Tunnel leakage identification method, device and system and medium

Abstract

The invention provides a tunnel leakage identification method, device and system and a medium. The method comprises the following steps: obtaining a tunnel thermodynamic diagram; inputting the tunnel thermodynamic diagram into a pre-trained deep learning fusion network, screening out the tunnel thermodynamic diagram containing a leakage point region, and segmenting the leakage point region in the tunnel thermodynamic diagram; calculating the area of a leakage point region in the segmented tunnel thermodynamic diagram; the invention provides an innovative scheme based on a deep learning fusion network and infrared thermal imaging, efficient and accurate detection of a tunnel leakage target and semantic segmentation of a leakage area are realized through combination with an advanced image processing technology, hidden leakage can be automatically identified, automatic extraction of leakage information is further realized, and the detection efficiency is improved. The intelligent and automatic level of leakage detection is greatly improved, and a more comprehensive and efficient solution is provided for tunnel detection.

Description

Technical Field

[0001] The invention relates to the field of engineering detection methods, and in particular to a method for quickly identifying tunnel leakage suitable for installing a fireproof layer. Background Art

[0002] Underwater tunnels are important urban transportation hubs. Due to factors such as geology, construction defects, and the environment, cracks of varying degrees will inevitably occur during the operation period, and water leakage will occur, affecting the structural strength of the lining concrete and pipe segments and the normal operation of the tunnel. At present, fireproof layers are generally installed inside underwater tunnels, so that when leakage occurs in the tunnel, the leakage cannot be directly observed or identified by the naked eye or visible light. This design limits the early discovery of leakage diseases, delays the implementation of preventive measures and the opportunity for timely repair, and increases the difficulty and cost of later repairs. Therefore, water leakage detection in tunnels with fireproof layers has become a technical problem that needs to be solved urgently.

[0003] The traditional method of detecting water leakage in tunnels is manual visual inspection, that is, inspectors patrol to find water leakage points. This method is simple to operate, but has shortcomings such as low efficiency and great impact on traffic. Installing hygrometers and thermometers (such as distributed fiber optic sensors) can accurately monitor water leakage in certain locations or coverage areas, and the data can be transmitted remotely, but the locations outside the coverage area of ​​the sensor cannot be sensed, there are many blind spots, and the subsequent system maintenance cost is high. The use of non-destructive testing technology (3D laser scanning, infrared thermal imaging, etc.) can be used to perform comprehensive and rapid scanning of tunnels, which is the development trend of rapid detection of water leakage. 3D laser scanning identifies water leakage by analyzing the cloud point image obtained by scanning, but it has the disadvantages of complex scanning equipment system, high cost, and time-consuming cloud point image processing.

[0004] As one of the non-destructive testing technologies, infrared thermal imaging technology has obvious advantages in water seepage defect detection. The infrared thermal imaging detection method does not require a radiation source and can completely rely on the thermal radiation of the water seepage area itself to form an image; therefore, it can penetrate the fireproof layer and directly detect whether there is invisible leakage inside. However, most of the current infrared technology methods require manual auxiliary identification, and the level of intelligence and automation is insufficient. To solve this problem, the present invention proposes an innovative solution based on a deep learning fusion network and infrared thermal imaging. By combining advanced image processing technology, it realizes efficient and accurate detection of tunnel leakage targets and semantic segmentation of leakage areas. It can automatically identify hidden leakage and further realize the automatic extraction of leakage information, which greatly improves the intelligence and automation level of leakage detection and provides a more comprehensive and efficient solution for tunnel detection. Summary of the invention

[0005] The purpose of the present invention is to provide a tunnel leakage identification method, device, system and medium, which, by combining image processing technology, realizes efficient and accurate detection of tunnel leakage targets and semantic segmentation of leakage areas, and can automatically identify hidden leakage.

[0006] In order to achieve the above objectives / solve the above technical problems, the present invention is implemented by adopting the following technical solutions.

