Tunnel construction inspection method and system based on cooperative operation of multiple unmanned aerial vehicles

Abstract

The invention discloses a tunnel construction inspection method and system based on multi-unmanned-aerial-vehicle cooperative operation, and relates to the field of multi-unmanned-aerial-vehicle cooperative tunnel construction inspection, and the method comprises the steps: obtaining the position information of each unmanned aerial vehicle, and when each unmanned aerial vehicle flies to a tunnel construction tunnel face, according to the position information and obstacle information, determining the position information of each unmanned aerial vehicle; and determining a first optimized path corresponding to each unmanned aerial vehicle by adopting a particle swarm optimization algorithm and a dynamic obstacle avoidance mechanism, and reaching the tunnel construction tunnel face according to the first optimized paths. After a tunnel construction tunnel face is reached, determining a second optimization path corresponding to each unmanned aerial vehicle by adopting a multi-target particle swarm optimization algorithm according to the position information corresponding to the unmanned aerial vehicles, scanning on the second optimization paths according to a preset flight instruction, and adjusting a light source according to a light condition to obtain image data and spectral data; and recognizing the image data and the spectral data by adopting a construction tunnel face recognition model. According to the invention, the detection efficiency and precision can be effectively improved.

Description

Technical Field

[0001] The present application relates to the field of multi-UAV collaborative tunnel construction inspection, and in particular to a tunnel construction inspection method and system based on multi-UAV collaborative operation. Background Art

[0002] In the field of tunnel construction inspection, manual inspection methods have always been the main means, and rely on inspectors to enter the tunnel and use traditional means to directly observe and record the internal conditions of the tunnel. It can be seen that although the above methods are intuitive, they have obvious limitations, including high labor intensity, slow inspection speed, difficulty in covering a large area, and threats to the safety of inspectors under certain circumstances. In addition, the accuracy and consistency of manual inspection data are also affected by human factors, making it difficult to achieve a high degree of standardization and automation. Traditional manual inspection methods face many challenges, especially in complex environments with dim light and many obstacles, the difficulty and risk of inspection work are significantly increased. These problems include not only slow inspection speed, low efficiency, and high labor costs, but also the accuracy and safety of inspections are difficult to guarantee due to environmental restrictions.

[0003] With the advancement of science and technology, drone technology has been introduced into tunnel construction inspection, bringing innovation to this field. The common drone image recognition technology uses a high-definition camera to shoot the inside of the tunnel. After obtaining the image data, it uses image processing software to analyze it to identify cracks, leakage and other problems inside the tunnel. Compared with manual inspection, this method improves the speed and range of inspection, reduces labor intensity, and can improve the accuracy of inspection to a certain extent.

[0004] Although the above attempts to combine drone images with intelligent recognition technology for tunnel inspection, the current methods also have some limitations. Existing drone photography and intelligent recognition systems often lack effective image acquisition capabilities in low-light environments. Inside the tunnel, due to the lack of natural light, existing image acquisition equipment may not be able to obtain sufficiently clear images, resulting in the inability of intelligent recognition algorithms to accurately identify defects and anomalies inside the tunnel. Existing intelligent recognition algorithms have not been fully optimized to adapt to the complexity of the tunnel construction environment, and the diversity of obstacles, construction equipment and materials inside the tunnel may interfere with the algorithm's recognition process, reducing the accuracy and reliability of detection. Furthermore, the existing drone inspection system lacks an effective collaborative operation mechanism. In tunnel construction, multiple areas need to be inspected simultaneously, and the existing system cannot achieve effective coordination between multiple drones, resulting in limited detection efficiency and coverage. In addition, the existing methods do not fully consider the real-time monitoring and dynamic adjustment needs during the construction process. At the same time, tunnel construction is a dynamically changing process, and the existing system cannot adjust the inspection plan and strategy in real time according to the construction progress and environmental changes, thus failing to provide continuous and comprehensive monitoring.

[0005] Therefore, based on the above problems, there is an urgent need to provide a new tunnel construction inspection method based on the collaborative operation of multiple UAVs, so as to realize the dynamic path adjustment of the collaborative operation of multiple UAVs to cope with environmental changes and real-time updates of detection needs, thereby effectively improving the detection efficiency and accuracy. Summary of the invention

[0006] The purpose of this application is to provide a tunnel construction inspection method and system based on the collaborative operation of multiple UAVs, which can realize dynamic path adjustment of the collaborative operation of multiple UAVs to cope with environmental changes and real-time updates of detection needs, thereby effectively improving detection efficiency and accuracy.

