Tunnel vehicle-mounted leakage sensing detection method based on infrared thermal imaging technology

Abstract

The invention relates to the technical field of image analysis, in particular to a tunnel vehicle-mounted leakage sensing detection method based on an infrared thermal imaging technology, which comprises the following steps: acquiring an infrared image frame sequence, and carrying out preprocessing and feature enhancement processing; extracting a gradient characteristic value and an outlier characteristic value from the enhanced infrared image, and identifying an initial leakage area; tracking each initial leakage region in the continuous enhanced infrared image frame sequence to obtain a region space sequence, and extracting a space gradient feature value of the initial leakage region; according to the gradient characteristic value, the outlier characteristic value and the space gradient characteristic value, a multi-characteristic comprehensive leakage index is calculated, and the initial leakage area is processed to obtain a leakage area; and performing feature analysis on each leakage region to obtain region multi-dimensional features, and performing severity evaluation according to the region multi-dimensional features. According to the invention, the reliability of leakage analysis and evaluation is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and in particular to a tunnel vehicle-mounted leakage perception and detection method based on infrared thermal imaging technology. Background Art

[0002] The current leakage perception method obtains the thermal radiation distribution data of the tunnel lining surface and uses digital image processing technology to construct a mapping relationship between temperature and grayscale images. The core is to convert temperature information into a multidimensional digital image matrix through grayscale mapping or pseudo-color coding algorithm, and use algorithms such as threshold segmentation and region growing to process the thermal radiation image. Morphological operations are combined to optimize the contour features of the thermal anomaly area. Specifically, a gradient edge detection method based on heat conduction difference is used to perform image processing on the lining surface temperature distribution matrix to generate a topological structure representation map of the thermal anomaly area.

[0003] However, existing technical solutions have limitations in image sequence processing in mobile detection, mainly reflected in the failure to analyze the spatiotemporal correlation between consecutive frame images. Since the continuous image sequence generated by mobile imaging acquisition has spatial displacement characteristics, the traditional analysis method based on single-frame static images cannot effectively capture the spatiotemporal development law of the target area as the detection equipment moves, resulting in a break in the spatial continuity of the detection and reducing the reliability of the results.

[0004] Therefore, a tunnel vehicle-mounted leakage perception and detection method based on infrared thermal imaging technology is proposed. Summary of the invention

[0005] The purpose of the present invention is to provide a tunnel vehicle-mounted leakage perception and detection method based on infrared thermal imaging technology. It includes: obtaining an infrared image frame sequence and performing preprocessing and feature enhancement processing; extracting gradient eigenvalues ​​and outlier eigenvalues ​​from the enhanced infrared image to identify the initial leakage area; tracking each initial leakage area in a continuous enhanced infrared image frame sequence to obtain a regional spatial sequence, extracting the spatial gradient eigenvalue of the initial leakage area; calculating a comprehensive leakage index based on the gradient eigenvalue, outlier eigenvalue and spatial gradient eigenvalue, processing the initial leakage area to obtain a leakage area; performing feature analysis on each leakage area to obtain regional multidimensional features, and performing severity assessment based on the regional multidimensional features. The present invention improves the reliability of leakage analysis and assessment.

[0006] To achieve the above object, the present invention provides the following technical solutions: A tunnel vehicle-mounted leakage perception and detection method based on infrared thermal imaging technology, comprising: Acquire an infrared image frame sequence of the tunnel detection area and perform preprocessing and feature enhancement processing on each frame of the infrared image to obtain an enhanced infrared image frame sequence; Extracting the gradient characteristic value and the outlier characteristic value of each pixel of each frame of enhanced infrared image, and identifying the initial leakage area by combining the gradient characteristic value and the outlier characteristic value; Tracking each of the initial leakage regions in the enhanced infrared image frame sequence to obtain a regional space sequence, and extracting a spatial gradient feature value of the initial leakage region; Calculating a comprehensive leakage index of the initial leakage area according to the gradient eigenvalue, the outlier eigenvalue and the spatial gradient eigenvalue; According to the comprehensive leakage index, the initial leakage area is processed by a region growing algorithm to obtain a leakage area; A feature analysis is performed on each leakage area to obtain a regional multi-dimensional feature, and a severity assessment is performed on the leakage area based on the regional multi-dimensional feature.