[0007] In one aspect, the present invention provides a method for identifying tunnel leakage, comprising the following steps:

[0008] Get tunnel thermal map;

[0009] Input the tunnel thermal map into the pre-trained deep learning fusion network, filter out the tunnel thermal map containing the leakage point area, and segment the leakage point area in the tunnel thermal map;

[0010] Calculate the area of ​​the leakage point region in the segmented tunnel thermal map;

[0011] Among them: The area calculation method of the leakage point area includes:

[0012] The shooting field of view is calculated according to the shooting angle of the segmented tunnel heat map, and the radial distortion of the lens of the shooting field of view is corrected. The pixel accuracy is calculated according to the corrected shooting field of view.

[0013] Perform image optimization processing on the segmented tunnel thermal map to obtain the number of effective pixels;

[0014] The area of ​​the leakage point region is calculated based on the pixel accuracy and the number of effective pixels.

[0015] Furthermore, the step of screening out the tunnel thermal map containing the leakage point area and segmenting the leakage point area in the tunnel thermal map includes:

[0016] The tunnel heat map is passed through several convolutional layers and pooling layers to extract the image features, and the image features are gradually downsampled and upsampled to increase the abstraction level of the image features. The probability map of each pixel in the tunnel heat map belonging to the leakage point category is generated through the activation function, and the tunnel heat map containing the leakage point area is screened out according to the probability map;

[0017] The screened tunnel heat map is compressed through encoding operation to extract multi-scale features, and then gradually restored through deconvolution or upsampling in decoding operation. At the same time, the features of the corresponding layer of the encoding operation are fused, and the detail information is transferred through jump connection. Finally, the decoding operation outputs the tunnel heat map leakage accurate segmentation image containing the leakage point area with the original resolution.

[0018] Furthermore, the calculating of the shooting field of view according to the shooting angle of the segmented tunnel heat map includes:

[0019] ,

[0020] ,

[0021] Where: θ h is the horizontal shooting angle of the tunnel thermal map, θ v W is the vertical shooting angle of the tunnel thermal map; sensor H is the width of the image sensor. sensor is the height of the shooting sensor; k 0 is the geometric difference adjustment coefficient between the shooting sensor and the lens, f is the focal length of the lens; d o is the object distance, d f is the actual working distance;

[0022] ,

[0023] ,

[0024] Where: W fov is the field width of the tunnel thermal map, H fov is the field of view height of the tunnel thermal image; MTF(f) is the modulation transfer function of the lens and imaging system at the spatial frequency f.

[0025] Furthermore, the radial distortion expression of the lens for correcting the shooting field of view is:

[0026] ,

[0027] Where: W' fov is the field of view width after considering the radial distortion correction of the lens, H' fov′ is the height of the field of view after considering the correction of the radial distortion of the lens, k 1 , k 2 , k 3 It is the coefficient of radial distortion of the set lens.

[0028] Furthermore, the pixel accuracy of the corrected shooting field of view is calculated according to the corrected shooting field of view, including:

[0029] ,

[0030] ,

[0031] ,

[0032] Where: P is the pixel accuracy of the tunnel thermal map, P x is the horizontal pixel accuracy of the tunnel thermal map, P y It is the vertical pixel accuracy of the tunnel thermal map.

[0033] Furthermore, the image optimization processing is performed on the segmented tunnel thermal map to obtain the number of effective pixels thereof, specifically comprising:

[0034] The image optimization process includes: performing mean filtering, grayscale conversion and binarization operations on the segmented tunnel thermal map in sequence:

[0035] Among them: The expression of the mean filter operation is:

[0036] ,

[0037] ,

[0038] ,

[0039] Where: I filtered (x,y) is the pixel value of the filtered tunnel heat map at position (x,y); m and n are the width and height of the window, I(x+i, y+j) represents the pixel value of the tunnel heat map at the x+ith row and y+jth column; a and b are the set filter kernel size parameters; i and j are the set filter kernel indexes;

[0040] The grayscale expression is:

[0041] ,

[0042] Where: g(x, y) is the gray value at the coordinate (x, y) in the tunnel thermal map, R is the value of the red channel of the tunnel thermal map, G is the value of the green channel of the tunnel thermal map, and B is the value of the black channel of the tunnel thermal map;