[0007] To achieve the above objectives, this application provides the following solutions:

[0008] In a first aspect, the present application provides a tunnel construction inspection method based on multi-UAV collaborative operation, comprising:

[0009] Get the location information of each drone;

[0010] When each UAV performs a flight inspection mission and flies to the tunnel construction face, a first optimization path corresponding to each UAV is determined by using a particle swarm optimization algorithm and a dynamic obstacle avoidance mechanism according to the position information and obstacle information corresponding to the UAV; and the UAV reaches the tunnel construction face according to the corresponding first optimization path;

[0011] After arriving at the tunnel construction face, a multi-objective particle swarm optimization algorithm is used to determine a second optimization path corresponding to each UAV according to the position information corresponding to the UAV;

[0012] Scanning is performed on the second optimized path according to the pre-set flight instructions; and the light source is adjusted according to the light conditions to obtain image data and spectral data;

[0013] According to the image data and the spectral data, a construction face recognition model is used for recognition; the construction face recognition model is constructed based on a convolutional neural network.

[0014] Optionally, obtaining the location information of each drone specifically includes:

[0015] Three signal transmitters are set up in the tunnel; the positions of the three signal transmitters are not in the same straight line;

[0016] According to the signals transmitted by the three signal transmitters, the triangulation positioning algorithm is used to obtain the corresponding position information of each drone.

[0017] Optionally, reaching the tunnel construction face according to the corresponding first optimized path specifically includes:

[0018] During each flight of the UAV, path detection data in the tunnel is obtained; the path detection data includes: tunnel wall images and videos obtained by high-definition visual sensors, spectral data of the tunnel wall obtained by high-resolution spectral analysis sensors, and temperature distribution of the tunnel wall obtained by infrared thermal imaging sensors; images and videos are used to obtain surface cracks and peeling conditions of the tunnel wall; spectral data are used to obtain surface peeling and weathering conditions of the tunnel wall; temperature distribution is used to obtain potential water seepage points, moisture distribution, and internal defects of materials on the tunnel wall;

[0019] The path detection data is stored.

[0020] Optionally, the step of adjusting the light source according to light conditions to obtain image data and spectral data specifically includes:

[0021] Using Formula I new =I old +α·(L desired -L current ) Adjust the light source;

[0022] Among them, I new is the adjusted light intensity, I old is the light source intensity before adjustment, α is the weight parameter for adjusting the light source intensity and amplitude, L desired For ideal brightness, L current is the current brightness.

[0023] Optionally, the identifying using a construction face identification model according to the image data and the spectral data specifically includes:

[0024] Preprocessing the image data and the spectral data; the preprocessing includes: error control coding;

[0025] The pre-processed data is transmitted to the control center at the drone base station in real time;

[0026] The control center uses the construction face recognition model to perform recognition based on the preprocessed data.

[0027] Optionally, the tunnel construction inspection method based on multi-UAV collaborative operation further includes:

[0028] Redundant design and dynamic power allocation strategy are used to adjust the power supply of each drone.

[0029] Optionally, the redundant design and dynamic power distribution strategy are used to adjust the power supply of each drone, specifically including the following formula:

[0030]

[0031] Among them, p i is the i-th detection point, p i+1 is the i+1th detection point, d(p i ,p i+1 ) is the distance between the detection points, R(p i ,p i+1 ) is the environmental risk function, n is the total number of detection points, and β is the risk coefficient adjusted in real time according to environmental changes.

[0032] Optionally, the step of using a construction face recognition model to perform recognition based on the image data and the spectral data may further include:

[0033] The first optimized path and the second optimized path and the environmental parameters during operation are saved, and a path log is generated; the environmental parameters include: spectral data, image data and obstacle information;

[0034] When the next operation is performed, the path is adjusted based on the saved path and the real-time environmental parameters; the saved path includes: a first optimized path and a second optimized path.

[0035] Optionally, when performing the next operation, the path adjustment is performed based on the saved path and the real-time environmental parameters, specifically including:

[0036] Using the formula p new =f(p old ,Ereal-time ,L history ) determine the adjusted path points;

[0037] Among them, p new is the adjusted path point, p old is the historical path point when the task was last executed, E real-time is the real-time environmental parameter in the current task process, L history is the historical path data of the drone when it last performed a mission, and f() is the path adjustment function.

