Multi-source data feature-based rail transit tunnel disease detection method and system

Abstract

The invention discloses a rail transit tunnel disease detection method and system based on multi-source data features, and relates to the technical field of tunnel health monitoring, and the method comprises the steps: obtaining the global visible light image, infrared image and laser radar point cloud data of a multi-stage rail transit tunnel; according to the visible light image, an image recognition algorithm is adopted, and a suspected water leakage area and a suspected segment slab staggering area in the tunnel are recognized; performing temperature difference judgment based on the infrared image, refining a suspected water leakage area, mapping the clear water leakage area into the visible light image, and judging the development trend of the water leakage according to the water leakage area in the multi-stage visible light image; according to the laser radar point cloud data, the contour deformation condition of different sections of the tunnel, the land subsidence condition, the multi-stage tunnel deformation condition and the rail component installation condition are recognized, the water accumulation amount in the tunnel is judged, and segment slab staggering information is obtained in combination with the suspected segment slab staggering area; and according to various disease detection results, more comprehensive rail traffic tunnel safety evaluation is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel safety and health monitoring, and in particular to a method and system for detecting rail transit tunnel diseases based on multi-source data characteristics. Background Art

[0002] The statements in this part only provide background technical information related to the present invention and do not necessarily constitute prior art.

[0003] Currently, as one of the main forces of important transportation in cities, urban rail transit has developed rapidly. However, rail transit has disadvantages such as a closed environment and a narrow space, and its safety is the current key concern. Therefore, the detection of safety information in rail transit tunnel engineering, that is, the detection of tunnel diseases such as cracks, cavities, water leakage, and deformation that affect the safety of rail transit tunnels, has become the top priority.

[0004] At present, the automated disease detection means for rail transit tunnels are generally imperfect, and most still remain in the stage of manual inspection. This method not only has a huge workload, but also has a long detection time and high safety risks. For example, tunnel deformation is mostly detected by total station for cross-section spot checks, with a large interval between spot-check cross-sections and low efficiency; the detection of tunnel cracks and water leakage mainly relies on manual inspections, occupying a large amount of human resources; during the tunnel operation and maintenance stage, the status of sleepers, rails, and connecting parts needs to be inspected regularly by personnel to find out the safety status of each component of the track, consuming a large amount of manpower and material resources.

[0005] Although various automated disease detection schemes have also been proposed in the prior art, such as using intelligent robots carrying data acquisition equipment for automatic inspection of rail transit tunnels, however, this method usually can only detect single diseases and cannot achieve a comprehensive evaluation of the safety of rail transit tunnels; in addition, disease detection is usually based on single-sensor detection and identification, which has certain discrimination errors and cannot guarantee the accuracy of detection and evaluation. Summary of the Invention

[0006] To solve the above deficiencies of the prior art, the present invention provides a method and system for detecting rail transit tunnel diseases based on multi-source data characteristics. By analyzing and processing the multi-period and multi-source data of the rail transit tunnel obtained, various disease conditions such as cracks, water leakage, segment misalignment, deformation, and settlement existing in the rail transit tunnel are comprehensively detected, so as to achieve a comprehensive safety assessment of the rail transit tunnel. Its detection accuracy and detection efficiency are high, and its versatility is strong, and it can be applied to the comprehensive disease detection and safety evaluation of different types of rail transit tunnels at different stages.

[0007] In the first aspect, the present invention provides a method for detecting rail transit tunnel diseases based on multi-source data characteristics.

[0008] A method for detecting rail transit tunnel diseases based on multi-source data characteristics includes:

[0009] Perform multi - period data collection along the data collection trajectory for the same rail transit tunnel line to obtain visible light images, infrared images, and lidar point cloud data of the entire rail transit tunnel for multiple periods.

[0010] Based on the visible light images, use image recognition algorithms to identify suspected water leakage areas and suspected segment misalignment areas in the tunnel.

[0011] Based on the infrared images, perform temperature difference judgment to refine the suspected water leakage areas, map the identified water leakage areas to the visible light images, and then judge the development trend of water leakage according to the water leakage areas in the multi - period visible light images.