[0043] The binary expression is:

[0044] ,

[0045] Where: B(x, y) is the pixel value at coordinate (x, y) in the binarized image, g(x, y) is the pixel value in the original grayscale image, T is the set threshold, T=0;

[0046] The expression for obtaining the number of effective pixels is:

[0047] ,

[0048] ,

[0049] in: is the pixel determination function, N is the number of valid pixels, a valid pixel is a pixel with a pixel value of 255, H is the height of the tunnel thermal map, W is the width of the tunnel thermal map, and I(i,j) represents the pixel value of the tunnel thermal map in the i-th row and j-th column.

[0050] Furthermore, the area of ​​the leakage point region is calculated according to the pixel accuracy and the number of effective pixels, including:

[0051] ,

[0052] Where: S is the area of ​​the leak point.

[0053] In a second aspect, the present invention provides a tunnel leakage identification device suitable for installing a fireproof layer, comprising:

[0054] Acquisition module: used to obtain tunnel thermal map;

[0055] Segmentation module: used to input the tunnel thermal map into the pre-trained deep learning fusion network, filter out the tunnel thermal map containing the leakage point area, and segment the leakage point area in the tunnel thermal map;

[0056] Identification module: used to calculate the area of ​​the leakage point region in the segmented tunnel thermal map;

[0057] Among them: The area calculation method of the leakage point area includes:

[0058] The shooting field of view is calculated according to the shooting angle of the segmented tunnel heat map, and the radial distortion of the lens of the shooting field of view is corrected. The pixel accuracy is calculated according to the corrected shooting field of view.

[0059] Perform image optimization processing on the segmented tunnel thermal map to obtain the number of effective pixels;

[0060] The area of ​​the leakage point region is calculated based on the pixel accuracy and the number of effective pixels.

[0061] In a third aspect, the present invention provides a tunnel leakage identification system, comprising:

[0062] Memory, for storing computer programs / instructions;

[0063] A processor is used to execute the computer program / instructions to implement the steps of the above-mentioned tunnel leakage identification method.

[0064] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program / instruction stored thereon, which, when executed by a processor, implements the steps of the above-mentioned tunnel leakage identification method.

[0065] Compared with the prior art, the present invention has the following beneficial effects: the present invention proposes an innovative solution based on deep learning fusion network and infrared thermal imaging. By combining advanced image processing technology, the present invention realizes efficient and accurate detection of tunnel leakage targets and semantic segmentation of leakage areas, and can automatically identify hidden leakage, further realize automatic extraction of leakage information, greatly improve the intelligence and automation level of leakage detection, and provide a more comprehensive and efficient solution for tunnel detection;

[0066] The present invention studies a method for rapid identification of tunnel leakage suitable for installing a fireproof layer. By combining infrared imaging with deep learning and image processing, the efficiency of tunnel water leakage detection is greatly improved, non-contact automatic rapid detection is realized, interference with tunnel operations is reduced, and safety and feasibility are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a flow chart for rapid identification of tunnel leakage provided by an example of the present invention;

[0068] Figure 2 It is a schematic diagram of the experimental site arrangement provided in Example 1 of the present invention;

[0069] Figure 3 is a schematic diagram of leakage point detection provided by Example 1 of the present invention;

[0070] Figure 4 It is a schematic diagram of the experimental site arrangement provided by Example 2 of the present invention;

[0071] Figure 5 This is a schematic diagram of leakage point detection provided by Example 2 of the present invention. DETAILED DESCRIPTION

[0072] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. The embodiments of the present invention and the technical features in the embodiments may be combined with each other unless there is a conflict.

[0073] The term "and / or" is only a way to describe the association relationship of related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " generally indicates that the related objects are in an "or" relationship.

