features to enhance texture features related to edge detection of leaking areas;

[0019] The enhanced texture features and the downsampled enhanced image are superimposed to obtain the first processed data.

[0020] In one embodiment, the processing of the cable tunnel image and the inverted image by a fusion sub-model in the recognition model to determine the dual-phase fusion feature includes:

[0021] Inputting the cable tunnel image and the inverted image into the fusion sub-model; the fusion sub-model includes a convolution module and a fusion module;

[0022] Performing multi-layer convolution processing on the cable tunnel image by the convolution module to obtain a first feature map;

[0023] Performing multi-layer convolution processing on the inverted image by the convolution module to obtain a second feature map;

[0024] The first feature map and the second feature map are fused to obtain a two-phase fusion feature.

[0025] In one embodiment, the convolution module includes multiple convolution layers, multiple activation function layers, and an attention mechanism layer, wherein the multiple activation function layers are arranged between some of the convolution layers in the multiple convolution layers, and the attention mechanism layer is arranged after the convolution layers;

[0026] The multi-layer convolutional layer includes a multi-layer first convolutional layer, a multi-layer second convolutional layer and a multi-layer third convolutional layer. The third convolutional layer corresponds to the first convolutional layer one-to-one, and the third convolutional layer is a skip convolutional layer of the corresponding first convolutional layer.

[0027] In one embodiment, the training process of the recognition model includes:

[0028] Constructing an initial recognition model; the initial recognition model includes an initial fusion sub-model, an initial enhancement sub-model and an initial leaky segmentation marking sub-model;

[0029] Constructing a no-reference loss function and a detection effect evaluation index; the detection effect evaluation index is used to evaluate the detection effect of the overlap and similarity between the true label and the predicted mask;

[0030] A cable tunnel water leakage dark light dataset is obtained, and the initial recognition model is trained according to the cable tunnel water leakage dark light dataset, a no-reference loss function, and the detection effect evaluation index to obtain the recognition model.

[0031] In one embodiment, the no-reference loss function is determined based on at least one of a spatial consistency loss function, an exposure control loss function, a color consistency loss function, a total variation loss function, and a channel consistency loss function.

[0032] In a second aspect, the present application further provides a device for identifying water seepage in cracks in cable tunnels, comprising:

[0033] An inversion module, used for performing inversion processing on the cable tunnel image to obtain an inverted image;

[0034] a fusion module, configured to process the cable tunnel image and the inverted-phase image by using a fusion sub-model in a recognition model to determine a dual-phase fusion feature;

[0035] an enhancement module, configured to perform data enhancement on the cable tunnel image based on the dual-phase fusion feature and an enhancement sub-model in the recognition model to obtain an enhanced image;

[0036] an acquisition module, configured to input the enhanced image and the dual-phase fusion feature into a water leakage segmentation mark sub-model in the recognition model, so as to obtain a tunnel water leakage segmentation mark map output by the water leakage segmentation mark sub-model;

[0037] The determination module is used to determine the water seepage area in the cable tunnel based on the tunnel water leakage segmentation mark map.

[0038] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0039] Inverting the cable tunnel image to obtain an inverted image;

[0040] Processing the cable tunnel image and the inverted image by using a fusion sub-model in the recognition model to determine a dual-phase fusion feature;

[0041] Based on the two-phase fusion feature and the enhancer model in the recognition model, data enhancement is performed on the cable tunnel image to obtain an enhanced image;

[0042] Inputting the enhanced image and the dual-phase fusion feature into the water leakage segmentation mark sub-model in the recognition model to obtain a tunnel water leakage segmentation mark map output by the water leakage segmentation mark sub-model;

[0043] Determine water seepage areas in cable tunnels based on tunnel water leakage segmentation marker maps.

[0044] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0045] Inverting the cable tunnel image to obtain an inverted image;

[0046] Processing the cable tunnel image and the inverted image by using a fusion sub-model in the recognition model to determine a dual-phase fusion feature;

[0047] Based on the two-phase fusion feature and the enhancer model in the recognition model, data enhancement is performed on the cable tunnel image to obtain an enhanced image;

[0048] Inputting the enhanced image and the dual-phase fusion feature into the water leakage segmentation mark sub-model in the recognition model to obtain a tunnel water leakage segmentation mark map output by the water leakage segmentation mark sub-model;

[0049] Determine water seepage areas in cable tunnels based on tunnel water leakage segmentation marker maps.

[0050] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the following steps:

[0051] Inverting the cable tunnel image to obtain an inverted image;

[0052] Processing the cable tunnel image and the inverted image by using a fusion sub-model in the recognition model to determine a dual-phase fusion feature;

[0053] Based on the two-phase fusion feature and the enhancer model in the recognition model, data enhancement is performed on the cable tunnel image to obtain an enhanced image;

[0054] Inputting the enhanced image and the dual-phase fusion feature into the water leakage segmentation mark sub-model in the recognition model to obtain a tunnel water leakage segmentation mark map output by the water leakage segmentation mark sub-model;

[0055] Determine water seepage areas in cable tunnels based on tunnel water leakage segmentation marker maps.

[0056] The above-mentioned cable tunnel crack seepage identification method, device, and computer equipment perform inversion processing on the cable tunnel image to obtain an inverted image; process the cable tunnel image and the inverted image using a fusion sub-model in the recognition model to determine a bi-phase fusion feature; perform data enhancement on the cable tunnel image based on the bi-phase fusion feature and the enhancement sub-model in the recognition model to obtain an enhanced image; input the enhanced image and the bi-phase fusion feature into the water leakage segmentation and marking sub-model in the recognition model to obtain a tunnel water leakage segmentation and marking map output by the water leakage segmentation and marking sub-model; and determine the water leakage area in the cable tunnel based on the tunnel water leakage segmentation and marking map. The inverted image corresponding to the cable tunnel image is obtained, and the bi-phase fusion feature of the cable tunnel image and the inverted image is used to perform data enhancement on the cable tunnel image, thereby improving the accuracy of the enhanced data recognition. The enhanced data and the bi-phase fusion feature are used as input to the water leakage segmentation and marking sub-model, making the input data more reliable and accurate, thereby improving the input of the water leakage segmentation and marking sub-model, and improving the accuracy of cable tunnel water leakage recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0058] Figure 1 A diagram showing an application environment of a method for identifying water seepage in cable tunnel cracks according to an embodiment;

[0059] Figure 2 This is a flow chart of a method for identifying water seepage in cracks in a cable tunnel according to one embodiment;

[0060] Figure 3This is a second flow chart of a method for identifying water seepage in cable tunnel cracks in one embodiment;

[0061] Figure 4 A schematic diagram of an arbitrary module of a fragment in one embodiment;