[0012] Based on the lidar point cloud data, identify the deformation of different cross - section contours of the tunnel, ground settlement, deformation of the tunnel for multiple periods, and the installation of rail components, determine the water accumulation in the tunnel, and combine with the suspected segment misalignment areas to obtain segment misalignment information.

[0013] Evaluate the safety of the rail transit tunnel based on the detection results of various diseases.

[0014] In a second aspect, the present invention provides a rail transit tunnel disease detection system based on multi - source data characteristics.

[0015] A rail transit tunnel disease detection system based on multi - source data characteristics, comprising:

[0016] A data collection module, which is used to perform multi - period data collection along the data collection trajectory for the same rail transit tunnel line to obtain visible light images, infrared images, and lidar point cloud data of the entire rail transit tunnel for multiple periods.

[0017] A disease detection module, which is used to, based on the visible light images, use image recognition algorithms to identify suspected water leakage areas and suspected segment misalignment areas in the tunnel; perform temperature difference judgment based on the infrared images to refine the suspected water leakage areas, map the identified water leakage areas to the visible light images, and then judge the development trend of water leakage according to the water leakage areas in the multi - period visible light images; based on the lidar point cloud data, identify the deformation of different cross - section contours of the tunnel, ground settlement, deformation of the tunnel for multiple periods, and the installation of rail components, determine the water accumulation in the tunnel, and combine with the suspected segment misalignment areas to obtain segment misalignment information.

[0018] A safety evaluation module, which is used to evaluate the safety of the rail transit tunnel based on the detection results of various diseases.

[0019] In a third aspect, the present invention further provides an electronic device, including a memory, a processor, and computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the steps of the method described in the first aspect are completed.

[0020] In a fourth aspect, the present invention also provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the steps of the method described in the first aspect.

[0021] The above one or more technical solutions have the following beneficial effects:

[0022] 1. The present invention provides a method and system for detecting rail transit tunnel diseases based on multi-source data features. For the same rail transit tunnel line, multi-phase and multi-source data such as global visible light images, infrared images, and lidar point cloud data of the rail transit tunnel are obtained along the data acquisition trajectory. By analyzing and processing the obtained multi-phase and multi-source data, various diseases existing in the rail transit tunnel are comprehensively detected, such as tunnel apparent cracks, leakage conditions, leakage development trends, tunnel segment misalignment, deformation conditions of different tunnel cross-section profiles, ground settlement conditions, tunnel deformation conditions, and installation conditions of rail components, etc., so as to realize comprehensive disease detection and safety assessment of the rail transit tunnel; this method proposed by the present invention can be applied to different stages of different types of rail transit tunnels, has strong versatility, high detection accuracy and detection efficiency, and combines multiple detection methods for evaluation, improving the comprehensiveness of the safety assessment of the rail transit tunnel.

[0023] 2. The method for detecting rail transit tunnel diseases based on multi-source data features proposed by the present invention identifies diseases such as tunnel apparent cracks, suspected leakage areas, and suspected segment misalignment areas through image recognition methods, and obtains the approximate locations of the diseases through positioning data. At the same time, on the basis of image recognition, infrared images are combined for analysis, and the leakage areas are refined by judging temperature and temperature difference, and the leakage development trend is judged by the range of leakage areas at the same part identified by multi-phase data. Combining information such as positioning data to achieve disease positioning; in addition, analyzing the point cloud data obtained by lidar, respectively identifying the deformation conditions of different tunnel cross-section profiles in a single time and the deformation analysis of multi-phase point clouds, and combining the recognition results of visible light images, verifying the suspected segment misalignment areas according to the morphological characteristics of segment installation to ensure the accuracy of the final detection results; using the obtained track point cloud data in the tunnel for deformation analysis of rails, etc. and ground settlement analysis, as well as detection of the accumulated water volume in the tunnel and the absence and looseness of components such as rail fixing bolts; through the above methods, comprehensive, efficient, and accurate detection and identification of various diseases in the rail transit tunnel are realized, so as to comprehensively and accurately evaluate the safety of the rail transit tunnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The specification drawings forming a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention.