[0007] Furthermore, the process of preprocessing and feature enhancement processing of each frame of the infrared image specifically includes: performing Gaussian filtering denoising, geometric correction and temperature calibration on the infrared image to obtain a preprocessed infrared image; and performing feature enhancement on the preprocessed infrared image through adaptive nonlinear mapping, and the formula is: ; in, and They are the pixels before and after enhancement. The temperature value at is the average temperature of the preprocessed infrared image; is the adaptive enhancement coefficient; is the gradient sensitivity coefficient; is the pixel point in the preprocessed infrared image The gradient at .

[0008] Furthermore, the process of identifying the initial leakage area specifically includes: extracting the temperature gradient of each pixel in the enhanced infrared image to obtain the gradient eigenvalue; dividing the enhanced infrared image into uniform local areas, calculating the temperature deviation of each pixel and the local area where it is located, and obtaining the outlier eigenvalue; calculating the potential leakage index of each pixel based on the gradient eigenvalue and the outlier eigenvalue; processing the pixels whose potential leakage index is greater than the potential leakage threshold through the connected component algorithm to obtain the initial leakage area set.

[0009] Furthermore, the process of tracking each of the initial leakage areas in the enhanced infrared image frame sequence specifically includes: assigning a different identifier to each of the initial leakage areas in the enhanced infrared image frame sequence and extracting a characteristic feature; combining the vehicle motion parameters and the image acquisition parameters to calculate the first The predicted position of the initial leakage area in the frame image in the subsequent frame image; in the subsequent frame image, a search window is set with the predicted position as the center, and the characterization matching degree between the initial leakage area to be matched in the search window and the initial leakage area is calculated according to the characterization feature, and the initial leakage area to be matched with the largest characterization matching degree in the search window is marked as the tracking object of the initial leakage area, if the characterization matching degree of the tracking object is greater than or equal to the matching degree threshold, the identifier of the tracking object is changed to the identifier of the initial leakage area, if the characterization matching degree of the tracking object is less than the matching degree threshold, the tracking of the initial leakage area is terminated; the initial leakage areas to be matched with the same identifier as the initial leakage area in the enhanced infrared image frame sequence are summarized as the regional space sequence of the initial leakage area.

[0010] Furthermore, for each initial leakage area, the spatial gradient characteristic value is calculated based on the regional spatial sequence, and the calculation formula is: ; in, Initial leakage area The spatial gradient eigenvalue of Initial leakage area The length of the regional spatial sequence; is the serial number of the regional space sequence; For the The weight coefficient of the sequence number; and Initial leakage area In the and The average temperature value in the enhanced infrared image corresponding to the serial number; Enhance the spatial distance of infrared images for two adjacent frames.

[0011] Furthermore, taking the pixel point with the highest comprehensive leakage index in the initial leakage area as the seed point, applying the regional growing algorithm to the initial leakage area to obtain the leakage area, the calculation formula of the comprehensive leakage index is: ; in, Initial leakage area Medium pixel The comprehensive leakage index of , and are the weight coefficients of outlier eigenvalue, gradient eigenvalue and spatial gradient eigenvalue respectively; is the outlier eigenvalue; is the gradient eigenvalue; Initial leakage area The maximum value of the gradient eigenvalue in the enhanced infrared image; Initial leakage area The spatial gradient eigenvalue of Initial leakage area The maximum value of the spatial gradient eigenvalue of all initial leakage areas in the enhanced infrared image.

[0012] Furthermore, the process of feature extraction and severity assessment of the leakage area includes: obtaining the regional spatial sequence of the leakage area according to the mapping relationship between each leakage area and the initial leakage area, extracting the regional multidimensional features to obtain a regional multidimensional feature sequence; the regional multidimensional features include regional morphological features, regional temperature features and regional permeability features; inputting the regional multidimensional feature sequence into a leakage severity assessment model, and outputting a leakage severity index.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. First, the infrared image is preprocessed and feature enhanced, and then the initial leakage area is identified using the features of the gradient eigenvalue and the outlier eigenvalue. The gradient eigenvalue can capture the temperature change rate between the leakage area and the surrounding environment, while the outlier eigenvalue identifies abnormal temperature points by calculating the deviation between the pixel temperature and the local area temperature. The combination of these two features can more accurately identify areas with abnormal temperature distribution and obvious boundary features. The connected component algorithm is used to process pixels with potential leakage index greater than the threshold to determine the initial leakage area set, which significantly improves the accuracy of preliminary leakage identification, effectively reduces the interference of environmental noise and non-leakage abnormal heat sources, and lays the foundation for subsequent leakage area analysis.