[0074] Example 1

[0075] like Figure 1 An embodiment shown in the figure provides a method for identifying tunnel leakage, comprising the following steps:

[0076] Step 1: Obtain tunnel thermal map using thermal infrared imager;

[0077] Step 2: Input the tunnel heat map into the pre-trained deep learning fusion network, filter out the tunnel heat map containing the leakage point area, and segment the leakage point area in the tunnel heat map. The deep learning fusion network combines the YOLOv7 and Unet networks. The YOLOv7 network training uses random gradient descent as the optimizer type, the optimizer momentum parameter is 0.937, the weight decay value is 5e-4, the model maximum learning rate is 1e-2, and the minimum learning rate is 1e-4 to ensure that the model can converge quickly. The Unet network training uses adaptive moment estimation as the optimizer type, the optimizer internal momentum parameter is 0.9, the weight decay value is 0, the model maximum learning rate is set to 1e-4, and the minimum learning rate is 1e-6.

[0078] Step 3: Calculate the area of ​​the leak point area;

[0079] Wherein step 3: the area calculation method of the leakage point area includes:

[0080] Calculate the viewing angle of the thermal infrared imager by combining the object distance of the leak area, the size of the thermal infrared imager, and the actual working distance;

[0081] The field of view is calculated by combining the viewing angle with the object distance in the leakage point area, and the radial distortion correction of the lens is performed on the field of view of the thermal infrared imager.

[0082] The pixel accuracy of the tunnel thermal image containing the leakage point area is calculated through the corrected field of view;

[0083] For the image segmented by the deep learning fusion network, filtering, grayscale and binarization processing are performed, and the number of pixels with a pixel value of 255 in the image is counted;

[0084] The area of ​​the leakage point region is calculated by the number of pixels with a pixel value of 255 in the image and the pixel accuracy.

[0085] The method of screening out the tunnel heat map containing the leakage point area and segmenting the leakage point area in the tunnel heat map includes: firstly adjusting the resolution of the tunnel heat map after inputting into the deep learning fusion network to 640×640, normalizing the pixel value of the scale-adjusted image from the range of [0, 255] to the range of [0, 1] to obtain a preprocessed image, extracting the image features of the preprocessed image through multiple convolution layers and pooling layers, and then gradually downsampling and upsampling to increase the abstraction level of the image features, and finally generating a probability map of each pixel in the tunnel heat map belonging to the leakage point category through an activation function, compressing the spatial resolution of the probability map through an encoding operation and extracting multi-scale features, and then gradually restoring the spatial resolution through deconvolution or upsampling through a decoding operation, while fusing the features of the corresponding layer of the encoding operation, transmitting detail information through a jump connection, and finally outputting a tunnel heat map leakage accurate segmentation image with the original resolution through a decoding operation.

[0086] The calculation of the viewing angle by introducing the object distance of the leak point area, the size of the thermal infrared imager, and the actual working distance includes:

[0087] ,

[0088] ,

[0089] Where: θ h is the horizontal shooting angle of the tunnel thermal map, θ v W is the vertical shooting angle of the tunnel thermal map; sensor H is the width of the image sensor. sensor is the height of the shooting sensor; k 0 is the geometric difference adjustment coefficient between the shooting sensor and the lens, f is the focal length of the lens; d o is the object distance, d f is the actual working distance;

[0090] ,

[0091] ,

[0092] Where: W fov is the field width of the tunnel thermal map, H fov is the field of view height of the tunnel thermal image; MTF(f) is the modulation transfer function of the lens and imaging system at the spatial frequency f.

[0093] The radial distortion expression of the lens for correcting the shooting field of view is:

[0094] ,

[0095] Where: W' fovis the field of view width after considering the radial distortion correction of the lens, H' fov′ is the height of the field of view after considering the correction of the radial distortion of the lens, k 1 , k 2 , k 3 is the coefficient of radial distortion of the lens.

[0096] The step of calculating the pixel accuracy of the corrected shooting field of view includes:

[0097] ,

[0098] ,

[0099] ,

[0100] Where: P is the pixel accuracy of the tunnel thermal map, P x is the horizontal pixel accuracy of the tunnel thermal map, P y It is the vertical pixel accuracy of the tunnel thermal map.

