Tunnel crack detection method and tunnel crack detection system

Abstract

The application discloses a tunnel crack detection method and a tunnel crack detection system and belongs to the field of tunnel safety detection. The method comprises the following steps: a tunnel inspection robot is used to inspect in a tunnel to obtain complete three-dimensional data inside the tunnel, to construct a three-dimensional map inside the tunnel, and to determine the positions of crack regions in the three-dimensional map; the corresponding inspection regions of the crack regions are determined and path planning is performed to generate an optimal inspection path of a UAV; the optimal inspection path enables the UAV to enter the tunnel, to pass through each inspection region in turn under the condition of meeting the obstacle avoidance condition, and to leave the tunnel with the shortest total time; the UAV moves according to the optimal inspection path, and a depth camera is used to scan the crack region each time the UAV moves to an inspection region to obtain fine three-dimensional data of the crack region; the fine three-dimensional data of the crack region is analyzed and processed to complete the tunnel crack detection. The application can improve the efficiency and precision of long and large tunnel crack detection.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of tunnel safety detection, and more particularly relates to a tunnel crack detection method and a tunnel crack detection system. BACKGROUND

[0002] Tunnel crack detection is crucial to ensure the safety of tunnel structure. Traditional manual inspection is inefficient, prone to missed detection, high in detection cost and high in danger. Currently, track inspection robots are gradually replacing traditional manual inspection. By carrying high-definition cameras, laser radars and various sensors, the inspection efficiency and safety are improved. However, the track inspection robot is limited in coverage and detection accuracy when moving and detecting on the fixed track. At the same time, the track inspection robot may find problems during inspection, and manual re-inspection is still needed to ensure accuracy and integrity. However, for some cracks that need multi-angle observation, defects located at complex structures such as curves or intersections, and other complex or more detailed evaluation situations.

[0003] The key to tunnel crack detection is to accurately obtain three-dimensional data of the crack area. Due to its high flexibility and strong maneuverability, a UAV can carry various sensors such as high-definition cameras and infrared sensors into the tunnel, making it convenient and fast to reach places that are difficult for humans and tunnel inspection robots to reach for detection. The inspection range is wider, and compared with the tunnel inspection robot, the three-dimensional data obtained by the UAV inspection is more accurate, which is more conducive to accurate detection of tunnel cracks.

[0004] Due to the above advantages, UAVs have been increasingly widely used in tunnel inspection. However, with the continuous development of transportation, current tunnels are mostly long tunnels. Long tunnels have a large length, and as they are used continuously, multiple cracks often occur in the same tunnel, which makes the UAV face the problem of short endurance time. Batteries need to be frequently replaced or charged, which is low in inspection efficiency. This will affect the normal and safe use of the tunnel. In addition, the space in a long tunnel is closed, and the lighting conditions are poor, which will affect the flight stability of the UAV and the quality of three-dimensional data acquisition, reducing the detection accuracy. SUMMARY

[0005] In view of the defects and improvement needs of the prior art, the present application provides a tunnel crack detection method and a tunnel crack detection system, which aims to improve the efficiency and accuracy of long tunnel crack detection.

[0006] To achieve the above purpose, according to one aspect of the present application, a tunnel crack detection method is provided, comprising:

[0007] using a tunnel inspection robot to start inspection in the tunnel from the tunnel entrance to obtain complete three-dimensional data inside the tunnel;

[0008] construct a three-dimensional map of the tunnel interior using the complete three-dimensional data of the tunnel interior, detect crack regions in the tunnel, and determine the positions of the crack regions in the three-dimensional map;

[0009] determine inspection regions corresponding to the crack regions, wherein the inspection regions satisfy that when a UAV carrying a depth camera is located in an inspection region, the crack region is located within the field of view of the depth camera;

[0010] perform path planning according to the three-dimensional map and the positions of the inspection regions in the three-dimensional map, and generate an optimal inspection path of the UAV, wherein the optimal inspection path causes the UAV to sequentially pass through each inspection region in the tunnel after entering the tunnel from the tunnel entrance under the condition of satisfying an obstacle avoidance condition, and then exit the tunnel from the tunnel exit, and the total time of the UAV from entering the tunnel to exiting the tunnel is the shortest;

[0011] cause the UAV to move according to the optimal inspection path, and each time the UAV moves to an inspection region, the crack region is scanned using the depth camera to obtain fine three-dimensional data of the crack region;

[0012] perform analysis and processing on the fine three-dimensional data of the crack region to complete tunnel crack detection;

[0013] The analysis and processing includes at least one of the following: three-dimensional reconstruction of the crack region, identification of the crack type, extraction of the three-dimensional shape of the crack, assessment of the crack danger level, and prediction of the crack development trend.

[0014] Further, before the tunnel inspection robot starts to inspect in the tunnel from the tunnel entrance, the method further includes: obtaining a relative pose between the tunnel inspection robot and the tunnel entrance at the tunnel entrance, denoted as an entrance reference relative pose;

[0015] And, before causing the UAV to move according to the optimal inspection path, the method further includes: obtaining a relative pose between the UAV and the tunnel entrance at the tunnel entrance, denoted as an entrance alignment relative pose, and aligning the entrance alignment relative pose to the entrance reference relative pose, so that the tunnel inspection robot and the UAV share a unified three-dimensional map coordinate system.