[0014] 2. For the dynamic tracking of the initial leakage area in the vehicle-mounted detection environment, a unique identifier is first assigned to each initial leakage area and the characterization features are extracted. Then, the position of the leakage area in the subsequent frames is predicted by combining the vehicle motion parameters and image acquisition parameters. By setting a search window at the predicted position and calculating the characterization match between the initial leakage area and the target area in the window, effective tracking of the leakage area is achieved. The concept of spatial gradient eigenvalue is introduced. By calculating the temperature change in the regional spatial sequence, the law of temperature change in the leakage area with space can be reflected, which significantly improves the adaptability and detection stability of leakage detection in a mobile environment, as well as the consistency and reliability of the detection process.

[0015] 3. By combining the three key features of outlier eigenvalue, gradient eigenvalue and spatial gradient eigenvalue, the comprehensive leakage index is calculated, which comprehensively considers the static temperature characteristics and dynamic spatial variation characteristics of the leakage area, and can more comprehensively characterize the leakage phenomenon. Based on the comprehensive leakage index, the leakage area is identified, and then the regional multidimensional feature sequence is further extracted to comprehensively describe the geometric shape, temperature distribution and infiltration dynamic characteristics of the leakage area. Finally, these multidimensional feature sequences are input into the leakage severity assessment model, and the leakage severity index is output to achieve a quantitative assessment of the leakage problem. The assessment method based on the regional multidimensional feature sequence improves the accuracy and reliability of the leakage severity assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of a tunnel vehicle-mounted leakage sensing and detection method based on infrared thermal imaging technology of the present invention; Figure 2 A schematic diagram of the process of tracking the initial leakage area of ​​the present invention; Figure 3 It is a data flow diagram of the present invention. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] See also Figures 1 to 3 The present invention provides a tunnel vehicle-mounted leakage perception detection method based on infrared thermal imaging technology, and the technical solution is as follows: Embodiment 1: This embodiment is applied to the leakage detection of a certain city tunnel. By using the vehicle-mounted infrared thermal imaging equipment, the infrared image of the tunnel wall is collected. In order to realize the automatic detection of the leakage area and the severity assessment, a tunnel vehicle-mounted leakage perception and detection method based on infrared thermal imaging technology is applied. Figure 1 ,include: Acquire an infrared image frame sequence of the tunnel detection area and perform preprocessing and feature enhancement processing on each frame of the infrared image to obtain an enhanced infrared image frame sequence; Extracting the gradient characteristic value and the outlier characteristic value of each pixel of each frame of enhanced infrared image, and identifying the initial leakage area by combining the gradient characteristic value and the outlier characteristic value; Tracking each of the initial leakage regions in the enhanced infrared image frame sequence to obtain a regional space sequence, and extracting a spatial gradient feature value of the initial leakage region; Calculating a comprehensive leakage index of the initial leakage area according to the gradient eigenvalue, the outlier eigenvalue and the spatial gradient eigenvalue; According to the comprehensive leakage index, the initial leakage area is processed by a region growing algorithm to obtain a leakage area; A feature analysis is performed on each leakage area to obtain a regional multi-dimensional feature, and a severity assessment is performed on the leakage area based on the regional multi-dimensional feature.

[0019] Furthermore, the process of preprocessing and feature enhancement processing of each frame of the infrared image specifically includes: performing Gaussian filtering denoising, geometric correction and temperature calibration on the infrared image to obtain a preprocessed infrared image; and performing feature enhancement on the preprocessed infrared image through adaptive nonlinear mapping, and the formula is: ; in, and They are the pixels before and after enhancement. The temperature value at is the average temperature of the preprocessed infrared image; It is an adaptive reinforcement factor, which is adjusted according to the type of fireproof panels installed in the tunnel; is the gradient sensitivity coefficient; is the pixel point in the preprocessed infrared image The gradient at .

[0020] The adjustment formula of the adaptive enhancement coefficient is: ;in, is the basic enhancement coefficient, which is 1.2; is the material adjustment coefficient. For example, the values ​​of different materials are: cement-based fireproof board is 0, calcium silicate fireproof board is 0.15, fiber-reinforced fireproof board is 0.3, and mineral fiber fireproof board is 0.25; is the thickness adjustment coefficient, and the calculation formula is: ;in, is the actual thickness of the fireproof board (mm), The standard thickness is 10mm.