[0025] Figure 1 This is a flowchart of the rail transit tunnel disease detection method based on multi-source data features according to the embodiments of the present invention. Specific embodiments

[0026] It should be noted that the following detailed description is exemplary only for describing specific embodiments, aiming to provide further explanation of the present invention, and is not intended to limit the exemplary embodiments according to the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs. In addition, it should also be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0027] Embodiment 1

[0028] This embodiment provides a rail transit tunnel disease detection method based on multi-source data features. This method consists of a multi-source feature recognition algorithm based on visible light images, infrared images, and lidar point cloud data. Multiple detection methods are jointly evaluated to reduce the discrimination error between single sensors, improve the accuracy and efficiency of disease detection, and ensure the accuracy and comprehensiveness of the safety evaluation of rail transit tunnels. As Figure 1 shown, the method proposed in this embodiment specifically includes the following steps:

[0029] Step S1: Perform multi-period data collection along the data collection trajectory for the same rail transit tunnel line to obtain multi-period visible light images, infrared images, and lidar point cloud data of the entire rail transit tunnel.

[0030] Specifically, for the same rail transit tunnel line, multi-period multi-source data collection is performed on the tunnel along the preset data collection trajectory. The multi-source data includes visible light images, infrared images, and lidar point cloud data of the entire rail transit tunnel, etc.

[0031] Step S2: According to the visible light images, use an image recognition algorithm to identify the suspected water leakage area and the suspected segment dislocation area in the tunnel.

[0032] In this embodiment, the visible light images of the rail transit tunnel are obtained through a camera, and an image recognition algorithm, such as a traditional YOLO-based image recognition algorithm, is used to detect and identify diseases such as apparent cracks in the tunnel, suspected water leakage areas, and suspected segment dislocation areas in the subway tunnel, and the approximate location of the diseases is obtained through positioning data.

[0033] Specifically, for the identification of suspected water leakage areas, based on visible light images, through an image recognition algorithm based on YOLO, the features of color changes, water marks, flowing traces, and moss areas are extracted. According to the extracted features, the edges are identified and the area contours are divided to determine the suspected water leakage areas. Similarly, for the identification of suspected segment misalignment areas, based on visible light images, feature extraction is performed through an image recognition algorithm based on YOLO, and the segment misalignment areas are identified according to the extracted features, thereby determining the suspected water leakage areas.

[0034] Step S3: Perform temperature difference judgment based on infrared images, refine the suspected water leakage areas, map the identified water leakage areas to visible light images, and judge the development trend of water leakage based on the water leakage areas in multi-period visible light images.

[0035] In this embodiment, based on the identification of suspected water leakage areas in images, temperature and temperature difference analysis are combined with infrared images to further judge and clarify the water leakage areas. Then, the range of water leakage areas at the same part identified by multi-period data is used to judge the development trend of water leakage, and disease location is achieved by combining information such as positioning data.

[0036] First, considering that water leakage mostly occurs in the rainy season in summer, the water temperature is higher than that in the normal segment area. For areas without water leakage, the temperature mean and standard deviation of this area will fluctuate within a certain range, while for areas with water leakage, the temperature of this area will change significantly. Infrared images can map temperature information into color information. Different color areas in infrared images indicate the existence of temperature differences. Based on this characteristic, the water leakage areas of rail transit tunnels can be further clarified by combining infrared images. Specifically, temperature difference judgment is performed based on infrared images to refine the suspected water leakage areas into water leakage areas, that is: map the suspected water leakage areas to infrared images, divide the suspected water leakage areas into different areas, and calculate the temperature mean of each area of the suspected water leakage respectively. Since the temperature mean of the water leakage area is significantly higher than that of the normal area, and the temperature at the edge of the water leakage area shows an obvious upward trend, the areas with temperature mean higher than the normal temperature are statistically counted, and the identified water leakage areas are obtained after integration.

[0037] Secondly, the development trend of water leakage can be further determined according to the change situation of the identified water leakage areas. Specifically, the development trend of water leakage is judged based on the water leakage areas in multi-period visible light images, that is: according to the water leakage areas in multi-period visible light images sorted in time series, calculate the number of corresponding pixels in the same water leakage area in each period. According to the change of the number of pixels in multi-periods, judge the development trend of the water leakage area. If the number of pixels is rising linearly or fluctuating upward, it indicates that the water leakage part is in a developing state.