[0062] Figure 5 This is a third flow chart of a method for identifying water seepage in cracks in a cable tunnel according to an embodiment;

[0063] Figure 6 is a schematic diagram of an adaptive module in one embodiment;

[0064] Figure 7 This is a fourth flow chart of a method for identifying water seepage in cracks in a cable tunnel according to an embodiment;

[0065] Figure 8 Schematic diagram of a convolution module in one embodiment;

[0066] Figure 9 This is a fifth flow chart of a method for identifying water seepage in cracks in a cable tunnel according to an embodiment;

[0067] Figure 10 A schematic diagram of a framework of an identification model in one embodiment;

[0068] Figure 11 1 is a structural block diagram of a device for identifying water seepage in cracks in cable tunnels in one embodiment. DETAILED DESCRIPTION

[0069] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0070] Cable tunnels are crucial infrastructure for power transmission, and their structural safety is directly linked to the stable operation of power systems. However, due to long-term environmental influences such as groundwater, soil pressure, and temperature fluctuations, cable tunnels are prone to cracks, which in turn lead to water seepage. Water seepage not only accelerates the aging of tunnel structures but can also cause cable short circuits, equipment damage, and even serious safety incidents. Therefore, timely identification and diagnosis of water seepage in cable tunnel cracks is crucial to ensuring the safe operation of power systems.

[0071] Traditionally, the identification and diagnosis of cracks and water seepage in cable tunnels relies primarily on manual inspections and traditional monitoring methods. Manual inspections typically involve professionals periodically entering the tunnel to determine the presence of cracks and water seepage through visual inspections, tapping, and listening. While intuitive, this method suffers from low efficiency, limited coverage, and subjectivity, posing a safety threat to inspectors in high-risk environments. Traditional monitoring methods, such as humidity and temperature sensors, indirectly determine the presence of water seepage by monitoring humidity and temperature changes within the tunnel. However, these methods often provide only limited information, cannot accurately locate cracks, and are susceptible to environmental interference, leading to false or missed reports. Furthermore, the high installation and maintenance costs of traditional monitoring equipment make it difficult to implement across large tunnel networks.

[0072] With technological advancements, advanced detection methods have been gradually introduced, such as infrared thermal imaging and ultrasonic testing. Infrared thermal imaging can indirectly identify areas of water seepage by capturing the temperature distribution on the tunnel surface, but its limited resolution makes it difficult to detect even tiny cracks. Ultrasonic testing, which utilizes the propagation characteristics of sound waves through materials, can detect cracks, but its operation is complex and requires high-level technical expertise.

[0073] The identification and diagnosis of water seepage in cable tunnel cracks primarily relies on manual inspections and traditional monitoring methods, which suffer from low efficiency, poor accuracy, and high costs. Therefore, an efficient, accurate, and low-cost method for identifying and diagnosing water seepage in cable tunnel cracks is urgently needed to meet the growing demand for power infrastructure maintenance.

[0074] The cable tunnel crack water seepage identification method provided by the embodiment of the present application can be applied to Figure 1 In the application environment shown in FIG. , the computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 1As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store cable tunnel crack seepage identification data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a cable tunnel crack seepage identification method is implemented.

[0075] Those skilled in the art will understand that Figure 1 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0076] In one embodiment, Figure 2 As shown, a method for identifying water seepage in cable tunnel cracks is provided. Figure 1 The following is an example of a server in the example, including:

[0077] S201 , performing inversion processing on the cable tunnel image to obtain an inverted image.

[0078] In an embodiment of the present application, the cable tunnel image is a dark light image, and the inversion processing is to perform color inversion processing on the cable tunnel image, convert the dark area (low pixel value) in the cable tunnel image into a bright area (high pixel value), and change the value of each pixel channel in the cable tunnel image from the current value I to 255-I (for an 8-bit image). Exemplarily, the cable tunnel image I is inverted, and the formula for obtaining the inverted image I' can be: I'=255-I.

[0079] S202 , processing the cable tunnel image and the inverted-phase image by using a fusion sub-model in the recognition model to determine a dual-phase fusion feature.

[0080] In an embodiment of the present application, the recognition model includes a fusion sub-model, an enhancement sub-model and a marking sub-model. The cable tunnel image and the inverted image are input into the fusion sub-model for feature extraction to obtain a first feature of the cable tunnel image and a second feature of the inverted image. Further, the first feature and the second feature are fused to obtain a dual-phase fusion feature.

[0081] For example, the fusion sub-model may merge the first feature and the second feature to obtain a two-phase fusion feature, wherein the merging operation may include any of the following:

[0082] (1) Concatenation: directly concatenate two feature vectors or feature maps in the channel dimension;

[0083] (2) Element-wise operations: addition, multiplication, averaging, etc.

[0084] (3) Weighted fusion: Use the attention mechanism to learn the importance weights of different channels or spatial positions, and then perform weighted fusion;

[0085] (4) Learning-based fusion: Use a small neural network to learn how to optimally combine two feature streams.

[0086] Optionally, the dual-phase fusion feature can be a curve parameter mapping of the adaptive light enhancement curve ,in Represents pixel coordinates.

[0087] S203 , based on the dual-phase fusion features and the enhancer model in the recognition model, data enhancement is performed on the cable tunnel image to obtain an enhanced image.

[0088] In the embodiment of the present application, the dual-phase fusion feature and the cable tunnel image are input into the enhancement sub-model, and the cable tunnel image is data enhanced to obtain an enhanced image. For example, the cable tunnel image can be input into the amplitude iteration controller, according to The brightness of the light is enhanced by and the number of iterations The adaptive light enhancement curve ALE is further iteratively adjusted to obtain an enhanced image.

[0089] Optionally, select a dark light image dataset such as SID or DPED, extract a certain number of images from each dataset, record the enhancement and suppression amplitude values ​​of the images at different exposure levels, and select the best fitting amplitude curve. To calculate the optimal enhancement or suppression amplitude required for the input image at different exposure levels. The amplitude curve is expressed as: ,in is the normalized image pixel average. Further, the iterative curve The iterative scheme required to fit the different exposure images is then Determine the number of iterations, the number of iterations , iterative curve It can be shown as formula 1:

[0090] (Formula 1)

[0091] Where n is the number of iterations, This is a rounding operation.

[0092] Furthermore, the curve parameters are mapped , amplitude intensity , number of iterations n, input image brightness By introducing the designed adaptive light enhancement (ALE) curve, the light level of the image is iteratively enhanced or suppressed to a specified interval. After multiple iterations, a dark-light enhanced image is obtained. The ALE curve is shown in Equation 2:

[0093] (Equation 2)

[0094] in, for The enhanced results, is the initial input image , = 0.6 as the target appropriate exposure level for the image.