[0021] The enhancement coefficient adjustment formula based on material and thickness is adopted, which can optimize parameters for fireproof boards of different materials and thicknesses, improving the pertinence and adaptability of image enhancement. Preprocessing and feature enhancement can effectively improve the quality of infrared images, making the leakage area more prominent in the image, and providing a more reliable data basis for subsequent leakage detection.

[0022] Furthermore, the process of identifying the initial leakage area specifically includes: extracting the temperature gradient of each pixel in the enhanced infrared image to obtain the gradient eigenvalue; dividing the enhanced infrared image into uniform local areas, calculating the temperature deviation of each pixel and the local area where it is located, and obtaining the outlier eigenvalue; calculating the potential leakage index of each pixel based on the gradient eigenvalue and the outlier eigenvalue; processing the pixels whose potential leakage index is greater than the potential leakage threshold through the connected component algorithm to obtain the initial leakage area set.

[0023] Specifically, the maximum inter-class variance method is used to determine the potential leakage threshold; the calculation formula of the potential leakage index is: ; in, is the pixel point in the initial leakage area Potential leakage index; and are the weight coefficients of the outlier eigenvalue and the gradient eigenvalue respectively; is the outlier eigenvalue; is the gradient eigenvalue; It is the maximum value of the gradient eigenvalue in the enhanced infrared image where the initial leakage area is located.

[0024] By calculating the gradient eigenvalue and the outlier eigenvalue and combining them to calculate the potential leakage index, the maximum inter-class variance method is used to determine the threshold for pixel screening, and finally the connected component algorithm is applied to obtain the initial leakage area set. This method takes into account both local temperature changes and global temperature anomalies, can capture leakage characteristics more comprehensively, reduce the misjudgment caused by a single feature, and improve the accuracy and reliability of initial leakage area identification.

[0025] Further, see Figure 2 The process of tracking each of the initial leakage areas in the enhanced infrared image frame sequence specifically includes: Assigning a different identifier to each of the initial leakage areas in the enhanced infrared image frame sequence and extracting characterizing features; Combine the vehicle motion parameters and image acquisition parameters to calculate the The predicted position of the initial leakage area in the frame image in the subsequent frame image; In the subsequent frame image, a search window is set with the predicted position as the center, and the characterization matching degree between the initial leakage area to be matched in the search window and the initial leakage area is calculated according to the characterization feature, and the initial leakage area to be matched with the largest characterization matching degree in the search window is marked as the tracking object of the initial leakage area, and if the characterization matching degree of the tracking object is greater than or equal to the matching degree threshold, the identifier of the tracking object is changed to the identifier of the initial leakage area, and if the characterization matching degree of the tracking object is less than the matching degree threshold, the tracking of the initial leakage area is terminated; The initial leakage regions to be matched that have the same identifier as the initial leakage region in the enhanced infrared image frame sequence are summarized as a region space sequence of the initial leakage region.

[0026] Specifically, the characterization features include geometric features, temperature features and texture features; the geometric features include the area, perimeter, circularity and rectangularity of the region; the temperature features include the average temperature, maximum temperature, minimum temperature and temperature standard deviation in the region; the texture features include entropy, contrast, correlation and homogeneity calculated by the gray level co-occurrence matrix; the geometric similarity, temperature similarity and shape similarity are calculated respectively according to the geometric features, temperature features and texture features, and the geometric similarity, temperature similarity and shape similarity are weightedly fused to obtain the characterization matching degree; The calculation method of the predicted position is as follows: considering that the imaging device is installed on the vehicle, the movement of the vehicle causes the relative displacement of the stationary object in the image. If the vehicle moves forward in the longitudinal direction of the tunnel, the stationary point on the tunnel wall will move backward in the image, and the default point will not change in the vertical direction. Taking the tunnel wall on the right side of the vehicle's driving direction as an example, the time interval between two frames of images of the vehicle is obtained. Speed ​​within and imaging device parameters to calculate the horizontal displacement of pixels , get the The initial leakage area in the frame image is Predicted position in the frame image , Indicates The centroid coordinates of the initial leakage area in the frame image, displacement The calculation formula is ;in, is the conversion coefficient of pixel displacement corresponding to unit actual distance; is the focal length of the infrared thermal imaging device; is the distance between the infrared thermal imaging device and the tunnel wall. Considering the uncertainty of the prediction, the size of the search window is set to twice the bounding box of the initial leakage area to ensure that the tracked object is within the search window.