[0016] Further, the tunnel crack detection method provided by the present application further includes: pre-setting a visual marker at the tunnel entrance;

[0017] And, the entrance reference relative pose and the entrance alignment relative pose are both determined by means of the visual marker.

[0018] Further, before determining the inspection regions corresponding to the crack regions, the method further includes: determining a visual field blind area of the tunnel inspection robot in the tunnel according to the three-dimensional map, and a position of the visual field blind area in the three-dimensional map;

[0019] The path planning is performed according to the three-dimensional map and the positions of the visual field blind areas in the three-dimensional map, and a supplementary inspection path of the unmanned aerial vehicle is generated; the supplementary inspection path is such that, after the unmanned aerial vehicle enters the tunnel from the tunnel entrance, the unmanned aerial vehicle sequentially passes through each visual field blind area under the condition that an obstacle avoidance condition is met, and then exits the tunnel from the tunnel exit, and the total time of the unmanned aerial vehicle from entering the tunnel to exiting the tunnel is the shortest;

[0020] The unmanned aerial vehicle moves according to the supplementary inspection path, and each time the unmanned aerial vehicle moves to a visual field blind area, the unmanned aerial vehicle scans the visual field blind area by using the depth camera to obtain fine three-dimensional data of the visual field blind area as supplementary depth information of the visual field blind area;

[0021] The three-dimensional map is updated by using the supplementary depth information of the visual field blind area.

[0022] According to still another aspect of the present application, a tunnel crack detection system is provided, comprising: a crack preliminary positioning module, a crack accurate scanning module, a communication module and a cloud computing server;

[0023] The crack preliminary positioning module comprises: a tunnel inspection robot and a mobile edge computing server carried on the tunnel inspection robot;

[0024] The tunnel inspection robot is used for inspecting in the tunnel to obtain complete three-dimensional data of the inside of the tunnel, and sends the complete three-dimensional data to the edge computing server;

[0025] The mobile edge computing server is used for constructing a three-dimensional map of the inside of the tunnel by using the complete three-dimensional data of the inside of the tunnel, detecting crack areas in the tunnel, determining positions of the crack areas in the three-dimensional map, and sending the three-dimensional map and the positions of the crack areas in the three-dimensional map to the crack accurate scanning module;

[0026] The crack accurate scanning module comprises an unmanned aerial vehicle carrying a depth camera and a pose calculation unit carried on the unmanned aerial vehicle;

[0027] The unmanned aerial vehicle is used for scanning a target in a field of view range of the depth camera to obtain fine three-dimensional data of the target;

[0028] The pose calculation unit is used for determining an inspection area corresponding to each crack area; the inspection area satisfies that when the unmanned aerial vehicle is located in the inspection area, the crack area is located in the field of view range of the depth camera;

[0029] The pose calculation unit is further used for performing path planning according to the three-dimensional map and the positions of the inspection areas in the three-dimensional map, and generating an optimal inspection path of the unmanned aerial vehicle; the optimal inspection path is such that, after the unmanned aerial vehicle enters the tunnel from the tunnel entrance, the unmanned aerial vehicle sequentially passes through each inspection area in the tunnel under the condition that an obstacle avoidance condition is met, and then exits the tunnel from the tunnel exit, and the total time of the unmanned aerial vehicle from entering the tunnel to exiting the tunnel is the shortest;

[0030] The pose calculation unit is also configured to control the UAV to move along the optimal inspection path, and to control the UAV to scan the crack region by using the depth camera to obtain fine three-dimensional data of the crack region and send the fine three-dimensional data to the cloud computing server each time the UAV moves to an inspection region;

[0031] The cloud computing server is located outside the tunnel and is configured to perform analysis and processing on the fine three-dimensional data of the crack region to complete the tunnel crack detection, and the analysis and processing includes at least one of the following: three-dimensional reconstruction of the crack region, identification of the crack type, extraction of the three-dimensional shape of the crack, assessment of the crack risk level, and prediction of the crack development trend.

[0032] The communication module is configured to realize communication between the crack preliminary positioning module and the crack accurate scanning module, and communication between the crack accurate scanning module and the cloud computing server.

[0033] Further, the communication module includes a wireless access point arranged in the tunnel, and WIFI communication modules arranged in the crack preliminary positioning module and the crack accurate scanning module.

[0034] In addition, the communication between the crack preliminary positioning module and the crack accurate scanning module is realized by the wireless access point and the WIFI communication modules.

[0035] Further, the communication module further includes a 5G communication module arranged in the crack accurate scanning module and the cloud service calculator.

[0036] In addition, the communication between the crack accurate scanning module and the cloud service calculator is realized by the 5G communication module.

[0037] Further, the tunnel inspection robot is also provided with an on-vehicle positioning sensor, and the UAV is also provided with an on-board positioning sensor.

[0038] The on-vehicle positioning sensor is configured to obtain a relative pose between the tunnel inspection robot and the tunnel entrance at the tunnel entrance, denoted as an entrance reference relative pose, and send the entrance reference relative pose to the crack accurate scanning module.