[0095] In the embodiments of this application, by simultaneously considering the input image and its inverted version, the model is able to learn more comprehensive lighting information, significantly improving its ability to process images with different exposure levels. This method designs an adaptive light enhancement curve and an amplitude iterative controller that can precisely control the intensity of image enhancement or suppression, effectively avoiding noise amplification and achieving high-quality enhancement of low-light images. This not only improves the versatility of the model but also enhances its application advantages on actual devices, providing reliable technical support for image processing in low-light environments.

[0096] S204 , inputting the enhanced image and the two-phase fusion features into the water leakage segmentation and marking sub-model in the recognition model to obtain a tunnel water leakage segmentation and marking map output by the water leakage segmentation and marking sub-model.

[0097] In an embodiment of the present application, the cable tunnel image after image enhancement processing and its corresponding two-phase fusion features are synchronously input into the water leakage segmentation marking sub-model within the recognition model. The water leakage segmentation marking sub-model can jointly analyze the two types of input data through the convolutional neural network architecture, and output a tunnel water leakage segmentation marking map with pixel-level prediction accuracy.

[0098] Optionally, the tunnel water leakage segmentation marker map can be a binary matrix or a probability heat map. When the tunnel water leakage segmentation marker map is a binary matrix, specific pixel values ​​represent the spatial location of the leakage. When the tunnel water leakage segmentation marker map is a probability heat map, high-probability areas represent the spatial location of the leakage.

[0099] S205 : Determine the water seepage area in the cable tunnel based on the tunnel water leakage segmentation marker map.

[0100] In this embodiment, based on the spatial coordinate mapping relationship in the tunnel water leakage segmentation marker map, the activated pixel areas (or connected domains with a probability exceeding a preset threshold) in the marker map are mapped to corresponding locations in the original cable tunnel image, accurately delineating the geometric boundaries of the water leakage area. Ultimately, a cable tunnel structure diagram with spatial annotations of the water leakage area and quantitative leakage range data are output. For example, the quantitative leakage range data includes area, location coordinates, etc.

[0101] In the above-mentioned application embodiment, the cable tunnel image is inverted to obtain an inverted image; the cable tunnel image and the inverted image are processed by the fusion sub-model in the recognition model to determine the bi-phase fusion feature; based on the bi-phase fusion feature and the enhancement sub-model in the recognition model, the cable tunnel image is data enhanced to obtain an enhanced image; the enhanced image and the bi-phase fusion feature are input into the water leakage segmentation labeling sub-model in the recognition model to obtain the tunnel water leakage segmentation labeling map output by the water leakage segmentation labeling sub-model; the water leakage area in the cable tunnel is determined based on the tunnel water leakage segmentation labeling map. The inverted image corresponding to the cable tunnel image is obtained, and the bi-phase fusion feature of the cable tunnel image and the inverted image is used to perform data enhancement on the cable tunnel image, thereby improving the accuracy of the enhanced data recognition. Moreover, the enhanced data and the bi-phase fusion feature are used as the input of the water leakage segmentation labeling sub-model, and the input data is more reliable and accurate, making the input of the water leakage segmentation labeling sub-model more reliable and accurate, thereby improving the accuracy of the recognition of water leakage in the cable tunnel.

[0102] In one embodiment, an implementation of the above S203 is provided, such as Figure 3 As shown, the water leakage segmentation and marking sub-model includes an adaptive module and a segment arbitrary module. The above-mentioned "inputting the enhanced image and the two-phase fusion feature into the water leakage segmentation and marking sub-model in the recognition model to obtain the tunnel water leakage segmentation and marking map output by the water leakage segmentation and marking sub-model" includes:

[0103] S301 , separating pixel data of the enhanced image according to pixel colors to obtain three-channel features corresponding to the enhanced image.

[0104] In an embodiment of the present application, a color component analysis operation can be performed on each pixel of the enhanced image based on an RGB color space model or a preset color separation algorithm. Optionally, a pixel-level color channel decoupling algorithm can be used to decompose the composite color value of each pixel into three independent spectral components: red (R), green (G), and blue (B). Furthermore, a two-dimensional feature matrix with the same size as the original image is constructed for each pixel:

[0105] (1) Red channel feature matrix: stores the R component values ​​of all pixels;

[0106] (2) Green channel feature matrix: stores the G component values ​​of all pixels;

[0107] (3) Blue channel feature matrix: stores the B component values ​​of all pixels.

[0108] Furthermore, the output is a three-channel separation feature set that constitutes the enhanced image.

[0109] S302: Input the three-channel features into the adaptive module for processing to obtain first processed data.

[0110] In an embodiment of the present application, the adaptive module can be a trained deep learning network model, and the three-channel features are input into the adaptive module for processing to enhance the texture features necessary for detecting the edges of the leaking area, and the first processed data is determined based on the enhanced texture features.

[0111] As an optional implementation, the three-channel separation feature set is input into the adaptive module, and the first processed data is generated through the following processing flow:

[0112] (1) Channel importance assessment: A learnable weight generator is used to analyze the statistical characteristics of each channel feature and dynamically generate a channel weight vector;

[0113] (2) Weighted feature fusion: Linearly fuse the three-channel features according to the weight vector to generate an enhanced feature matrix;

[0114] (3) Spatial adaptive enhancement: Use the dilated convolutional layer to extract multi-scale context information and combine it with the spatial attention mechanism to output the first processed data.

[0115] S303 , stacking the first processed data, the three-channel features, and the two-phase fusion features through a fragment arbitrary module to obtain a tunnel water leakage segmentation labeling map output by a water leakage segmentation labeling sub-model.

[0116] In the embodiment of the present application, the first processed data output by the adaptive module is used as input, and the position information in the feature sequence is marked by the position encoding module PE. Here, the dual-phase fusion feature output is introduced to perform multi-layer original feature stacking, such as Figure 4As shown in the figure, the leaky segmentation labeling submodel is a Spatial Attention Multi-scale Aggregation (SAMA) module. Adaptive modules are inserted before and after each self-attention feedforward Transformer block. The input of each adaptive module is the parameter adjustment output of the previous adaptive module. Through multiple adaptive modules, effective adjustment of network parameters is achieved. The self-attention feedforward Transformer block consists of a multi-head attention layer (MHA) and a multi-layer perceptron (MLP) layer with layer normalization. A pair of MLP layers are added between each Transformer block as adaptive modules, as shown in Equation 3:

[0117] (Equation 3)

[0118] in Through the adaptive module The obtained prompt, Pi is attached to each Transformer block of the SAM encoder, GELU is the activation function, Embedding features for patches and high frequency component characteristics The sum can be expressed as: .

[0119] In an embodiment of the present application, the first processed data, three-channel features and two-phase fusion features are input into the trained segment arbitrary module for stacking processing to obtain processed feature data, and the processed feature data is converted into a tunnel water leakage segmentation label map.