[0027] In this embodiment, the vehicle equipped with the infrared thermal imaging device travels at a speed of 20 km / h to 30 km / h, and the sampling frequency of the infrared thermal imaging device is 25 Hz.

[0028] By assigning a unique identifier to each initial leakage area and extracting characterization features, the position of the area in subsequent frames is predicted by combining vehicle motion parameters and image acquisition parameters, a search window is set and the feature matching degree is calculated for tracking matching. The relative motion characteristics in the vehicle scene are considered, and the image displacement problem caused by vehicle movement is effectively solved. The same leakage area can maintain a consistent identification in a continuous image sequence, providing a basis for subsequent spatial change characteristic analysis.

[0029] Furthermore, for each initial leakage area, the spatial gradient characteristic value is calculated based on the regional spatial sequence, and the calculation formula is: ; in, Initial leakage area The spatial gradient eigenvalue of Initial leakage area The length of the regional spatial sequence; is the serial number of the regional space sequence; For the The weight coefficient of the sequence number; and Initial leakage area In the and The average temperature value in the enhanced infrared image corresponding to the serial number; Enhance the spatial distance of infrared images for two adjacent frames.

[0030] The relationship between the temperature change and spatial distance of adjacent frames in the regional spatial sequence is analyzed by calculating the spatial gradient eigenvalue, and the spatial temperature change trend of the leakage area is quantified, which can reflect the spatial diffusion characteristics of the leakage phenomenon and help distinguish between real leakage and surface temperature anomalies. Real leakage often shows certain gradient eigenvalue distribution characteristics in space, while ordinary temperature anomaly points lack this characteristic. By introducing the spatial gradient eigenvalue, the system's ability to identify leakage phenomena is significantly improved.

[0031] Furthermore, taking the pixel point with the highest comprehensive leakage index in the initial leakage area as the seed point, applying the regional growing algorithm to the initial leakage area to obtain the leakage area, the calculation formula of the comprehensive leakage index is: ; in, Initial leakage area Medium pixel The comprehensive leakage index of , and are the weight coefficients of outlier eigenvalue, gradient eigenvalue and spatial gradient eigenvalue respectively; is the outlier eigenvalue; is the gradient eigenvalue; Initial leakage area The maximum value of the gradient eigenvalue in the enhanced infrared image; Initial leakage area The spatial gradient eigenvalue of Initial leakage area The maximum value of the spatial gradient eigenvalue of all initial leakage areas in the enhanced infrared image.

[0032] A comprehensive leakage index is constructed by comprehensively considering the outlier eigenvalue, gradient eigenvalue and spatial gradient eigenvalue. The regional segmentation method based on multi-feature fusion can more accurately determine the boundary of the leakage area, avoiding the over-segmentation or under-segmentation problem caused by a single feature. At the same time, it can adapt to different leakage detection scenarios and improve the accuracy and robustness of the segmentation results.

[0033] Furthermore, the process of feature extraction and severity assessment of the leakage area includes: obtaining the regional spatial sequence of the leakage area according to the mapping relationship between each leakage area and the initial leakage area, extracting the regional multidimensional features to obtain a regional multidimensional feature sequence; the regional multidimensional features include regional morphological features, regional temperature features and regional permeability features; inputting the regional multidimensional feature sequence into a leakage severity assessment model, and outputting a leakage severity index.

[0034] By extracting regional multidimensional features and constructing feature sequences, and inputting them into the leakage severity assessment model for evaluation, all aspects of the leakage phenomenon are fully considered. It can not only detect the existence of leakage, but also assess the severity of leakage, providing a more specific quantitative basis for tunnel maintenance decisions. This is of great value for timely discovering dangerous tunnel conditions, formulating maintenance priorities and preventive maintenance strategies.