[0038] Step S4: Based on the lidar point cloud data, identify the deformation conditions of different cross-section profiles of the tunnel, the deformation conditions of the tunnel in multiple periods, the ground settlement conditions, and the installation conditions of the track components, determine the water accumulation volume in the tunnel, and combine the suspected segment misalignment areas to obtain segment misalignment information.

[0039] In this embodiment, based on the point cloud data obtained by the lidar, the tunnel deformation conditions are detected, which can be respectively used to identify the deformation conditions of different cross-section profiles of a single (or single-period) tunnel and the tunnel deformation conditions of multi-period point clouds. In addition, based on the image recognition results, the segment misalignment can be verified according to the morphological characteristics of the segment installation; the obtained track point cloud data is used for deformation analysis of the track and ground settlement analysis, determine the water accumulation volume in the tunnel, and identify the missing and loose conditions of components such as track fixing bolts. The identification of the above various conditions specifically includes:

[0040] (1) Based on the lidar point cloud data, identify the deformation conditions of different cross-section profiles of a single-period tunnel. Considering that the designed cross-section dimensions of a certain section of the same tunnel are the same, at this time, a typical cross-section can be selected for sliding detection based on the solution trajectory, and the deformation threshold is set according to the point cloud resolution to detect the deformation conditions of this section in real time.

[0041] Specifically, for each tunnel cross-section, based on the tunnel contour point cloud of the tunnel cross-section, calculate the central point coordinates of the tunnel contour point cloud; move the central point coordinates along the data acquisition trajectory, and the tunnel contour point cloud also moves along the trajectory accordingly, so as to obtain the central point coordinates of the next tunnel cross-section and its corresponding tunnel contour point cloud after solution; then, according to the distance fluctuation between the tunnel contour point cloud and the point cloud after solution, judge whether the different tunnel cross-section profiles are deformed. Among them, if the distance between the tunnel contour point cloud and the point cloud obtained by solution fluctuates within a certain range (this range is preset) at a fixed angular position, it is judged that the tunnel has not deformed; otherwise, if it exceeds the limit value of this range, it is judged that the tunnel has deformed.

[0042] (2) Based on the lidar point cloud data, identify the ground settlement conditions and the tunnel deformation conditions in multiple periods.

[0043] (2.1) To ensure the accuracy of subsequent multi-period cross-section point cloud registration, before registration, first identify the ground settlement diseases.

[0044] Specifically, for the same rail transit tunnel line, the point cloud data of the multi-period data acquisition trajectories is subjected to ICP registration, and then the point cloud data is compared after registration. Among them, during the comparison process, if the comparison error at a certain position in the comparison errors of adjacent two periods exceeds the set fluctuation range, it is considered that there is tunnel ground deformation at this place, and the ground settlement position is identified; otherwise, if the comparison error does not exceed the set fluctuation range, it is considered that there is no tunnel ground deformation. In this way, diseases such as tunnel ground settlement can be quickly identified and can be used for tunnel disaster warning.

[0045] As another implementation method, according to the tunnel deformation situation judged in the previous step, for the tunnel area where no deformation has occurred, first overlap the point clouds of the upper contour of the tunnel in the previous and subsequent periods of this area, and then compare the point cloud data of the two cross-sections by the standard slicing method (this point cloud data can be registered by the ICP algorithm, etc.). If the distance between the point clouds of the two cross-sections in the previous and subsequent periods fluctuates within the set range, it is judged that there is no ground settlement; otherwise, if it exceeds this range limit, it is judged that ground settlement has occurred. In this way, the calculation cost of matching and comparison can be further reduced, and the recognition efficiency can be improved.

[0046] (2.2) For the multi-period point cloud data of the tunnel obtained by long-term monitoring, on the basis of the ICP algorithm registration, a standard step distance can be set according to requirements to extract cross-sections and compare them to analyze the tunnel deformation information. Further, if the equipment performance is high and there is enough time, the overall point cloud deformation of the two periods of data can be compared, and the deformation information can be visualized by coloring the deformation amount of each point.