[0120] In the above application embodiment, the three-channel features are first processed by the adaptive module, and then the first processed data obtained is used as the input of the fragment arbitrary module, and stacked with the three-channel features and the two-phase fusion features, thereby improving the accuracy of the tunnel water leakage segmentation marker map.

[0121] In one embodiment, an implementation of the above S302 is provided, such as Figure 5 As shown, the above “inputting the three-channel features into the adaptive module for processing to obtain first processed data” includes:

[0122] S401, input the three-channel features into the convolution layer of the adaptive module for feature extraction to obtain single-channel features corresponding to the three-channel features; the single-channel features include red channel data, green channel data and blue channel data.

[0123] S402: performing a maximum pooling operation on the single-channel features to enhance texture features related to edge detection of the leaking area.

[0124] S403: Superimpose the enhanced texture features and the downsampled enhanced image to obtain first processed data.

[0125] In the embodiment of the present application, the enhanced output image output Perform R, G, B three-channel segmentation and input into the adaptive module. Figure 6 As shown in the figure, the adaptive module first extracts single-channel features through a 1×1 convolution layer, and then enhances the texture features necessary for detecting the edges of the leaking area through a maximum pooling layer. The enhanced texture features are repeatedly superimposed on the downsampled image by channel and element to retain the texture features extracted from the convolution layer across channels as the input features of the adaptive module of any model of the fragment The design of the adaptive module can be expressed as: .in, is the dark light enhanced output detection image. is downsampling. It is a convolutional layer with a kernel size of 1×1. is a max pooling layer with a kernel size of 2×2.

[0126] In the above-mentioned application embodiment, an adaptive module is introduced within the model so that any model of the fragment can adapt to a specific inspection task without retraining or fine-tuning the entire model, effectively reducing training time and memory consumption.

[0127] In one embodiment, an implementation of the above S202 is provided, such as Figure 7 As shown, the above “processing the cable tunnel image and the inverted image by the fusion sub-model in the identification model to determine the dual-phase fusion feature” includes:

[0128] S501, inputting the cable tunnel image and the inverted image into a fusion sub-model; the fusion sub-model includes a convolution module and a fusion module.

[0129] In an embodiment of the present application, the cable tunnel image and the inverted image are used as inputs of the fusion sub-model, so that the convolution module and the fusion module in the fusion sub-model are used to process the input data.

[0130] Optionally, the convolution module includes multiple layers of convolutional layers, multiple layers of activation function layers and attention mechanism layers, the multiple layers of activation function layers are arranged between some convolutional layers in the multiple layers of convolutional layers, and the attention mechanism layer is arranged after the multiple layers of convolutional layers; wherein, the multiple layers of convolutional layers include multiple layers of first convolutional layers, multiple layers of second convolutional layers and multiple layers of third convolutional layers, the first convolutional layer is the input convolutional layer of the jump connection, the second convolutional layer is the convolutional layer not participating in the jump connection, the third convolutional layer is the convolutional layer formed by the jump connection, the third convolutional layer corresponds one-to-one to the first convolutional layer, and the third convolutional layer is the convolutional layer formed by the jump connection of the corresponding first convolutional layer.

[0131] S502: Perform multi-layer convolution processing on the cable tunnel image through a convolution module to obtain a first feature map.

[0132] In the embodiment of the present application, 9 layers of parallel ordered convolution are performed on the cable tunnel image to obtain the first feature map. Figure 8 As shown, the convolutional module consists of 16 3×3 convolutional kernels with a stride of 1. Rectified Linear Unit (ReLU) activation functions are added after layers 2, 3, 5, and 7. A coordinated attention (CA) mechanism is added after the activation function of layer 8 to compress global spatial information into two dimensions and embed positional information into the spatial information. This incorporates the mapping relationship between channels and enhances the network's learning expressiveness. Layers 5, 7, and 9 are skip-connection convolutional layers, representing the 2nd, 3rd, and 1st layers, respectively. They preserve and propagate low-level features such as details and texture in the original input image, allowing them to be directly passed to subsequent classification stages for low-level feature reuse. This increases feature diversity, provides an unobstructed path for gradients, effectively alleviates the vanishing gradient problem, and accelerates training. A Ghost module and Tanh activation function are added after the last layer of the convolutional module. The Ghost module first compresses the input feature map using a 1×1 convolution to extract cross-channel features. After obtaining the compressed features, a 3×3 convolution kernel is used layer by layer to generate an additional feature map. Finally, the 1×1 convolution result is superimposed with the layer-by-layer convolution result to obtain the final feature map, which greatly improves the utilization efficiency of the feature map, reduces the amount of calculation, and achieves richer feature fusion.

[0133] Optionally, the first convolutional layer of the multi-layer is Figure 8 The 2nd, 3rd, and 1st layers in the multi-layer third convolutional layer are Figure 8 The 5th, 7th, and 9th layers in ,correspond one to one to the 2nd, 3rd, and 1st layers respectively, and the ,rest of the convolutional layers are multi-layer second convolutional layers.

[0134] S503: Perform multi-layer convolution processing on the inverted image through a convolution module to obtain a second feature map.

[0135] In an embodiment of the present application, 9 layers of parallel ordered convolution are performed on the inverted input image to obtain a second feature map.

[0136] S504: Fusing the first feature map and the second feature map to obtain a two-phase fusion feature.

[0137] In the embodiment of the present application, the three-channel (R, G, B) curve parameter mappings of the two outputs are merged to obtain the curve parameter mapping of the adaptive light enhancement curve for each iteration. ,in Represents pixel coordinates.

[0138] In the above application embodiment, the first feature map corresponding to the cable tunnel image and the second feature map corresponding to the inverted image are fused to obtain a dual-phase fusion feature, making the data more comprehensive and reliable.

[0139] In one embodiment, Figure 9 As shown in Figure 2, the training process of the above recognition model includes:

[0140] S601, constructing an initial recognition model; the initial recognition model includes an initial fusion sub-model, an initial enhancement sub-model and an initial leaky segmentation labeling sub-model.

[0141] Optionally, the overall network structure in the embodiment of the present application is as follows: Figure 10 As shown in the figure, the fusion sub-model and the enhancement sub-model can be Zero-Reference Deep Curve Estimation (Zero-DDCE), where the fusion sub-model is a double-phase of Deep Curve EstimationNet (DDCE-Net), and the leaky segmentation labeling sub-model is a Spatial Attention Multi-scale Aggregation (SAMA).

[0142] S602, constructing a no-reference loss function and a detection effect evaluation index; the detection effect evaluation index is used to evaluate the detection effect of the overlap and similarity between the true label and the predicted mask.