[0035] This embodiment uses a gradient boosting tree model as the basic model for leakage severity assessment, and adjusts and optimizes it in combination with expert knowledge. Model training uses 300 sets of leakage case data evaluated by experts. During the training process, the depth of the tree is set to 6, the learning rate is 0.05, the number of iterations is 500, and a five-fold cross validation is used to ensure model stability. In order to improve the sensitivity of the model to samples of different severity, a weighted loss function is introduced to give higher weights to samples of high severity. The regional multidimensional feature sequence is used as the input of the leakage severity assessment model, the features are normalized to eliminate the dimensional effect, and principal component analysis is used to reduce the redundancy between features. The model output is a leakage severity index from 0 to 100, and the original output of the model is mapped to the interval [0,100] through the activation function. The leakage severity index ranges from [0,20] for slight leakage, [21,40] for relatively light leakage, [41,60] for moderate leakage, [61,80] for relatively severe leakage, and [81,100] for severe leakage requiring immediate attention.

[0036] The present invention firstly utilizes infrared thermal imaging technology for non-contact detection, thus realizing high-efficiency and all-round monitoring of tunnel leakage, thus avoiding the disadvantages of low efficiency, high risk and strong subjectivity of traditional manual detection. Secondly, through image preprocessing and feature enhancement, the recognizability of the leakage area in the image is significantly improved. Thirdly, the initial leakage area is divided by combining the dual features of gradient eigenvalue and outlier eigenvalue, thus improving the reliability of preliminary detection. The introduction of regional tracking and spatial gradient feature analysis solves the problem of target tracking in a vehicle-mounted environment, realizes consistent recognition of the same leakage area in continuous images, and effectively distinguishes real leakage from surface temperature anomalies by analyzing its spatial temperature variation characteristics. The comprehensive leakage index and the regional growing algorithm based on the index further improve the accuracy of leakage area segmentation. Finally, through the regional multi-dimensional feature extraction and severity assessment model, the quantitative assessment of the degree of leakage is realized, providing a scientific basis for tunnel maintenance decision-making.

[0037] Embodiment 2: A tunnel vehicle-mounted leakage detection method based on infrared thermal imaging technology. The data flow is shown in Figure 3 , the specific steps include: Acquire an infrared image frame sequence of the tunnel detection area and perform preprocessing and feature enhancement processing on each frame of the infrared image to obtain an enhanced infrared image frame sequence; Furthermore, the process of preprocessing and feature enhancement processing of each frame of the infrared image specifically includes: performing Gaussian filtering denoising, geometric correction and temperature calibration on the infrared image to obtain a preprocessed infrared image; and performing feature enhancement on the preprocessed infrared image through adaptive nonlinear mapping, and the formula is: ; in, and They are the pixels before and after enhancement. The temperature value at is the average temperature of the preprocessed infrared image; is the adaptive enhancement coefficient; is the gradient sensitivity coefficient; is the pixel point in the preprocessed infrared image The gradient at .

[0038] Extracting the gradient characteristic value and the outlier characteristic value of each pixel of each frame of enhanced infrared image, and identifying the initial leakage area by combining the gradient characteristic value and the outlier characteristic value; Furthermore, the process of identifying the initial leakage area specifically includes: extracting the temperature gradient of each pixel in the enhanced infrared image to obtain the gradient eigenvalue; dividing the enhanced infrared image into uniform local areas, calculating the temperature deviation of each pixel and the local area where it is located, and obtaining the outlier eigenvalue; calculating the potential leakage index of each pixel based on the gradient eigenvalue and the outlier eigenvalue; processing the pixels whose potential leakage index is greater than the potential leakage threshold through the connected component algorithm to obtain the initial leakage area set.

[0039] Table 1 Initial leakage area identification results Detection area number Number of initial leakage areas Average gradient eigenvalue Average outlier eigenvalue False alarm rate (%) 1 12 0.38 2.7 8.3 2 9 0.42 3.1 6.7 3 15 0.51 3.5 9.8 Table 1 gives an example of the initial leakage area identification results. The initial leakage area division algorithm based on gradient eigenvalue and outlier degree achieves an average false alarm rate of ≤10% in a complex tunnel environment.

[0040] Tracking each of the initial leakage regions in the enhanced infrared image frame sequence to obtain a regional space sequence, and extracting a spatial gradient feature value of the initial leakage region; Furthermore, the process of tracking each of the initial leakage areas in the enhanced infrared image frame sequence specifically includes: assigning a different identifier to each of the initial leakage areas in the enhanced infrared image frame sequence and extracting a characteristic feature; combining the vehicle motion parameters and the image acquisition parameters to calculate the first The predicted position of the initial leakage area in the frame image in the subsequent frame image; in the subsequent frame image, a search window is set with the predicted position as the center, and the characterization matching degree between the initial leakage area to be matched in the search window and the initial leakage area is calculated according to the characterization feature, and the initial leakage area to be matched with the largest characterization matching degree in the search window is marked as the tracking object of the initial leakage area, if the characterization matching degree of the tracking object is greater than or equal to the matching degree threshold, the identifier of the tracking object is changed to the identifier of the initial leakage area, if the characterization matching degree of the tracking object is less than the matching degree threshold, the tracking of the initial leakage area is terminated; the initial leakage areas to be matched with the same identifier as the initial leakage area in the enhanced infrared image frame sequence are summarized as the regional space sequence of the initial leakage area.