[0101] The image optimization processing of the segmented tunnel thermal map to obtain the number of effective pixels specifically includes:

[0102] The image optimization process includes: performing mean filtering, grayscale conversion and binarization operations on the segmented tunnel thermal map in sequence:

[0103] Among them: The expression of the mean filter operation is:

[0104] ,

[0105] ,

[0106] ,

[0107] Where: I filtered (x,y) is the pixel value of the filtered tunnel heat map at position (x,y); m and n are the width and height of the window, I(x+i, y+j) represents the pixel value of the tunnel heat map at the x+ith row and y+jth column; a and b are the set filter kernel size parameters; i and j are the set filter kernel indexes;

[0108] The grayscale expression is:

[0109] ,

[0110] Where: g(x, y) is the gray value at the coordinate (x, y) in the tunnel thermal map, R is the value of the red channel of the tunnel thermal map, G is the value of the green channel of the tunnel thermal map, and B is the value of the black channel of the tunnel thermal map;

[0111] The binary expression is:

[0112] ,

[0113] Where: B(x, y) is the pixel value at coordinate (x, y) in the binarized image, g(x, y) is the pixel value in the original grayscale image, T is the set threshold, T=0;

[0114] The expression for obtaining the number of effective pixels is:

[0115] ,

[0116] ,

[0117] in: is the pixel determination function, N is the number of valid pixels, a valid pixel is a pixel with a pixel value of 255, H is the height of the tunnel thermal map, W is the width of the tunnel thermal map, and I(i,j) represents the pixel value of the tunnel thermal map in the i-th row and j-th column.

[0118] Calculating the area of ​​the leakage point region according to the pixel accuracy and the number of effective pixels includes:

[0119] ,

[0120] Where: S is the area of ​​the leak point.

[0121] Specific: such as Figure 2~3 As shown, this embodiment is carried out indoors, and calcium silicate and glass magnesium fireproof boards commonly used in underwater tunnels are used to simulate the actual tunnel environment, and water-absorbing non-woven fabrics are used to simulate leakage. Water does not seep out from the front of the fireproof board, but there is a small leakage on the back, which simulates the leakage points in the board and the leakage points in the board seams respectively; an infrared thermal imager is fixed 3m away from the test board, and a deep learning fusion network is trained in advance to observe the recognition of hidden leakage and the calculation of the area.

[0122] The method of the present invention has a good effect on leakage point identification and image segmentation. An infrared thermal imager with a focal length of 8 mm, a resolution of 640*480, and a sensor size of 1 / 2.7 inches, i.e., 5.37 mm (width) × 4.04 mm (height), is used, i.e., the object distance is 3000 mm;

[0123] The tunnel thermal map obtained by the infrared imager is input into the trained deep learning fusion network to filter out the tunnel thermal map containing the leakage point area and segment the leakage point area in the tunnel thermal map;

[0124] For the segmented tunnel thermal map containing the leakage point area, the horizontal and vertical field angles are calculated to be 37.36° and 28.10° respectively; the corrected field width and field height are calculated to be 2000 mm and 1500 mm by substituting into the field calculation formula; the horizontal pixel accuracy and vertical pixel accuracy are calculated to be 3.125 mm / pix, and the final calculated leakage area is about 75 mm 2 The actual leakage area is 77mm 2 The area calculation error is 2.6%. The results show that the technical solution of the present invention has good measurement and calculation accuracy.

[0125] Embodiment 2:

[0126] like Figures 4 and 5 As shown, in addition to the technical solution of Example 1, this Example 2 is carried out in an outdoor tunnel environment, the deep learning fusion network is trained in advance, the water leakage detection calculation is performed on the tunnel vault, and the equipment is used to carry out the tunnel on-site water leakage detection experiment.