[0143] In the embodiment of the present application, a reference-free loss function can be designed with the goals of spatial consistency, spatial consistency, color constancy, and illumination smoothness, and a detection effect evaluation index can be constructed based on historical data.

[0144] Optionally, in an embodiment of the present application, the no-reference loss function is determined based on at least one of a spatial consistency loss function, an exposure control loss function, a color consistency loss function, a total variation loss function, and a channel consistency loss function. Wherein:

[0145] (1) Spatial consistency loss function , which can promote the spatial consistency of the enhanced image, as shown in Equation 4:

[0146] (Formula 4)

[0147] Among them, K is the number of local areas, are the four adjacent areas above, below, left and right centered on area i A collection of . and It is the average pixel value of the local area in the enhanced image and the original image. The size of the local area is set to 4×4.

[0148] (2) Exposure control loss function , the exposure level can be controlled as shown in Equation 5:

[0149] (Formula 5)

[0150] Where M is the number of non-overlapping local regions of size 16×16, the average pixel value of the enhanced local region is recorded as Y, and E is the median value of the image brightness, which is taken as 0.6.

[0151] (3) Color consistency loss function , the color deviation of the enhanced image can be reduced, as shown in Equation 6:

[0152] (Equation 6)

[0153] in, represents the pixel average of channel p in the enhanced image, Represents the pixel average of the q channel in the enhanced image, and the variables p and q traverse the two combinations of all possible enhanced color channels , Represents a pair of channels.

[0154] (4) Total variation loss function ,Can , as shown in Equation 7:

[0155] (Equation 7)

[0156] Where C, H, and W represent the number of channels, height, and width of the image, respectively. ∇x and ∇y represent the horizontal and vertical gradient operations, respectively. To enhance the image channel , high value is , width is The pixel value of .

[0157] (5) Channel consistency loss function , the KL divergence can be used to enhance the consistency of the original image and the enhanced image in the channel pixel difference, suppress the generation of noise information and invalid features, and improve the image enhancement effect, as shown in Equation 8:

[0158] (Equation 8)

[0159] Among them, R, G, and B represent the color channels of the original image. 、 、 represents the three color channels of the enhanced image, and the KL divergence represents the difference between the two distributions. If the difference between the two is small, the KL divergence is small. When the two distributions are consistent, the KL divergence value is 0.

[0160] Furthermore, the total loss function It can be shown as formula 9:

[0161] (Equation 9)

[0162] in, 、 、 They are the weights of spatial consistency loss, color consistency loss, total variation loss, and channel consistency loss, respectively, depending on the actual emphasis of the detection effect.

[0163] Optionally, in the embodiment of the present application, by averaging the intersection / combination , average dice coefficient , as a detection performance evaluation metric to measure the overlap and similarity between the predicted and true masks. It can be expressed as formula 10, It can be expressed as formula 11:

[0164] (Equation 10)

[0165] (Equation 11)

[0166] in, represents the number of classes in the segmentation process, and denote the intersection and union of the ground truth mask and the predicted water seepage mask, respectively.

[0167] S603 , obtaining a cable tunnel water leakage dark light dataset, and training an initial recognition model based on the cable tunnel water leakage dark light dataset, a no-reference loss function, and a detection effect evaluation index to obtain a recognition model.

[0168] In the embodiment of the present application, a cable tunnel water leakage dark light dataset is collected, and the cable tunnel water leakage dark light dataset, the no-reference loss function and the detection effect evaluation index are used to analyze the cable tunnel water leakage dark light dataset. Figure 10The Zero-DDCE-SAMA (Zero-Reference Double-phase of Deep Curve Estimation Segment Anything Model Adapetr) initial recognition model shown is trained until the loss of the initial recognition model reaches the preset loss and meets the detection effect evaluation index, so as to realize the recognition of water leakage in the cable tunnel under dim light conditions.

[0169] In the above-mentioned application embodiment, by constructing loss functions in multiple dimensions, the recognition accuracy of the trained recognition model is made higher, and the recognition model is evaluated according to the detection effect evaluation index, thereby further improving the recognition accuracy.

[0170] In one embodiment, a complete method for identifying water seepage in cracks in cable tunnels is provided, comprising:

[0171] S1, build the initial recognition model; the initial recognition model includes the initial fusion sub-model, the initial enhancement sub-model and the initial leaky segmentation labeling sub-model.

[0172] S2, constructs a no-reference loss function and detection effect evaluation index; the detection effect evaluation index is used to evaluate the detection effect of the overlap and similarity between the true label and the predicted mask.

[0173] S3, obtaining a cable tunnel water leakage dark light dataset, and training an initial recognition model based on the cable tunnel water leakage dark light dataset, a no-reference loss function, and a detection effect evaluation index to obtain a recognition model.

[0174] S4, performing inversion processing on the cable tunnel image to obtain an inverted image.

[0175] S5, inputs the cable tunnel image and the inverted image into the fusion sub-model; the fusion sub-model includes a convolution module and a fusion module.

[0176] S6, performing multi-layer convolution processing on the cable tunnel image through a convolution module to obtain a first feature map.

[0177] S7, performing multi-layer convolution processing on the inverted image through a convolution module to obtain a second feature map.

[0178] S8, fusing the first feature map and the second feature map to obtain a two-phase fusion feature.

[0179] S9, based on the dual-phase fusion features and the enhancer model in the recognition model, performs data enhancement on the cable tunnel image to obtain an enhanced image.

[0180] S10, separating pixel data of the enhanced image according to pixel colors to obtain three-channel features corresponding to the enhanced image.

[0181] S11, input the three-channel features into the convolution layer of the adaptive module for feature extraction to obtain single-channel features corresponding to the three-channel features; the single-channel features include red channel data, green channel data and blue channel data.

[0182] S12, performs a maximum pooling operation on the single-channel features to enhance the texture features related to the edge detection of the leaking area.

[0183] S13, superimposing the enhanced texture features and the downsampled enhanced image to obtain first processed data.

[0184] S14, stacking the first processed data, the three-channel features, and the two-phase fusion features through the segment arbitrary module to obtain a tunnel water leakage segmentation labeling map output by the water leakage segmentation labeling sub-model.

[0185] S15, determining a water seepage area in the cable tunnel based on the tunnel water leakage segmentation marker map.