[0039] In addition, the on-board positioning sensor is configured to obtain a relative pose between the UAV and the tunnel entrance at the tunnel entrance before the UAV moves along the optimal inspection path, denoted as an entrance alignment relative pose, and align the entrance alignment relative pose to the entrance reference relative pose.

[0040] Further, the tunnel crack detection system provided by the application further includes a visual marker arranged at the tunnel entrance.

[0041] In addition, the entrance reference relative pose and the entrance alignment relative pose are determined by means of the visual marker.

[0042] Overall, the above technical solutions conceived by the present application can achieve the following beneficial effects:

[0043] (1) The present application uses a tunnel inspection robot to obtain three-dimensional data inside the tunnel to establish a three-dimensional map inside the tunnel and preliminarily detect all crack areas inside the tunnel and the positions of the crack areas in the three-dimensional map, and accordingly plans an optimal inspection path for the unmanned aerial vehicle, so that the unmanned aerial vehicle moves to the vicinity of the crack areas in turn according to the optimal inspection path for precise scanning. Since the unmanned aerial vehicle only needs to scan the crack areas during the inspection process, and the path planning ensures the shortest inspection time, the unmanned aerial vehicle can quickly collect precise three-dimensional data of all crack areas using the depth camera mounted thereon, with high data quality and high inspection efficiency without frequent battery replacement. When planning the inspection path for the unmanned aerial vehicle, the present application considers the obstacle avoidance problem, thereby improving the flight stability of the unmanned aerial vehicle in the tunnel. Overall, the present application uses the tunnel inspection robot and the unmanned aerial vehicle to work cooperatively, thereby effectively improving the efficiency and accuracy of crack detection in the tunnel, especially in long and large tunnels.

[0044] (2) In the preferred scheme of the present application, when the unmanned aerial vehicle moves according to the optimal inspection path, its relative pose with respect to the tunnel entrance (entrance alignment relative pose) is first aligned to the relative pose of the tunnel inspection robot with respect to the tunnel entrance (entrance reference relative pose), so that the unmanned aerial vehicle can be precisely positioned and navigated in the three-dimensional map constructed by the tunnel inspection robot, thereby improving the control accuracy of the unmanned aerial vehicle's pose during the inspection process and ensuring that the fine three-dimensional data collected by the unmanned aerial vehicle can be accurately registered and fused with the three-dimensional map constructed by the tunnel inspection robot. In the further preferred scheme, the relative pose of the unmanned aerial vehicle / tunnel inspection robot with respect to the tunnel entrance is determined by means of visual markers, so that the relative pose can be quickly and accurately determined.

[0045] (3) In the preferred scheme of the present application, after establishing the three-dimensional map of the tunnel based on the three-dimensional data obtained by the tunnel inspection robot, the visual blind area of the tunnel inspection robot inside the tunnel is further determined, and a supplementary inspection path is planned for the visual blind area. The unmanned aerial vehicle collects precise three-dimensional data of the visual blind area as supplementary depth information for the visual blind area, and updates and improves the three-dimensional map of the tunnel using the supplementary depth information, and then generates an optimal inspection path for the unmanned aerial vehicle based on the updated and improved three-dimensional map. The optimal inspection path more comprehensively utilizes the three-dimensional information inside the tunnel, thereby further improving the inspection efficiency and flight stability of the unmanned aerial vehicle.

[0046] (4) The tunnel crack inspection system provided by the application adopts a collaborative architecture of a cloud computing server and an edge computing server, the edge computing server is used to complete relatively simple calculations such as three-dimensional map establishment and preliminary tunnel crack detection, complex calculations for analyzing and processing accurate three-dimensional data are completed by the cloud computing server, thereby a large amount of data transmission is reduced, the bandwidth utilization rate is optimized, the real-time response capability and flexibility of the system are ensured, powerful cloud analysis capability is provided, the system can quickly identify potential risks and perform in-depth evaluation, and the data processing efficiency and intelligent level of the entire detection process are greatly improved. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 A tunnel crack detection method flowchart is provided for the embodiment of the application.

[0048] Figure 2 A function block diagram of a tunnel crack detection system is provided for the embodiment of the application.

[0049] Figure 3 A specific structure schematic diagram of a tunnel crack detection system is provided for the embodiment of the application. DETAILED DESCRIPTION

[0050] In order to make the objectives, technical solutions and advantages of the application clearer, the application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the application and should not be used to limit the application. In addition, the technical features involved in each embodiment of the application described below can be combined with each other as long as there is no conflict.

[0051] In the application, the terms "first", "second", etc. (if any) in the application and the drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence.

[0052] In order to solve the technical problem that the existing unmanned aerial vehicle inspection method cannot efficiently and accurately realize crack detection of long and large tunnels, the application provides a tunnel crack detection method and a tunnel crack detection system, and the overall concept is that, in the case of collecting three-dimensional data of a crack area by using an unmanned aerial vehicle carrying a depth camera to ensure data collection accuracy, a pre-planned inspection path is used to make the unmanned aerial vehicle not need to scan other areas outside the crack area during the inspection process to improve the inspection efficiency, solve the endurance problem, and realize obstacle avoidance and improve flight stability.