[0047] Specifically, the method of using traditional ICP (Iterative Closest Point, precise point cloud registration) for point cloud registration is as follows: First, taking a cross-section point cloud as a reference, the distance between each point in other cross-section point clouds and the closest point in the reference point cloud is calculated through iterative calculation to determine the closest point pairs; then, by minimizing the distance error between these point pairs, a rigid body transformation matrix is calculated to align the point cloud to be registered with the reference point cloud; after multiple iterations, the cross-section point clouds of each period are accurately overlapped in three-dimensional space, thus completing the registration of multi-period point clouds.

[0048] Considering that there are still certain errors in the above traditional ICP registration, which will smooth the cumulative error to all points and it is difficult to attract key attention. Therefore, in this embodiment, when it is determined that there is no track ground deformation, based on the multi-period tunnel point cloud data, an improved ICP algorithm is used to register the multi-period tunnel cross-section point clouds to identify the multi-period tunnel deformation situation, which specifically includes the following steps:

[0049] First, obtain the point cloud data of the same tunnel cross-section in multiple periods. Through weighted processing, based on the center point of the tunnel cross-section and multiple fixed angular values, determine the crown, springing, waist, and ground intersection points of the same tunnel.

[0050] Secondly, based on the point cloud data of the determined special multi-point positions, perform the first ICP registration with relatively high requirements. Then, perform secondary ICP registration on the point cloud data of other positions and conduct local optimization under the condition of ensuring that the above special positions remain basically unchanged.

[0051] After that, for the multi-period point cloud data processed as above, use the nearest neighbor method to search for corresponding points and calculate the distance between corresponding points. In this way, the deformation value of local points can be made larger than that of the traditional ICP method, and the cumulative error will not be smoothed to all points, making it easier to attract key attention.

[0052] Finally, compare the distance between corresponding points with the set threshold to judge the deformation situation of the tunnel between two adjacent periods.

[0053] By registering the special positions first to ensure that the special points remain unchanged and then registering other points, the cumulative error can be effectively reduced, and the accuracy of the final registration can be guaranteed.

[0054] (3) According to the lidar point cloud data, combined with the suspected segment misalignment area, obtain the segment misalignment information.

[0055] Specifically, before jointly using the camera and lidar, equipment calibration is required. After calibration, there is a one-to-one correspondence between the point cloud data points and the image pixel points, and the corresponding point cloud range can be determined according to the suspected area range identified by image recognition. On this basis, combine the obtained image and point cloud data to identify the segment misalignment and obtain the segment misalignment information. Among them, segment misalignment refers to the height difference between two adjacent segments, forming a stepped shape.

[0056] First, establish a tunnel lining misalignment point cloud database. After training with pointnet++, divide the tunnel into 1m scales in the circumferential direction, with the misalignment height ranging from 1mm to 5cm and a gradient of 1mm, so as to construct a misalignment training set based on different misalignment positions and different misalignment heights.

[0057] Secondly, automatically segment the suspected segment misalignment area identified by the image and obtain the corresponding point cloud data. Compare the obtained data with the data in the misalignment training set to automatically identify the working conditions with the same misalignment position and misalignment height, and obtain the segment misalignment information.

[0058] (4) According to the lidar point cloud data, determine the water accumulation in the tunnel.

[0059] Specifically, first, based on the tunnel point cloud data, the tunnel contour is compared with the designed contour through the point cloud data, and whether there is a water accumulation area in the tunnel is judged according to the coincidence situation. Among them, considering that in the area with water accumulation, the water surface usually intersects with the ground contour, resulting in the inconsistency between the lower contour and the designed contour. Therefore, if the upper contour of the tunnel coincides but the lower contour does not coincide, it is considered that there is a water accumulation area at the non-coincident position.

[0060] Secondly, if there is a water accumulation area, the height difference between the tunnel ground contour and the designed contour is compared based on the point cloud data, and the water accumulation depth is calculated. This is based on the assumption that the missing part between the ground contour and the designed section is caused by water accumulation, so that the depth of the water accumulation can be obtained.

[0061] After that, based on the water accumulation depth and the range of the tunnel section where the water accumulation is located, the water accumulation area at the current tunnel section is calculated. In this step, within the water accumulation area, the water accumulation area of each section is calculated by geometric calculation or numerical integration. If the section shape is complex, it can be divided into several simple areas to calculate the area separately and then sum them up.