[0041] Table 2 Initial leakage area tracking results Vehicle speed (km / h) Tracking success rate (%) Search window size (pixels) Average matching threshold 10 98.5 40×40 0.92 20 95.2 60×60 0.88 30 89.7 80×80 0.85 Table 2 shows the initial leakage area tracking results obtained under different vehicle speeds and search window sizes. The search window size is dynamically adjusted in combination with the vehicle motion parameters, and the tracking success rate is still greater than 85% at a speed of 30 km / h, meeting the vehicle dynamic detection requirements.

[0042] Furthermore, for each initial leakage area, the spatial gradient characteristic value is calculated based on the regional spatial sequence, and the calculation formula is: ; in, Initial leakage area The spatial gradient eigenvalue of Initial leakage area The length of the regional spatial sequence; is the serial number of the regional space sequence; For the The weight coefficient of the sequence number; and Initial leakage area In the and The average temperature value in the enhanced infrared image corresponding to the serial number; Enhance the spatial distance of infrared images for two adjacent frames.

[0043] Calculating a comprehensive leakage index of the initial leakage area according to the gradient eigenvalue, the outlier eigenvalue and the spatial gradient eigenvalue; According to the comprehensive leakage index, the initial leakage area is processed by a region growing algorithm to obtain a leakage area; Furthermore, taking the pixel point with the highest comprehensive leakage index in the initial leakage area as the seed point, applying the regional growing algorithm to the initial leakage area to obtain the leakage area, the calculation formula of the comprehensive leakage index is: ; in, Initial leakage area Medium pixel The comprehensive leakage index of , and are the weight coefficients of outlier eigenvalue, gradient eigenvalue and spatial gradient eigenvalue respectively; is the outlier eigenvalue; is the gradient eigenvalue; Initial leakage area The maximum value of the gradient eigenvalue in the enhanced infrared image; Initial leakage area The spatial gradient eigenvalue of Initial leakage area The maximum value of the spatial gradient eigenvalue of all initial leakage areas in the enhanced infrared image.

[0044] Performing feature analysis on each leakage area to obtain a regional multi-dimensional feature, and evaluating the severity of the leakage area according to the regional multi-dimensional feature; Furthermore, the process of feature extraction and severity assessment of the leakage area includes: obtaining the regional spatial sequence of the leakage area according to the mapping relationship between each leakage area and the initial leakage area, extracting the regional multidimensional features to obtain a regional multidimensional feature sequence; the regional multidimensional features include regional morphological features, regional temperature features and regional permeability features; inputting the regional multidimensional feature sequence into a leakage severity assessment model, and outputting a leakage severity index.

[0045] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A tunnel vehicle-mounted leakage perception and detection method based on infrared thermal imaging technology, characterized in that: include: Acquire an infrared image frame sequence of the tunnel detection area and perform preprocessing and feature enhancement processing on each frame of the infrared image to obtain an enhanced infrared image frame sequence; Extracting the gradient characteristic value and the outlier characteristic value of each pixel of each frame of enhanced infrared image, and identifying the initial leakage area by combining the gradient characteristic value and the outlier characteristic value; Tracking each of the initial leakage regions in the enhanced infrared image frame sequence to obtain a regional space sequence, and extracting a spatial gradient feature value of the initial leakage region; Calculating a comprehensive leakage index of the initial leakage area according to the gradient eigenvalue, the outlier eigenvalue and the spatial gradient eigenvalue; According to the comprehensive leakage index, the initial leakage area is processed by a region growing algorithm to obtain a leakage area; A feature analysis is performed on each leakage area to obtain a regional multi-dimensional feature, and a severity assessment is performed on the leakage area based on the regional multi-dimensional feature.