[0053] Based on the above concept, the whole inspection process is divided into two stages: the first stage, the tunnel inspection robot collects three-dimensional data in the tunnel, which is used to establish a three-dimensional map of the tunnel and preliminarily locate the positions of all crack areas in the three-dimensional map, providing a basis for the planning of the UAV inspection path; the second stage, based on the three-dimensional map of the tunnel and the positions of the crack areas in the three-dimensional map, the path planning is carried out to make the UAV inspection time shortest and realize obstacle avoidance, to generate the optimal inspection path for the UAV, and make the UAV collect the accurate three-dimensional data of each crack area in turn according to the optimal inspection path for subsequent analysis and processing.

[0054] The following is an example.

[0055] Example 1:

[0056] A tunnel crack detection method, as shown in Figure 1 , comprising:

[0057] The tunnel inspection robot starts from the tunnel entrance to inspect in the tunnel to obtain complete three-dimensional data inside the tunnel;

[0058] The three-dimensional map inside the tunnel is constructed by using the complete three-dimensional data inside the tunnel, the crack areas in the tunnel are detected, and the positions of the crack areas in the three-dimensional map are determined;

[0059] The inspection area corresponding to each crack area is determined; the inspection area satisfies that when the UAV carrying a depth camera is located in the inspection area, the crack area is located in the field of view range of the depth camera;

[0060] According to the three-dimensional map and the positions of the inspection areas in the three-dimensional map, path planning is performed to generate the optimal inspection path of the UAV; the optimal inspection path makes the UAV enter the tunnel from the tunnel entrance, pass through each inspection area in the tunnel in turn under the condition of meeting the obstacle avoidance condition, and then exit the tunnel through the tunnel exit, and the total time of the UAV from entering the tunnel to exiting the tunnel is the shortest;

[0061] The UAV moves according to the optimal inspection path, and when moving to an inspection area, the depth camera is used to scan the crack area to obtain the fine three-dimensional data of the crack area;

[0062] The fine three-dimensional data of the crack area is analyzed and processed to complete the tunnel crack detection.

[0063] In order to improve the control accuracy of the pose of the UAV in the inspection process, as an optional embodiment, before the tunnel inspection robot starts to inspect in the tunnel from the tunnel entrance, the relative pose of the tunnel inspection robot at the tunnel entrance and the tunnel entrance is obtained, which is recorded as the entrance reference relative pose.

[0064] And, before making the UAV move along the optimal inspection path, the method further comprises: obtaining a relative pose of the UAV at the tunnel entrance and the tunnel entrance, denoted as an entrance alignment relative pose, and aligning the entrance alignment relative pose to the entrance reference relative pose, so that the tunnel inspection robot and the UAV share a unified three-dimensional map coordinate system.

[0065] In order to facilitate the acquisition of the relative pose of the UAV / tunnel inspection robot relative to the tunnel entrance, the embodiment pre-provides an ArUco marker at the tunnel entrance; the ArUco marker is easy to quickly identify, has high robustness and low sensitivity to light and occlusion, and is suitable for accurate positioning in dynamic environments such as UAVs / tunnel inspection robots. The UAV / tunnel inspection robot often carries a positioning sensor, which uses the Harris algorithm to quickly locate image feature points, uses the SIFT algorithm to extract feature point descriptors, ensures the scale and rotation invariance of feature recognition, uses the FLANN algorithm for efficient feature matching, uses the RANSAC algorithm to filter false matches, ensures the recognition accuracy, uses the depth information and the matching result to perform PnP algorithm triangulation to obtain the pose of the ArUco marker in the coordinate system of the positioning sensor. For the UAV, the positioning sensor can be an onboard positioning sensor carried thereon, and for the tunnel inspection robot, the positioning sensor can be a vehicle-mounted positioning sensor carried thereon.

[0066] It should be noted that using the ArUco marker to determine the relative pose is only one optional implementation, and should not be understood as the only limitation of the present application. In some other embodiments of the present application, other visual markers such as AprilTag can also be used. In some other embodiments of the present application, visual markers can not be used, and the determination of the relative pose can be directly realized based on coordinate transformation and the like.

[0067] Optionally, in the present embodiment, the three-dimensional map is constructed by the Gmapping algorithm; and the preliminary detection of the crack region in the tunnel is detected by a pre-trained detection model. It should be noted that the method for constructing the three-dimensional map and the preliminary detection of the crack region described herein is only exemplary and should not be understood as the only limitation of the present application.

[0068] In actual applications, according to specific detection requirements, the analysis and processing performed on the fine three-dimensional data of the crack region include at least one of the following:

[0069] The data is processed by three-dimensional reconstruction technology to identify the general structure and the approximate position of the cracks, and to realize preliminary three-dimensional reconstruction;

[0070] After completing the preliminary reconstruction, data cleaning is performed to remove noise, and then feature enhancement technology is used to highlight the crack features and improve their visibility in the model.

[0071] By using three-dimensional segmentation technology, the cracks are accurately distinguished from other surface features to accurately segment the cracks and realize crack identification.

[0072] After identifying the cracks, a deep learning algorithm is used to extract the three-dimensional shape and parameters of the cracks, including length, width, and depth, etc.

[0073] Based on the extracted three-dimensional shape and parameters of the cracks, the crack type is identified, and detailed three-dimensional reconstruction is performed to comprehensively analyze the morphology and distribution of the cracks.