[0062] Finally, based on the water accumulation area, combined with the section step distance (i.e., the distance between two sections) between two tunnel sections, the water accumulation volume in the interval where the water accumulation area is located is calculated. Among them, for two adjacent sections, if the shape change between the water accumulation areas is linear, the trapezoidal formula or other volume calculation formulas are used to calculate the water accumulation volume (i.e., the water accumulation amount) between these two sections; if the shape is not linear, the section spacing can be locally reduced and the differential principle is used to achieve approximate linearity.

[0063] In this embodiment, through the above method, on the basis of the coincidence of the upper contour of the tunnel, the water accumulation depth can be obtained by comparing the missing height of the ground contour, and the sectional water accumulation amount can be obtained by combining the designed section information. Further, combined with the section step distance, the differential sectional water accumulation model can effectively calculate the water accumulation volume in the interval.

[0064]

[0065] Among them, V i is the water accumulation volume of the i-th section, A i and A i+1 are the water accumulation areas of the i-th section and the (i + 1)-th section respectively, and d i is the section step distance of the i-th section.

[0066] Furthermore, when calculating the water accumulation volume in the interval where the water accumulation area is located, the water accumulation amounts at multiple sections are calculated by the difference method, and the total water accumulation amount in the whole interval can be estimated. This method can provide a relatively accurate estimation of the water accumulation amount and is applicable to the situation where the water accumulation depths in different areas of the tunnel are uneven.

[0067] Accumulatively calculate the total accumulated water volume in the interval, that is: accumulate the water volume between each adjacent cross-section to obtain the total accumulated water volume of the entire tunnel section, which is:

[0068]

[0069] where n is the total number of cross-sections.

[0070] (5) Identify the installation status of rail components based on the lidar point cloud data.

[0071] Specifically, first, partition the track based on the point cloud data of the data acquisition track and establish the vault plane for each section.

[0072] Secondly, project the point cloud data of each section of the track onto the corresponding vault plane, where the track point cloud data includes the point cloud data of sleepers, bolts, rails, etc.

[0073] After that, calculate the projection distance of each component of the track onto the vault plane and the corresponding point data, and use this as the label for each component to establish a database with different track gauges, section heights, and acquisition precisions. Combine the set fluctuation threshold to identify the installation status of each rail component. Among them, for each unit section, if the distance meets the requirements but the overall number of points does not meet the requirements, identify whether there is a missing component according to the comparison of the number of points; if the distance does not meet the requirements but the overall number of points meets the requirements, determine whether the component is loose according to the distance distribution of each point in the component.

[0074] In the above method, the vault plane refers to the tangent plane passing through the survey line at the top of the tunnel. By projecting the data onto this plane, the projection distance (i.e., the normal distance) of each point of different components to this plane can be calculated. The calculation method is the distance from each point to the known plane. After that, take the average distance of each point from the upper plane of different components to the vault plane as the label for each component, and combine the set fluctuation threshold to identify the installation status of each rail component.

[0075] Step S5: Evaluate the safety of the rail transit tunnel based on the detection results of multiple diseases.

[0076] Specifically, according to the detection results of multiple diseases obtained from the above detection, compare them with the corresponding monitoring index requirements, and comprehensively evaluate the safety of the rail tunnel according to the comparison results.

[0077] Embodiment 2

[0078] This embodiment provides a rail transit tunnel disease detection system based on multi-source data characteristics, specifically including:

[0079] A data acquisition module for performing multi-phase data acquisition along a data acquisition trajectory for the same rail transit tunnel line to obtain visible light images, infrared images, and lidar point cloud data of the entire rail transit tunnel in multiple phases;

[0080] A disease detection module for identifying suspected water leakage areas and suspected segment dislocation areas in the tunnel according to visible light images using image recognition algorithms; making a temperature difference judgment based on infrared images to refine the suspected water leakage areas, mapping the identified water leakage areas to the visible light images, and then judging the development trend of water leakage according to the water leakage areas in multiple-phase visible light images; identifying the deformation conditions of different cross-section contours of the tunnel, ground settlement conditions, multi-phase tunnel deformation conditions, and the installation conditions of rail components according to the lidar point cloud data, determining the water accumulation volume in the tunnel, and combining with the suspected segment dislocation areas to obtain segment dislocation information;

[0081] A safety evaluation module for evaluating the safety of the rail transit tunnel based on the detection results of various diseases.