2. According to claim 1, a tunnel vehicle-mounted leakage perception and detection method based on infrared thermal imaging technology is characterized in that: The process of preprocessing and feature enhancement of each frame of the infrared image specifically includes: performing Gaussian filtering denoising, geometric correction and temperature calibration on the infrared image to obtain a preprocessed infrared image; and performing feature enhancement on the preprocessed infrared image through adaptive nonlinear mapping, the formula is: ； in, and They are the pixels before and after enhancement. The temperature value at is the average temperature of the preprocessed infrared image; is the adaptive enhancement coefficient; is the gradient sensitivity coefficient; is the pixel point in the preprocessed infrared image The gradient at .

3. According to claim 1, a tunnel vehicle-mounted leakage perception and detection method based on infrared thermal imaging technology is characterized in that: The process of identifying the initial leakage area specifically includes: extracting the temperature gradient of each pixel in the enhanced infrared image to obtain the gradient eigenvalue; dividing the enhanced infrared image into uniform local areas, calculating the temperature deviation of each pixel and the temperature of the local area, and obtaining the outlier eigenvalue; calculating the potential leakage index of each pixel according to the gradient eigenvalue and the outlier eigenvalue; processing the pixels whose potential leakage index is greater than the potential leakage threshold through the connected component algorithm to obtain the initial leakage area set.

4. According to claim 1, a tunnel vehicle-mounted leakage perception and detection method based on infrared thermal imaging technology is characterized in that: The process of tracking each of the initial leakage areas in the enhanced infrared image frame sequence specifically includes: assigning a different identifier to each of the initial leakage areas in the enhanced infrared image frame sequence and extracting a characteristic feature; calculating the first leakage area by combining the vehicle motion parameters and the image acquisition parameters; The predicted position of the initial leakage area in the frame image in the subsequent frame image; in the subsequent frame image, a search window is set with the predicted position as the center, and the characterization matching degree between the initial leakage area to be matched in the search window and the initial leakage area is calculated according to the characterization feature, and the initial leakage area to be matched with the largest characterization matching degree in the search window is marked as the tracking object of the initial leakage area, if the characterization matching degree of the tracking object is greater than or equal to the matching degree threshold, the identifier of the tracking object is changed to the identifier of the initial leakage area, if the characterization matching degree of the tracking object is less than the matching degree threshold, the tracking of the initial leakage area is terminated; the initial leakage areas to be matched with the same identifier as the initial leakage area in the enhanced infrared image frame sequence are summarized as the regional space sequence of the initial leakage area.

5. The tunnel vehicle-mounted leakage sensing and detection method based on infrared thermal imaging technology according to claim 1 is characterized in that: For each initial leakage area, the spatial gradient characteristic value is calculated based on the regional spatial sequence, and the calculation formula is: ； in, Initial leakage area The spatial gradient eigenvalue of Initial leakage area The length of the regional spatial sequence; is the serial number of the regional space sequence; For the The weight coefficient of the sequence number; and Initial leakage area In the and The average temperature value in the enhanced infrared image corresponding to the serial number; Enhance the spatial distance of infrared images for two adjacent frames.

6. The tunnel vehicle-mounted leakage sensing and detection method based on infrared thermal imaging technology according to claim 1 is characterized in that: The pixel point with the highest comprehensive leakage index in the initial leakage area is used as the seed point, and the region growing algorithm is applied to the initial leakage area to obtain the leakage area. The calculation formula of the comprehensive leakage index is: ； in, Initial leakage area Medium pixel The comprehensive leakage index of , and are the weight coefficients of outlier eigenvalue, gradient eigenvalue and spatial gradient eigenvalue respectively; is the outlier eigenvalue; is the gradient eigenvalue; Initial leakage area The maximum value of the gradient eigenvalue in the enhanced infrared image; Initial leakage area The spatial gradient eigenvalue of Initial leakage area The maximum value of the spatial gradient eigenvalue of all initial leakage areas in the enhanced infrared image.

7. The tunnel vehicle-mounted leakage sensing and detection method based on infrared thermal imaging technology according to claim 1 is characterized in that: The process of feature extraction and severity assessment of the leakage area includes: obtaining the regional spatial sequence of the leakage area according to the mapping relationship between each leakage area and the initial leakage area, extracting the regional multidimensional features to obtain a regional multidimensional feature sequence; the regional multidimensional features include regional morphological features, regional temperature features and regional permeability features; inputting the regional multidimensional feature sequence into a leakage severity assessment model, and outputting a leakage severity index.

Images