[0186] In the above-mentioned application embodiment, the cable tunnel image is inverted to obtain an inverted image; the cable tunnel image and the inverted image are processed by the fusion sub-model in the recognition model to determine the bi-phase fusion feature; based on the bi-phase fusion feature and the enhancement sub-model in the recognition model, the cable tunnel image is data enhanced to obtain an enhanced image; the enhanced image and the bi-phase fusion feature are input into the water leakage segmentation labeling sub-model in the recognition model to obtain the tunnel water leakage segmentation labeling map output by the water leakage segmentation labeling sub-model; the water leakage area in the cable tunnel is determined based on the tunnel water leakage segmentation labeling map. The inverted image corresponding to the cable tunnel image is obtained, and the bi-phase fusion feature of the cable tunnel image and the inverted image is used to perform data enhancement on the cable tunnel image, thereby improving the accuracy of the enhanced data recognition. Moreover, the enhanced data and the bi-phase fusion feature are used as the input of the water leakage segmentation labeling sub-model, and the input data is more reliable and accurate, making the input of the water leakage segmentation labeling sub-model more reliable and accurate, thereby improving the accuracy of the recognition of water leakage in the cable tunnel.

[0187] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0188] Based on the same inventive concept, embodiments of the present application also provide a device for identifying water seepage in cable tunnel cracks, for implementing the aforementioned method for identifying water seepage in cable tunnel cracks. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the device for identifying water seepage in cable tunnel cracks provided below can be found in the aforementioned limitations of the method for identifying water seepage in cable tunnel cracks, and will not be further elaborated here.

[0189] In one embodiment, Figure 11 As shown, a device for identifying water seepage in cracks in cable tunnels is provided, comprising: an inversion module 10, a fusion module 11, an enhancement module 12, an acquisition module 13 and a determination module 14, wherein:

[0190] An inversion module 10 is used to perform inversion processing on the cable tunnel image to obtain an inverted image;

[0191] A fusion module 11 is used to process the cable tunnel image and the inverted phase image through a fusion sub-model in the recognition model to determine a dual-phase fusion feature;

[0192] An enhancement module 12 is configured to perform data enhancement on the cable tunnel image based on the dual-phase fusion feature and the enhancement sub-model in the recognition model to obtain an enhanced image;

[0193] An acquisition module 13 is configured to input the enhanced image and the dual-phase fusion features into a water leakage segmentation and marking sub-model in the recognition model to obtain a tunnel water leakage segmentation and marking map output by the water leakage segmentation and marking sub-model;

[0194] The determination module 14 is configured to determine the water seepage area in the cable tunnel based on the tunnel water leakage segmentation marker map.

[0195] In one embodiment, the acquisition module 13 includes: a separation unit, a processing unit, and a stacking unit, wherein:

[0196] The separation unit is used to separate the pixel data of the enhanced image according to the pixel color to obtain the three-channel features corresponding to the enhanced image.

[0197] The processing unit is used to input the three-channel features into the adaptive module for processing to obtain first processed data.

[0198] The stacking unit is used to stack the first processed data, the three-channel features and the two-phase fusion features through the fragment arbitrary module to obtain the tunnel water leakage segmentation mark map output by the water leakage segmentation mark sub-model.

[0199] In one embodiment, the above-mentioned processing unit is specifically used to input the three-channel features into the convolution layer of the adaptive module for feature extraction to obtain single-channel features corresponding to the three-channel features; the single-channel features include red channel data, green channel data and blue channel data; the single-channel features are subjected to maximum pooling operation to enhance texture features related to edge detection of the leaking area; the enhanced texture features and the downsampled enhanced image are superimposed to obtain first processed data.

[0200] In one embodiment, the fusion module 11 includes: an input unit, a first convolution unit, a second convolution unit, and a fusion unit, wherein:

[0201] The input unit is used to input the cable tunnel image and the inverted image into the fusion sub-model; the fusion sub-model includes a convolution module and a fusion module.

[0202] The first convolution unit is configured to perform multi-layer convolution processing on the cable tunnel image through a convolution module to obtain a first feature map.

[0203] a second convolution unit, configured to perform multi-layer convolution processing on the inverted image through a convolution module to obtain a second feature map;

[0204] The fusion unit is used to fuse the first feature map and the second feature map to obtain a two-phase fusion feature.

[0205] In one embodiment, the above-mentioned convolution module includes multiple layers of convolutional layers, multiple layers of activation function layers and attention mechanism layers. The multiple layers of activation function layers are arranged between some convolutional layers in the multiple layers of convolutional layers, and the attention mechanism layers are arranged after the multiple layers of convolutional layers; wherein, the multiple layers of convolutional layers include multiple layers of first convolutional layers, multiple layers of second convolutional layers and multiple layers of third convolutional layers, the third convolutional layer corresponds one-to-one to the first convolutional layer, and the third convolutional layer is a skip convolution layer of the corresponding first convolutional layer.

[0206] In one embodiment, the cable tunnel crack water seepage identification device further includes: a first construction module, a second construction module and a training module, wherein:

[0207] The first construction module is used to construct an initial recognition model; the initial recognition model includes an initial fusion sub-model, an initial enhancement sub-model and an initial leaky segmentation marking sub-model.

[0208] The second building block is used to construct a no-reference loss function and a detection effect evaluation index; the detection effect evaluation index is used to evaluate the detection effect of the overlap and similarity between the true label and the predicted mask.

[0209] The training module is used to obtain a cable tunnel water leakage dark light dataset, and train an initial recognition model based on the cable tunnel water leakage dark light dataset, a no-reference loss function, and a detection effect evaluation index to obtain a recognition model.

[0210] In one embodiment, the above-mentioned no-reference loss function is determined according to at least one of a spatial consistency loss function, an exposure control loss function, a color consistency loss function, a total variation loss function, and a channel consistency loss function.

[0211] Each module in the aforementioned cable tunnel crack seepage identification device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in hardware form, or can be stored in a computer device memory in software form, so that the processor can call and execute the corresponding operations of each module.

[0212] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0213] Inverting the cable tunnel image to obtain an inverted image;

[0214] The cable tunnel image and the inverted phase image are processed by the fusion sub-model in the recognition model to determine the dual-phase fusion features;

[0215] Based on the dual-phase fusion features and the enhancer model in the recognition model, the cable tunnel image is enhanced to obtain an enhanced image.

[0216] The enhanced image and the dual-phase fusion features are input into the water leakage segmentation and marking sub-model in the recognition model to obtain the tunnel water leakage segmentation and marking map output by the water leakage segmentation and marking sub-model;

[0217] Determine water seepage areas in cable tunnels based on tunnel water leakage segmentation marker maps.

[0218] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0219] Separate the pixel data of the enhanced image according to pixel color to obtain the three-channel features corresponding to the enhanced image;

[0220] Inputting the three-channel features into the adaptive module for processing to obtain first processed data;

[0221] The first processed data, three-channel features and two-phase fusion features are stacked through the fragment arbitrary module to obtain the tunnel water leakage segmentation mark map output by the water leakage segmentation mark sub-model.