[0082] Embodiment III

[0083] This embodiment provides an electronic device including a memory, a processor, and computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the steps in a rail transit tunnel disease detection method based on multi-source data features as described above are completed.

[0084] Embodiment IV

[0085] This embodiment also provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by the processor, the steps in a rail transit tunnel disease detection method based on multi-source data features as described above are completed.

[0086] The steps involved in Embodiments II to IV above correspond to those in Method Embodiment I. For specific implementation manners, reference may be made to the relevant description part of Embodiment I. The term "computer-readable storage medium" should be understood to include a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode, or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0087] Those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computer device. Optionally, they can be implemented by program codes executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0088] The above are only the preferred embodiments of the present invention. Although the specific implementation manners of the present invention have been described in conjunction with the accompanying drawings, it is not a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications or deformations that can be made without creative efforts on the basis of the technical solutions of the present invention are still within the protection scope of the present invention.

Claims

1. A rail transit tunnel disease detection method based on multi-source data features, characterized in that: include: Carry out multi-period data collection along the data collection track of the same rail transit tunnel line to obtain visible light images, infrared images and lidar point cloud data of the global rail transit tunnel; Based on visible light images, image recognition algorithms are used to identify suspected water leakage areas and suspected segment misalignment areas in the tunnel; Based on the infrared image, the temperature difference is judged, the suspected water leakage area is refined, the clear water leakage area is mapped to the visible light image, and then the water leakage development trend is judged according to the water leakage area in the multi-period visible light image; Based on the LiDAR point cloud data, the deformation of different tunnel sections, ground settlement, multi-stage tunnel deformation, and rail component installation conditions are identified to determine the amount of water accumulation in the tunnel. In addition, segment misalignment information is obtained based on suspected segment misalignment areas. The safety of rail transit tunnels is evaluated based on the detection results of various diseases.

2. A rail transit tunnel disease detection method based on multi-source data features as claimed in claim 1, characterized in that: The determination of the water leakage area includes: Based on visible light images, the features of color changes, water marks, flow marks, and moss areas are extracted through the YOLO-based image recognition algorithm. The edges are combined and the area contours are divided according to the extracted features to determine the suspected water leakage area; Based on the infrared image, the temperature difference judgment is performed to refine the suspected water leakage area into the water leakage area, including: mapping the suspected water leakage area to the infrared image, dividing the suspected water leakage area into different areas, calculating the temperature mean of each suspected water leakage area, counting the areas with temperature mean higher than the normal temperature, and integrating them to obtain the water leakage area.

3. A rail transit tunnel disease detection method based on multi-source data features as claimed in claim 1, characterized in that: According to the leakage area in the multi-period visible light image, the leakage development trend is judged, including: According to the water leakage areas in the visible light images of multiple periods sorted by time series, the number of pixels corresponding to the same water leakage area in each period is calculated, and the development trend of the water leakage area is determined according to the changes in the number of pixels in multiple periods.

4. A rail transit tunnel disease detection method based on multi-source data features as claimed in claim 1, characterized in that: Based on the LiDAR point cloud data, the deformation of different section profiles of a single-phase tunnel is identified, including: For each tunnel section, based on the tunnel contour point cloud of the tunnel section, the coordinates of the center point of the tunnel contour point cloud are calculated; The center point coordinates are moved along the data acquisition track to obtain the center point coordinates of the next tunnel section and the corresponding solved tunnel contour point cloud; According to the distance fluctuation between the tunnel contour point cloud and the solved point cloud, it is judged whether the contours of different tunnel sections are deformed.