[0074] In-depth analysis is performed on the crack data to evaluate the crack risk level, predict the crack development trend, evaluate the repair demand, and simulate the potential impact of the cracks on the structural performance to complete the scientific support for structural health detection and maintenance decision-making.

[0075] Overall, the tunnel crack detection method provided by the embodiment overcomes the limitations of traditional single detection methods through the air-ground cooperation of the tunnel inspection robot and the unmanned aerial vehicle. The tunnel inspection robot is responsible for building a three-dimensional tunnel map and preliminarily positioning the crack area, and the unmanned aerial vehicle performs close-range accurate scanning, achieving comprehensive and efficient tunnel detection. This air-ground cooperation mode significantly improves the coverage and accuracy of detection, effectively solves the blind area problem in traditional tunnel detection, and can more comprehensively detect and identify potential risks in the tunnel structure.

[0076] The embodiment uses an unmanned aerial vehicle equipped with a depth camera to collect three-dimensional data of the crack area, combining the maneuverability and flexibility of the unmanned aerial vehicle, which can perform high-precision close-range scanning of the crack area, improving the image acquisition quality. This method not only improves the accuracy of crack detection, but also obtains the three-dimensional geometric features of the cracks, providing rich data support for subsequent crack analysis and evaluation, greatly improving the accuracy of tunnel structural health condition evaluation.

[0077] The invention adopts the detection strategy of "preliminary positioning + accurate scanning", effectively solving the problem of short endurance of the unmanned aerial vehicle in long tunnels. During the inspection process, the unmanned aerial vehicle only needs to generate an optimal inspection path based on the pose of the unmanned aerial vehicle in the three-dimensional map and the position of the crack area in the three-dimensional map to perform close-range accurate scanning of the potential crack area, without comprehensive detailed investigation, greatly reducing the workload and flight time of the unmanned aerial vehicle, and improving the flight stability. This strategy not only improves the detection efficiency, but also significantly prolongs the effective working time of the unmanned aerial vehicle, enabling it to play a greater role in long tunnels.

[0078] Embodiment 2:

[0079] A tunnel crack detection method. The embodiment is similar to the above-mentioned embodiment 1, the difference is that, in the process of tunnel inspection robot inspection, due to the existence of visual blind area, the three-dimensional data collected may be missing, resulting in that the three-dimensional map constructed therefrom cannot accurately reflect the three-dimensional environment in the tunnel, so after the preliminary establishment of the three-dimensional map of the tunnel and the preliminary positioning of the crack area, before determining the corresponding inspection area of each crack area, it further includes:

[0080] According to the three-dimensional map, the visual blind area of the tunnel inspection robot in the tunnel is determined, and the position of the visual blind area in the three-dimensional map is determined;

[0081] According to the three-dimensional map and the position of each visual blind area in the three-dimensional map, path planning is performed to generate a supplementary inspection path of the unmanned aerial vehicle; the supplementary inspection path makes the unmanned aerial vehicle enter the tunnel from the tunnel entrance, and then pass through each visual blind area in turn under the condition of meeting the obstacle avoidance condition, and then leave the tunnel through the tunnel exit, and the total time of the unmanned aerial vehicle from entering the tunnel to leaving the tunnel is the shortest;

[0082] Make the unmanned aerial vehicle move according to the supplementary inspection path, and make the unmanned aerial vehicle scan the visual blind area by using the depth camera when the unmanned aerial vehicle moves to each visual blind area, so as to obtain the fine three-dimensional data of the visual blind area as the supplementary depth information of the visual blind area;

[0083] Update the three-dimensional map by using the supplementary depth information of the visual blind area.

[0084] After updating the three-dimensional map by using the supplementary depth information, the corresponding inspection area of each crack area is determined, and path planning is performed according to the updated three-dimensional map and the position of the inspection area in the three-dimensional map.

[0085] The visual blind area of the tunnel inspection robot, that is, due to the limitation of the field of view angle of the sensor, the obstacle shielding and other factors, the sensor carried by the tunnel inspection robot cannot completely cover or is difficult to obtain accurate data. By analyzing the point cloud data of the three-dimensional map preliminarily constructed by the tunnel inspection robot, these visual blind areas can be identified and located. Specifically, in the point cloud data analysis, the visual blind area of the tunnel inspection robot usually has the following characteristics: point cloud data missing, poor point cloud data quality or point cloud density significantly lower than the surrounding area.

[0086] In the embodiment, the remaining steps are the same as those of the above-mentioned embodiment 1, and the specific implementation manner can refer to the description in the above-mentioned embodiment 1.

[0087] The embodiment collects precise three-dimensional data of the visual blind area by the unmanned aerial vehicle as supplementary depth information of the visual blind area, updates and improves the three-dimensional map of the tunnel by using the supplementary depth information, and generates an optimal inspection path of the unmanned aerial vehicle based on the updated and improved three-dimensional map. The optimal inspection path more comprehensively utilizes the three-dimensional information in the tunnel, and can further improve the inspection efficiency and flight stability of the unmanned aerial vehicle.

[0088] Embodiment 3

[0089] A tunnel crack detection system, as shown in Figure 2 includes a crack preliminary positioning module, a crack precise scanning module, a communication module and a cloud computing server.