[0127] The present invention has a better effect on tunnel leakage detection, using an infrared thermal imager with the same parameters as in Example 1, that is, the infrared thermal imager has a focal length of 8 mm and a sensor size of 1 / 2.7 inches. The measured width is about 3550mm. After calculation, the horizontal and vertical angles of view are 37.36° and 28.10° respectively. The corrected width and height of the field of view are 2394 mm and 1774 mm respectively. The horizontal pixel accuracy is 3.74 mm / pi, the vertical pixel accuracy is 3.7mm / pi, and the pixel accuracy is 3.72 mm / pix. The final calculated leakage area is 83mm 2 , the actual leakage area is 80mm 2 The area calculation error is 3.6%. The results show that the technical solution of the present invention has good measurement and calculation accuracy.

[0128] Embodiment 3:

[0129] This embodiment provides a tunnel leakage identification device suitable for installing a fireproof layer, including:

[0130] Acquisition module: used to obtain tunnel thermal map;

[0131] Segmentation module: used to input the tunnel thermal map into the pre-trained deep learning fusion network, filter out the tunnel thermal map containing the leakage point area, and segment the leakage point area in the tunnel thermal map;

[0132] Calculation module: used to calculate the area of ​​the leakage point area;

[0133] The area calculation method of the leakage point area includes:

[0134] Calculate the camera's viewing angle based on the object distance, sensor size, and actual working distance in the leak area;

[0135] The field of view is calculated by combining the viewing angle with the object distance of the leakage point area, and the radial distortion of the lens is corrected for the field of view.

[0136] The pixel accuracy of the tunnel thermal image containing the leakage point area is calculated through the corrected field of view;

[0137] For the image segmented by the deep learning fusion network, filtering, grayscale and binarization processing are performed, and the number of pixels with a pixel value of 255 in the image is counted;

[0138] The area of ​​the leak point region is calculated by the number of pixels with a pixel value of 255 and the pixel accuracy in the image.

[0139] Embodiment 4:

[0140] This embodiment provides a tunnel leakage identification system, comprising:

[0141] Memory, for storing computer programs / instructions;

[0142] A processor is used to execute the computer program / instructions to implement the steps of the above-mentioned tunnel leakage identification method.

[0143] Embodiment 5:

[0144] This embodiment provides a computer-readable storage medium on which a computer program / instruction is stored. When the computer program / instruction is executed by a processor, the steps of the above-mentioned tunnel leakage identification method are implemented.

[0145] The invention is innovatively applicable to the rapid identification of tunnel leakage in which a fireproof layer is installed, realizes automatic detection of leakage points and area calculation, saves costs, and has broad prospects.

[0146] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0147] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0148] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0149] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0150] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. A tunnel leakage identification method, characterized in that: The following steps are involved: Get tunnel thermal map; Input the tunnel thermal map into the pre-trained deep learning fusion network, filter out the tunnel thermal map containing the leakage point area, and segment the leakage point area in the tunnel thermal map; Calculate the area of ​​the leakage point region in the segmented tunnel thermal map; Among them: The area calculation method of the leakage point area includes: The shooting field of view is calculated according to the shooting angle of the segmented tunnel heat map, and the radial distortion of the lens of the shooting field of view is corrected. The pixel accuracy is calculated according to the corrected shooting field of view. Perform image optimization processing on the segmented tunnel thermal map to obtain the number of effective pixels; The area of ​​the leakage point region is calculated based on the pixel accuracy and the number of effective pixels.

2. The tunnel leakage identification method according to claim 1, characterized in that: The step of screening out the tunnel thermal map containing the leakage point area and segmenting the leakage point area in the tunnel thermal map comprises: The tunnel heat map is passed through several convolutional layers and pooling layers to extract the image features, and the image features are gradually downsampled and upsampled to increase the abstraction level of the image features. The probability map of each pixel in the tunnel heat map belonging to the leakage point category is generated through the activation function, and the tunnel heat map containing the leakage point area is screened out according to the probability map; The screened tunnel heat map is compressed through encoding operation to extract multi-scale features, and then gradually restored through deconvolution or upsampling in decoding operation. At the same time, the features of the corresponding layer of the encoding operation are fused, and the detail information is transferred through jump connection. Finally, the decoding operation outputs the tunnel heat map leakage accurate segmentation image containing the leakage point area with the original resolution.