5. The rail transit tunnel disease detection method based on multi-source data features as claimed in claim 1, characterized in that: Based on the LiDAR point cloud data, the ground subsidence and multi-stage tunnel deformation are identified, including: For the same rail transit tunnel line, the point cloud data of multiple data collection trajectories are ICP-registered, and then the point cloud data are compared after registration; during the comparison process, if the comparison error at a certain position in the comparison errors of two adjacent periods exceeds the set fluctuation range, it is considered that there is tunnel ground deformation here, and the ground settlement position is identified; conversely, if the comparison error does not exceed the set fluctuation range, it is considered that there is no tunnel ground deformation; When it is determined that there is no track ground deformation, the improved ICP algorithm is used to register the multi-phase tunnel section point clouds based on the multi-phase tunnel point cloud data to identify the multi-phase tunnel deformation, including: Obtain the point cloud data of the same tunnel section in multiple phases of data, and determine the crown, spandrel, waist and intersection points with the ground of the same tunnel based on the center point of the tunnel section and multiple fixed angle values; The first ICP registration is performed based on the point cloud data of the determined multiple points, and then the point cloud data of other points are re-registered with ICP and locally optimized; The nearest neighbor method is used to search for corresponding points and calculate the distance between corresponding points. The deformation of two adjacent tunnels is determined by comparing the distance between corresponding points with the set threshold.

6. The rail transit tunnel disease detection method based on multi-source data features as claimed in claim 1, characterized in that: Determine the amount of water in the tunnel based on the LiDAR point cloud data, including: Based on the tunnel point cloud data, the tunnel contour is compared with the design contour through the point cloud data, and whether there is a water accumulation area in the tunnel is determined according to the overlap situation; among them, if the upper tunnel contour overlaps but the lower tunnel contour does not overlap, it is considered that there is a water accumulation area at the non-overlapping position; If there is a waterlogged area, the height difference between the tunnel ground contour and the designed contour is compared based on the point cloud data to calculate the waterlogged depth; Based on the depth of water accumulation and the scope of the tunnel section where the water is accumulated, the water accumulation area at the current tunnel section is calculated; Based on the waterlogged area and the step distance between the two tunnel sections, the amount of water accumulated in the area where the waterlogged area is located is calculated.

7. The rail transit tunnel disease detection method based on multi-source data features as claimed in claim 1, characterized in that: Based on the LiDAR point cloud data and the suspected segment misalignment area, segment misalignment information is obtained, including: A point cloud database of tunnel lining misalignment was established. After training with pointnet++, a misalignment training set based on different misalignment positions and different misalignment heights was constructed. Obtain the point cloud data corresponding to the suspected segment misalignment area, compare the acquired data with the data in the misalignment training set, identify the working conditions where the misalignment position and misalignment height are consistent, and obtain the segment misalignment information; Based on the LiDAR point cloud data, the installation status of rail components is identified, including: Based on the point cloud data of the data acquisition track, the track is divided into sections and the vault plane of each section is established; Projecting the track point cloud data of each section onto the corresponding vault plane; the track point cloud data includes sleepers, bolts, and rail point cloud data; Calculate the projection distance and corresponding point data of each track component onto the vault plane, and identify the installation status of each track component based on the set fluctuation threshold; Among them, for each unit section, if the distance meets the requirements but the overall number of points does not meet the requirements, the number of points is compared to identify whether the component is missing; if the distance does not meet the requirements but the overall number of points meets the requirements, the distance distribution of each point in the component is used to determine whether the component is loose.

8. A rail transit tunnel defect detection system based on multi-source data features, characterized in that: include: The data acquisition module is used to collect multiple data along the data acquisition track of the same rail transit tunnel line, and obtain visible light images, infrared images and lidar point cloud data of the global rail transit tunnel for multiple periods; The disease detection module is used to identify suspected water leakage areas and suspected segment misalignment areas in the tunnel based on visible light images using image recognition algorithms; to make temperature difference judgments based on infrared images, refine suspected water leakage areas, map clear water leakage areas to visible light images, and then judge the water leakage development trend based on the water leakage areas in multi-phase visible light images; to identify the deformation of different cross-section contours of the tunnel, ground settlement, multi-phase tunnel deformation, and rail component installation based on the lidar point cloud data, determine the amount of water accumulation in the tunnel, and obtain segment misalignment information in combination with suspected segment misalignment areas; The safety evaluation module is used to evaluate the safety of rail transit tunnels based on the detection results of various diseases.

9. An electronic device, characterized in that: The invention comprises a memory and a processor and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps of a rail transit tunnel defect detection method based on multi-source data features as described in any one of claims 1 to 7 are completed.

10. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the steps of a rail transit tunnel defect detection method based on multi-source data features as described in any one of claims 1-7.

Images