[0090] The crack preliminary positioning module includes a tunnel inspection robot and a mobile edge computing server carried on the tunnel inspection robot.

[0091] The tunnel inspection robot is used for inspection in the tunnel to obtain complete three-dimensional data inside the tunnel, and sends the data to the edge computing server.

[0092] The mobile edge computing server is used for constructing a three-dimensional map inside the tunnel by using the complete three-dimensional data inside the tunnel, detecting crack regions in the tunnel, determining positions of the crack regions in the three-dimensional map, and sending the three-dimensional map and the positions of the crack regions in the three-dimensional map to the crack precise scanning module.

[0093] The crack precise scanning module includes an unmanned aerial vehicle carrying a depth camera and a pose calculation unit carried on the unmanned aerial vehicle.

[0094] The unmanned aerial vehicle is used for scanning a target in a field of view of the depth camera to obtain fine three-dimensional data of the target.

[0095] The pose calculation unit is used for determining an inspection region corresponding to each crack region, wherein the inspection region satisfies that the crack region is located in the field of view of the depth camera when the unmanned aerial vehicle is located in the inspection region.

[0096] The pose calculation unit is further used for performing path planning according to the three-dimensional map and the positions of the inspection regions in the three-dimensional map to generate an optimal inspection path of the unmanned aerial vehicle. The optimal inspection path makes the unmanned aerial vehicle pass through each inspection region in the tunnel in turn after entering the tunnel from the tunnel entrance under the condition of satisfying the obstacle avoidance condition, and then leaves the tunnel through the tunnel exit, and the total time of the unmanned aerial vehicle from entering the tunnel to leaving the tunnel is the shortest.

[0097] The pose calculation unit is further used for making the unmanned aerial vehicle move according to the optimal inspection path, and making the unmanned aerial vehicle scan the crack region by using the depth camera when the unmanned aerial vehicle moves to each inspection region, to obtain fine three-dimensional data of the crack region, and send the data to the cloud computing server.

[0098] The cloud computing server is located outside the tunnel and is configured to perform analysis processing on the fine three-dimensional data of the crack area to complete the crack detection of the tunnel.

[0099] The communication module is configured to realize communication between the crack preliminary positioning module and the crack accurate scanning module, and communication between the crack accurate scanning module and the cloud computing server.

[0100] As shown in Figure 3 the communication module includes a wireless access point, a WIFI communication module and a 5G communication module. The wireless access point is arranged at intervals inside the tunnel to realize WiFi network coverage in the tunnel; the WIFI communication module is arranged on the tunnel inspection robot and the unmanned aerial vehicle respectively; the tunnel inspection robot communicates with the unmanned aerial vehicle through the WIFI network; the 5G communication module is carried on the ground terminal device; the 5G network is composed of a 5G communication base station and a 5G router; the cloud computing server communicates with the ground terminal device through the 5G network; the ground terminal device communicates with the unmanned aerial vehicle through the WIFI network; and the unmanned aerial vehicle communicates with the cloud computing server through the WIFI network and the 5G network.

[0101] The embodiment also includes a visual marker arranged at the tunnel entrance to facilitate determination of the relative pose of the tunnel inspection robot and the unmanned aerial vehicle relative to the tunnel entrance, thereby realizing alignment of the initial relative pose of the inspection. Optionally, in the embodiment, the visual marker arranged at the tunnel entrance is specifically an ArUco marker, and it should be noted that in some other embodiments of the application, other visual markers such as AprilTag can also be used.

[0102] Optionally, in the embodiment, the tunnel inspection robot is internally carried with a three-dimensional laser radar, an inertial measurement unit, an odometer, a vehicle-mounted positioning sensor and a vehicle-mounted main controller; the three-dimensional laser radar is configured to scan the basic three-dimensional data of the tunnel in real time; the inertial measurement unit is configured to obtain the roll angle, the pitch angle and the heading angle of the rotation of the inspection robot relative to the horizontal plane in real time; the odometer is configured to measure the mileage of the tunnel inspection robot; the vehicle-mounted positioning sensor is configured to detect the visual marker at the tunnel entrance to obtain the relative pose of the visual marker at the tunnel entrance and the tunnel inspection robot, which is recorded as the relative pose at the entrance reference; and the vehicle-mounted main controller is configured to construct a three-dimensional map inside the tunnel by processing and fusing the data of the laser radar sensor, the inertial measurement unit and the odometer, and to obtain the pose of the tunnel inspection robot in the map coordinate system of the three-dimensional map in real time.

[0103] Optionally, in the embodiment, the unmanned aerial vehicle is further provided with an on-board inertial measurement unit, an on-board positioning sensor and an on-board main controller in addition to the depth camera, and comprises a multi-rotor frame; the multi-rotor frame is used to carry the depth camera, the on-board inertial measurement unit, the on-board positioning sensor and the on-board main controller, and provides the capability of vertical take-off and landing, hovering and rapid maneuvering in a narrow or complex environment; the depth camera is used to capture high-resolution depth images of the tunnel surface and structure to obtain accurate three-dimensional spatial information, supplement the three-dimensional map constructed by the tunnel inspection robot and obtain three-dimensional data of the crack area; the on-board inertial measurement unit is used to measure the angular velocity and acceleration of the corresponding coordinate axes in the three-dimensional space; the on-board positioning sensor is used to detect the visual marker at the tunnel entrance to obtain the relative pose of the visual marker at the tunnel entrance and the unmanned aerial vehicle, denoted as the entrance alignment relative pose; the on-board main controller is used to receive the positioning information and control the unmanned aerial vehicle to complete the corresponding action; and the pose calculation unit is used to obtain the pose of the unmanned aerial vehicle in the three-dimensional map in combination with the entrance reference relative pose, the entrance alignment relative pose and the pose of the tunnel inspection robot in the three-dimensional map.