[0222] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0223] The three-channel features are input into the convolution layer of the adaptive module for feature extraction to obtain the single-channel features corresponding to the three-channel features; the single-channel features include red channel data, green channel data, and blue channel data;

[0224] Perform a maximum pooling operation on the single-channel features to enhance the texture features related to edge detection of leaking areas;

[0225] The enhanced texture features and the downsampled enhanced image are superimposed to obtain first processed data.

[0226] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0227] Input the cable tunnel image and the inverted image into the fusion sub-model; the fusion sub-model includes a convolution module and a fusion module;

[0228] Performing multi-layer convolution processing on the cable tunnel image through a convolution module to obtain a first feature map;

[0229] Perform multi-layer convolution processing on the inverted image through the convolution module to obtain a second feature map;

[0230] The first feature map and the second feature map are fused to obtain a two-phase fusion feature.

[0231] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0232] The convolution module includes multiple convolution layers, multiple activation function layers and attention mechanism layers. The multiple activation function layers are set between some convolution layers in the multiple convolution layers, and the attention mechanism layer is set after the convolution layers.

[0233] Among them, the multi-layer convolution layer includes a multi-layer first convolution layer, a multi-layer second convolution layer and a multi-layer third convolution layer. The third convolution layer corresponds to the first convolution layer one by one, and the third convolution layer is a skip convolution layer of the corresponding first convolution layer.

[0234] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0235] Construct an initial recognition model; the initial recognition model includes an initial fusion sub-model, an initial enhancement sub-model, and an initial leaky segmentation labeling sub-model;

[0236] Construct a no-reference loss function and a detection effect evaluation index; the detection effect evaluation index is used to evaluate the detection effect of the overlap and similarity between the true label and the predicted mask;

[0237] A cable tunnel water leakage dark light dataset is obtained, and an initial recognition model is trained based on the cable tunnel water leakage dark light dataset, a no-reference loss function, and a detection effect evaluation index to obtain a recognition model.

[0238] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0239] The no-reference loss function is determined based on at least one of a spatial consistency loss function, an exposure control loss function, a color consistency loss function, a total variation loss function, and a channel consistency loss function.

[0240] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0241] Inverting the cable tunnel image to obtain an inverted image;

[0242] The cable tunnel image and the inverted phase image are processed by the fusion sub-model in the recognition model to determine the dual-phase fusion features;

[0243] Based on the dual-phase fusion features and the enhancer model in the recognition model, the cable tunnel image is enhanced to obtain an enhanced image.

[0244] The enhanced image and the dual-phase fusion features are input into the water leakage segmentation and marking sub-model in the recognition model to obtain the tunnel water leakage segmentation and marking map output by the water leakage segmentation and marking sub-model;

[0245] Determine water seepage areas in cable tunnels based on tunnel water leakage segmentation marker maps.

[0246] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0247] The water leakage segmentation and marking sub-model includes an adaptive module and a segment arbitrary module. The enhanced image and the two-phase fusion feature are input into the water leakage segmentation and marking sub-model in the recognition model to obtain the tunnel water leakage segmentation and marking map output by the water leakage segmentation and marking sub-model, including:

[0248] Separate the pixel data of the enhanced image according to pixel color to obtain the three-channel features corresponding to the enhanced image;

[0249] Inputting the three-channel features into the adaptive module for processing to obtain first processed data;

[0250] The first processed data, three-channel features and two-phase fusion features are stacked through the fragment arbitrary module to obtain the tunnel water leakage segmentation mark map output by the water leakage segmentation mark sub-model.

[0251] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0252] The three-channel features are input into the convolution layer of the adaptive module for feature extraction to obtain the single-channel features corresponding to the three-channel features; the single-channel features include red channel data, green channel data, and blue channel data;

[0253] Perform a maximum pooling operation on the single-channel features to enhance the texture features related to edge detection of leaking areas;

[0254] The enhanced texture features and the downsampled enhanced image are superimposed to obtain first processed data.

[0255] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0256] Input the cable tunnel image and the inverted image into the fusion sub-model; the fusion sub-model includes a convolution module and a fusion module;

[0257] Performing multi-layer convolution processing on the cable tunnel image through a convolution module to obtain a first feature map;

[0258] Perform multi-layer convolution processing on the inverted image through the convolution module to obtain a second feature map;

[0259] The first feature map and the second feature map are fused to obtain a two-phase fusion feature.

[0260] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0261] The convolution module includes multiple convolution layers, multiple activation function layers and attention mechanism layers. The multiple activation function layers are set between some convolution layers in the multiple convolution layers, and the attention mechanism layer is set after the convolution layers.

[0262] Among them, the multi-layer convolution layer includes a multi-layer first convolution layer, a multi-layer second convolution layer and a multi-layer third convolution layer. The third convolution layer corresponds to the first convolution layer one by one, and the third convolution layer is a skip convolution layer of the corresponding first convolution layer.

[0263] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0264] Construct an initial recognition model; the initial recognition model includes an initial fusion sub-model, an initial enhancement sub-model, and an initial leaky segmentation labeling sub-model;

[0265] Construct a no-reference loss function and a detection effect evaluation index; the detection effect evaluation index is used to evaluate the detection effect of the overlap and similarity between the true label and the predicted mask;

[0266] A cable tunnel water leakage dark light dataset is obtained, and an initial recognition model is trained based on the cable tunnel water leakage dark light dataset, a no-reference loss function, and a detection effect evaluation index to obtain a recognition model.

[0267] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0268] The no-reference loss function is determined based on at least one of a spatial consistency loss function, an exposure control loss function, a color consistency loss function, a total variation loss function, and a channel consistency loss function.

[0269] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0270] Inverting the cable tunnel image to obtain an inverted image;

[0271] The cable tunnel image and the inverted phase image are processed by the fusion sub-model in the recognition model to determine the dual-phase fusion features;

[0272] Based on the dual-phase fusion features and the enhancer model in the recognition model, the cable tunnel image is enhanced to obtain an enhanced image.

[0273] The enhanced image and the dual-phase fusion features are input into the water leakage segmentation and marking sub-model in the recognition model to obtain the tunnel water leakage segmentation and marking map output by the water leakage segmentation and marking sub-model;

[0274] Determine water seepage areas in cable tunnels based on tunnel water leakage segmentation marker maps.

[0275] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0276] Separate the pixel data of the enhanced image according to pixel color to obtain the three-channel features corresponding to the enhanced image;

[0277] Inputting the three-channel features into the adaptive module for processing to obtain first processed data;

[0278] The first processed data, three-channel features and two-phase fusion features are stacked through the fragment arbitrary module to obtain the tunnel water leakage segmentation mark map output by the water leakage segmentation mark sub-model.