3. The tunnel leakage identification method according to claim 1, characterized in that: The step of calculating the shooting field of view of the segmented tunnel heat map according to the shooting angle includes: ， ， Where: θ h is the horizontal shooting angle of the tunnel thermal map, θ v W is the vertical shooting angle of the tunnel thermal map; sensor H is the width of the image sensor. sensor is the height of the shooting sensor; k0 is the geometric difference adjustment coefficient between the shooting sensor and the lens, f is the focal length of the lens; d o is the object distance, d f is the actual working distance; ， ， Where: W fov is the field width of the tunnel thermal map, H fov is the field of view height of the tunnel thermal image; MTF(f) is the modulation transfer function of the lens and imaging system at the spatial frequency f.

4. The tunnel leakage identification method according to claim 3, characterized in that: The radial distortion expression of the lens for correcting the shooting field of view is: ， Where: W' fov is the field of view width after considering the radial distortion correction of the lens, H' fov′ is the height of the field of view after correction of the radial distortion of the lens, and k1, k2, and k3 are the coefficients of the set radial distortion of the lens.

5. The tunnel leakage identification method according to claim 4, characterized in that: The step of calculating the pixel accuracy of the corrected shooting field of view includes: ， ， ， Where: P is the pixel accuracy of the tunnel thermal map, P x is the horizontal pixel accuracy of the tunnel thermal map, P y It is the vertical pixel accuracy of the tunnel thermal map.

6. The tunnel leakage identification method according to claim 5, characterized in that: The image optimization processing of the segmented tunnel thermal map to obtain the number of effective pixels specifically includes: The image optimization process includes: performing mean filtering, grayscale conversion and binarization operations on the segmented tunnel thermal map in sequence: Among them: The expression of the mean filter operation is: ， ， ， Where: I filtered (x,y) is the pixel value of the filtered tunnel heat map at position (x,y); m and n are the width and height of the window, I(x+i, y+j) represents the pixel value of the tunnel heat map at the x+ith row and y+jth column; a and b are the set filter kernel size parameters; i and j are the set filter kernel indexes; The grayscale expression is: ， Where: g(x, y) is the gray value at the coordinate (x, y) in the tunnel thermal map, R is the value of the red channel of the tunnel thermal map, G is the value of the green channel of the tunnel thermal map, and B is the value of the black channel of the tunnel thermal map; The binary expression is: ， Where: B(x, y) is the pixel value at coordinate (x, y) in the binarized image, g(x, y) is the pixel value in the original grayscale image, and T is the set threshold; The expression for obtaining the number of effective pixels is: ， ， in: is the pixel determination function, N is the number of valid pixels, a valid pixel is a pixel with a pixel value of 255, H is the height of the tunnel thermal map, W is the width of the tunnel thermal map, and I(i,j) represents the pixel value of the tunnel thermal map in the i-th row and j-th column.

7. The tunnel leakage identification method according to claim 6, characterized in that: Calculating the area of ​​the leakage point region according to the pixel accuracy and the number of effective pixels includes: ， Where: S is the area of ​​the leak point.

8. A tunnel leakage identification device, characterized in that: include: Acquisition module: used to obtain tunnel thermal map; Segmentation module: used to input the tunnel thermal map into the pre-trained deep learning fusion network, filter out the tunnel thermal map containing the leakage point area, and segment the leakage point area in the tunnel thermal map; Identification module: used to calculate the area of ​​the leakage point region in the segmented tunnel thermal map; Among them: The area calculation method of the leakage point area includes: The shooting field of view is calculated according to the shooting angle of the segmented tunnel heat map, and the radial distortion of the lens of the shooting field of view is corrected. The pixel accuracy is calculated according to the corrected shooting field of view. Perform image optimization processing on the segmented tunnel thermal map to obtain the number of effective pixels; The area of ​​the leakage point region is calculated based on the pixel accuracy and the number of effective pixels.

9. A tunnel leakage identification system, characterized in that: include: Memory, for storing computer programs / instructions; A processor, configured to execute the computer program / instructions to implement the steps of the tunnel leakage identification method according to any one of claims 1 to 7.

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

Images