[0104] Optionally, in the embodiment, the unmanned aerial vehicle is a hexacopter. In other embodiments of the application, other types of unmanned aerial vehicles can also be used. The depth camera carried by the unmanned aerial vehicle can be a ToF (Time of Flight) camera. The ToF camera measures the distance between the target object and the camera using the time-of-flight principle, generates depth images and three-dimensional point cloud data by emitting light waves and measuring the time required for their reflection, has high detection accuracy, accurate depth map and is not limited by environmental lighting conditions, and can work normally even in a poorly lit tunnel environment. During the inspection process, the unmanned aerial vehicle moves to each inspection area and performs high-precision close-range scanning of the tunnel surface of the crack area by the ToF camera, and collects three-dimensional data of the crack area. The three-dimensional data describes the geometric characteristics of the crack, such as the size, shape and spatial distribution of the crack.

[0105] Optionally, in the embodiment, after the unmanned aerial vehicle collects the accurate three-dimensional data of the crack area, it packages the three-dimensional data of the crack area and the pose of the corresponding crack area in the map coordinate system of the tunnel three-dimensional map together and marks the time, obtains the detection data packet of the crack area, and then transmits the detection data packet to the cloud computing server in real time through the 5G network for further analysis and processing.

[0106] In this embodiment, after the cloud service computer receives the detection data packet of the crack area uploaded by the unmanned aerial vehicle, it extracts the fine three-dimensional data of the crack area and performs analysis processing to complete the tunnel crack detection. In actual application, according to specific detection requirements, the analysis processing performed by the cloud service computer on the fine three-dimensional data of the crack area includes at least one of the following: processing the data through three-dimensional reconstruction technology to identify the general structure and the approximate position of the crack, and realize preliminary three-dimensional reconstruction;

[0107] After completing the preliminary reconstruction, data cleaning is performed to remove noise, and then feature enhancement technology is used to highlight the crack features and improve their visibility in the model;

[0108] Three-dimensional segmentation technology is used to accurately distinguish cracks from other surface features to accurately segment the cracks and realize crack identification;

[0109] After identifying the cracks, a deep learning algorithm is used to extract the three-dimensional shape and parameters of the cracks, including length, width, and depth, etc.

[0110] Based on the extracted three-dimensional shape and parameters of the cracks, the crack type is identified, and detailed three-dimensional reconstruction is performed to comprehensively analyze the morphology and distribution of the cracks.

[0111] In-depth analysis is performed on the crack data to evaluate the risk level of the cracks, predict the development trend of the cracks, evaluate the repair needs, and simulate the potential impact of the cracks on the structural performance to complete the scientific support for structural health detection and maintenance decision-making.

[0112] Those skilled in the art will readily understand that the above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for detecting tunnel cracks, characterized in that, include: A tunnel inspection robot is used to inspect the tunnel from the tunnel entrance to obtain complete three-dimensional data of the tunnel interior. A three-dimensional map of the tunnel interior is constructed using complete three-dimensional data of the tunnel interior, crack areas within the tunnel are detected, and the location of each crack area in the three-dimensional map is determined. Determine the inspection area corresponding to each crack area; The inspection area satisfies the following condition: when a drone equipped with a depth camera is located in the inspection area, the crack area is within the field of view of the depth camera; Based on the 3D map and the location of each inspection area in the 3D map, path planning is performed to generate the optimal inspection path for the UAV. The optimal inspection path ensures that after the UAV enters the tunnel from the tunnel entrance, it passes through each inspection area in the tunnel in sequence while meeting obstacle avoidance conditions, and then leaves the tunnel through the tunnel exit, with the shortest total time from entering the tunnel to leaving the tunnel. The drone moves along the optimal inspection path, and each time it moves to an inspection area, it uses the depth camera to scan the crack area to obtain detailed three-dimensional data of the crack area; the drone only scans the crack area during the inspection process. Analyze and process detailed three-dimensional data of the crack area to complete tunnel crack detection; The analysis and processing includes at least one of the following: performing three-dimensional reconstruction of the crack area, identifying the crack type, extracting the three-dimensional shape of the crack, assessing the degree of crack hazard, and predicting the crack development trend.

2. The tunnel crack detection method as described in claim 1, characterized in that, Before using the tunnel inspection robot to start inspecting inside the tunnel from the tunnel entrance, the process also includes: acquiring the relative pose of the tunnel inspection robot and the tunnel entrance at the tunnel entrance, which is recorded as the entrance reference relative pose. Furthermore, before the drone moves along the optimal inspection path, the process further includes: acquiring the relative pose of the drone and the tunnel entrance at the tunnel entrance, denoted as the entrance alignment relative pose, and aligning the entrance alignment relative pose to the entrance reference relative pose, so that the tunnel inspection robot and the drone share a unified three-dimensional map coordinate system.