[0279] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0280] The three-channel features are input into the convolution layer of the adaptive module for feature extraction to obtain the single-channel features corresponding to the three-channel features; the single-channel features include red channel data, green channel data, and blue channel data;

[0281] Perform a maximum pooling operation on the single-channel features to enhance the texture features related to edge detection of leaking areas;

[0282] The enhanced texture features and the downsampled enhanced image are superimposed to obtain first processed data.

[0283] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0284] Input the cable tunnel image and the inverted image into the fusion sub-model; the fusion sub-model includes a convolution module and a fusion module;

[0285] Performing multi-layer convolution processing on the cable tunnel image through a convolution module to obtain a first feature map;

[0286] Perform multi-layer convolution processing on the inverted image through the convolution module to obtain a second feature map;

[0287] The first feature map and the second feature map are fused to obtain a two-phase fusion feature.

[0288] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0289] The convolution module includes multiple convolution layers, multiple activation function layers and attention mechanism layers. The multiple activation function layers are set between some convolution layers in the multiple convolution layers, and the attention mechanism layer is set after the convolution layers.

[0290] Among them, the multi-layer convolution layer includes a multi-layer first convolution layer, a multi-layer second convolution layer and a multi-layer third convolution layer. The third convolution layer corresponds to the first convolution layer one by one, and the third convolution layer is a skip convolution layer of the corresponding first convolution layer.

[0291] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0292] Construct an initial recognition model; the initial recognition model includes an initial fusion sub-model, an initial enhancement sub-model, and an initial leaky segmentation labeling sub-model;

[0293] Construct a no-reference loss function and a detection effect evaluation index; the detection effect evaluation index is used to evaluate the detection effect of the overlap and similarity between the true label and the predicted mask;

[0294] A cable tunnel water leakage dark light dataset is obtained, and an initial recognition model is trained based on the cable tunnel water leakage dark light dataset, a no-reference loss function, and a detection effect evaluation index to obtain a recognition model.

[0295] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0296] The no-reference loss function is determined based on at least one of a spatial consistency loss function, an exposure control loss function, a color consistency loss function, a total variation loss function, and a channel consistency loss function.

[0297] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, 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 various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), data processing logic devices based on quantum computing, and the like.

[0298] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, 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.

[0299] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for identifying water seepage in cable tunnel cracks, characterized in that: The method comprises: Inverting the cable tunnel image to obtain an inverted image; Processing the cable tunnel image and the inverted image by using a fusion sub-model in the recognition model to determine a dual-phase fusion feature; Based on the two-phase fusion feature and the enhancer model in the recognition model, data enhancement is performed on the cable tunnel image to obtain an enhanced image; Inputting the enhanced image and the dual-phase fusion feature into the water leakage segmentation mark sub-model in the recognition model to obtain a tunnel water leakage segmentation mark map output by the water leakage segmentation mark sub-model; Determine water seepage areas in cable tunnels based on tunnel water leakage segmentation marker maps.

2. The method according to claim 1, characterized in that The water leakage segmentation and marking sub-model includes an adaptive module and a segment arbitrary module. The step of inputting the enhanced image and the two-phase fusion feature into the water leakage segmentation and marking sub-model in the recognition model to obtain a tunnel water leakage segmentation and marking map output by the water leakage segmentation and marking sub-model includes: Separating pixel data of the enhanced image according to pixel colors to obtain three-channel features corresponding to the enhanced image; Inputting the three-channel features into an adaptive module for processing to obtain first processed data; The first processed data, the three-channel features and the two-phase fusion features are stacked by the segment arbitrary module to obtain a tunnel water leakage segmentation mark map output by the water leakage segmentation mark sub-model.

3. The method according to claim 2, characterized in that Inputting the three-channel features into the adaptive module for processing to obtain first processed data includes: Inputting the three-channel features into the convolution layer of the adaptive module for feature extraction to obtain single-channel features corresponding to the three-channel features; the single-channel features include red channel data, green channel data, and blue channel data; performing a maximum pooling operation on the single-channel features to enhance texture features related to edge detection of leaking areas; The enhanced texture features and the downsampled enhanced image are superimposed to obtain the first processed data.

4. The method according to claim 1, wherein The processing of the cable tunnel image and the inverted image by a fusion sub-model in the recognition model to determine a dual-phase fusion feature includes: Inputting the cable tunnel image and the inverted image into the fusion sub-model; the fusion sub-model includes a convolution module and a fusion module; Performing multi-layer convolution processing on the cable tunnel image by the convolution module to obtain a first feature map; Performing multi-layer convolution processing on the inverted image by the convolution module to obtain a second feature map; The first feature map and the second feature map are fused to obtain a two-phase fusion feature.

5. The method according to claim 4, characterized in that The convolution module includes multiple convolution layers, multiple activation function layers and an attention mechanism layer, wherein the multiple activation function layers are arranged between some convolution layers in the multiple convolution layers, and the attention mechanism layer is arranged after the convolution layers; The multi-layer convolutional layer includes a multi-layer first convolutional layer, a multi-layer second convolutional layer and a multi-layer third convolutional layer. The third convolutional layer corresponds to the first convolutional layer one-to-one, and the third convolutional layer is a skip convolutional layer of the corresponding first convolutional layer.

6. The method according to any one of claims 1 to 5, characterized in that The training process of the recognition model includes: Constructing an initial recognition model; the initial recognition model includes an initial fusion sub-model, an initial enhancement sub-model and an initial leaky segmentation marking sub-model; Constructing a no-reference loss function and a detection effect evaluation index; the detection effect evaluation index is used to evaluate the detection effect of the overlap and similarity between the true label and the predicted mask; A cable tunnel water leakage dark light dataset is obtained, and the initial recognition model is trained according to the cable tunnel water leakage dark light dataset, a no-reference loss function, and the detection effect evaluation index to obtain the recognition model.

7. The method according to claim 6, characterized in that The no-reference loss function is determined based on at least one of a spatial consistency loss function, an exposure control loss function, a color consistency loss function, a total variation loss function, and a channel consistency loss function.

8. A device for identifying water seepage in cable tunnel cracks, characterized in that: The device comprises: An inversion module, used for performing inversion processing on the cable tunnel image to obtain an inverted image; a fusion module, configured to process the cable tunnel image and the inverted-phase image by using a fusion sub-model in a recognition model to determine a dual-phase fusion feature; an enhancement module, configured to perform data enhancement on the cable tunnel image based on the dual-phase fusion feature and an enhancement sub-model in the recognition model to obtain an enhanced image; an acquisition module, configured to input the enhanced image and the dual-phase fusion feature into a water leakage segmentation mark sub-model in the recognition model, so as to obtain a tunnel water leakage segmentation mark map output by the water leakage segmentation mark sub-model; The determination module is used to determine the water seepage area in the cable tunnel based on the tunnel water leakage segmentation mark map.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images