3. The tunnel crack detection method as described in claim 2, characterized in that, Also includes: Visual markers were placed at the tunnel entrance in advance; Furthermore, both the relative pose of the entrance reference and the relative pose of the entrance alignment are determined using the visual markers.

4. The tunnel crack detection method according to any one of claims 1 to 3, characterized in that, Before determining the inspection area corresponding to each crack area, the method further includes: determining the blind spot of the tunnel inspection robot in the tunnel based on the three-dimensional map, and the location of the blind spot in the three-dimensional map; Based on the 3D map and the location of each blind spot in the 3D map, path planning is performed to generate a supplementary inspection path for the UAV. The supplementary inspection path enables the UAV to enter the tunnel from the tunnel entrance, pass through each blind spot in sequence while meeting obstacle avoidance conditions, and then leave the tunnel through the tunnel exit, with the total time from entering the tunnel to leaving the tunnel being the shortest. The drone is made to move along the supplementary inspection path, and each time the drone moves to a blind spot, the depth camera is used to scan the blind spot to obtain detailed three-dimensional data of the blind spot as supplementary depth information of the blind spot. The 3D map is updated using supplementary depth information from blind spots.

5. A tunnel crack detection system, characterized in that, include: The system includes a preliminary crack location module, a precise crack scanning module, a communication module, and a cloud computing server. The crack preliminary location module includes: a tunnel inspection robot and a mobile edge computing server mounted on the tunnel inspection robot; The tunnel inspection robot is used to inspect inside the tunnel to obtain complete three-dimensional data of the tunnel interior and send it to the edge computing server. The mobile edge computing server is used to construct a three-dimensional map of the tunnel using complete three-dimensional data inside the tunnel, detect crack areas inside the tunnel, determine the location of each crack area in the three-dimensional map, and send the three-dimensional map and the location of each crack area in the three-dimensional map to the crack precision scanning module. The crack precision scanning module includes a drone equipped with a depth camera and a pose calculation unit mounted on the drone. The drone is used to scan targets within the field of view of its depth camera to obtain detailed three-dimensional data of the targets; The pose calculation unit is used to determine the inspection area corresponding to each crack area; the inspection area satisfies the following condition: when the UAV is located in the inspection area, the crack area is within the field of view of the depth camera; The pose calculation unit is also used to perform path planning based on the three-dimensional map and the position of each inspection area in the three-dimensional map to generate the optimal inspection path of the UAV. The optimal inspection path enables the UAV to enter the tunnel from the tunnel entrance, pass through each inspection area in the tunnel in sequence under the condition of meeting obstacle avoidance, and then leave the tunnel through the tunnel exit, and the total time of the UAV from entering the tunnel to leaving the tunnel is the shortest. The pose calculation unit is also used to make the UAV move according to the optimal inspection path, and to make the UAV scan the crack area with the depth camera every time it moves to an inspection area to obtain detailed three-dimensional data of the crack area and send it to the cloud computing server; the UAV only scans the crack area during the inspection process. The cloud computing server is located outside the tunnel and is used to perform detailed three-dimensional data analysis and processing on the crack area to complete tunnel crack detection. The analysis and processing includes at least one of the following: three-dimensional reconstruction of the crack area, identification of crack type, extraction of crack three-dimensional shape, assessment of crack hazard level, and prediction of crack development trend. The communication module is used to enable communication between the crack preliminary location module and the crack precision scanning module, as well as between the crack precision scanning module and the cloud computing server.

6. The tunnel crack detection system as described in claim 5, characterized in that, The communication module includes: a wireless access point installed in the tunnel, and a WIFI communication module installed in the preliminary positioning module and the crack precision scanning module; Furthermore, communication between the initial crack location module and the precise crack scanning module is achieved through the wireless access point and the WIFI communication module.

7. The tunnel crack detection system as described in claim 6, characterized in that, The communication module further includes a 5G communication module disposed in the crack precision scanning module and the cloud computing server; Furthermore, the communication between the crack precision scanning module and the cloud computing server is achieved through the 5G communication module.

8. The tunnel crack detection system as described in any one of claims 5 to 7, characterized in that, The tunnel inspection robot is also equipped with a vehicle-mounted positioning sensor, and the drone is also equipped with an airborne positioning sensor. The vehicle-mounted positioning sensor is used to acquire the relative pose of the tunnel inspection robot at the tunnel entrance and the tunnel entrance itself, which is recorded as the entrance reference relative pose and sent to the crack precision scanning module. Furthermore, the airborne positioning sensor is used to acquire the relative pose of the UAV and the tunnel entrance at the tunnel entrance before the UAV moves according to the optimal inspection path, and denot it as the entrance alignment relative pose, and align the entrance alignment relative pose to the entrance reference relative pose.

9. The tunnel crack detection system as described in claim 8, characterized in that, Also includes: Visual markers placed at the tunnel entrance; Furthermore, both the relative pose of the entrance reference and the relative pose of the entrance alignment are determined using the visual markers.

Images