[0038] In the second aspect, the present application provides a tunnel construction inspection system based on multi-UAV collaborative operation, including:

[0039] An information acquisition module is used to obtain the location information of each drone;

[0040] The first UAV collaborative control module is used to determine the first optimized path corresponding to each UAV by using a particle swarm optimization algorithm and a dynamic obstacle avoidance mechanism according to the position information and obstacle information corresponding to the UAV when each UAV performs a flight inspection mission and flies to the tunnel construction face; and arrive at the tunnel construction face according to the corresponding first optimized path;

[0041] The second UAV cooperative control module is used to determine the second optimization path corresponding to each UAV by using a multi-objective particle swarm optimization algorithm according to the position information corresponding to the UAV after arriving at the tunnel construction face;

[0042] A special light source auxiliary shooting module is used to scan on the second optimized path according to a preset flight instruction, and adjust the light source according to the light conditions to obtain image data and spectral data;

[0043] An intelligent data processing and analysis module is used to identify the construction face according to the image data and the spectral data using a construction face recognition model; the construction face recognition model is constructed based on a convolutional neural network.

[0044] According to the specific embodiments provided in this application, this application has the following technical effects:

[0045] The present application provides a tunnel construction inspection method and system based on the collaborative operation of multiple UAVs. The UAVs automatically avoid obstacles and collect data information along the way in real time during the flight to the tunnel face, so as to realize dynamic path adjustment of the collaborative operation of multiple UAVs to cope with environmental changes and real-time updates of detection needs. The scanning path of the tunnel construction face is optimized, and the light source is adjusted according to the tunnel lighting conditions to obtain a sufficiently clear image, so that the intelligent recognition algorithm can accurately identify defects and anomalies inside the tunnel, thereby improving the accuracy and reliability of detection. In tunnel construction, multiple UAVs are used to perform collaborative operations on multiple areas, so as to achieve effective coordination between multiple UAVs and improve detection efficiency and coverage. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 This is a flow chart of a tunnel construction inspection method based on collaborative operation of multiple drones in one embodiment of the present application;

[0048] Figure 2 This is a schematic diagram of a drone flying to a tunnel construction face according to a corresponding first optimized path in an embodiment of the present application;

[0049] Figure 3 This is a schematic diagram of a drone scanning on a second optimized path according to a pre-set flight instruction in an embodiment of the present application;

[0050] Figure 4 This is a schematic diagram of the configuration of drone sensors in one embodiment of the present application.

[0051] Figure numerals: 1-UAV, 2-data storage center and UAV loading device, 3-signal transmitter, 4-high-definition visual sensor, 5-infrared thermal imaging sensor, 6-humidity sensor, 7-ultrasonic sensor, 8-high-resolution spectral analysis sensor. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0053] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0054] In an exemplary embodiment, Figure 1 As shown, the tunnel construction inspection method based on multi-UAV collaborative operation provided by this application specifically includes:

[0055] S1: Get the location information of each drone.

[0056] S2: When each UAV performs a flight inspection mission and flies to the tunnel construction face, a first optimization path corresponding to each UAV is determined using a particle swarm optimization algorithm and a dynamic obstacle avoidance mechanism according to the UAV's corresponding position information and obstacle information; and the UAV reaches the tunnel construction face according to the corresponding first optimization path.

[0057] S3: After arriving at the tunnel construction face, a second optimization path corresponding to each UAV is determined using a multi-objective particle swarm optimization algorithm according to the position information corresponding to the UAV.

[0058] S4: Scanning is performed on the second optimized path according to the pre-set flight instructions; and the light source is adjusted according to the light conditions to obtain image data and spectral data.

[0059] S5: According to the image data and the spectral data, a construction face recognition model is used for recognition; the construction face recognition model is constructed based on a convolutional neural network.

[0060] When multiple drones perform tasks in a tunnel at the same time, the implementation of S1 to S5 above can achieve dynamic path adjustment for the collaborative operation of multiple drones to cope with environmental changes and real-time updates of detection requirements, thereby effectively improving detection efficiency and accuracy.

[0061] In an exemplary embodiment, Figure 2 As shown, S1 specifically includes:

[0062] S110, three signal transmitters 3 are arranged in the tunnel; the positions of the three signal transmitters 3 are not in the same straight line.

[0063] S111, according to the signals transmitted by the three signal transmitters 3, a triangulation positioning algorithm is used to obtain the position information corresponding to each UAV.

[0064] In an exemplary embodiment, three fixed signal transmitters 3 that are not on the same straight line are selected in the tunnel, such as Figure 2As shown, the signal transmitters 3 at these three positions transmit signals to provide positioning mark information for the UAV. Before each flight detection mission starts, the tunnel space position data is calibrated according to the position information of the signal transmitter 3 to ensure that the UAV can be accurately positioned and navigated.

[0065] Specifically, the three signal transmitters 3 constituting the triangular plane are numbered A, B, and C respectively, and the multiple drones are numbered U n , the coordinates of signal transmitters A, B, and C are A(x A ,y A ,z A )、B(x B ,y B ,z B ) and C(x C ,y C ,z C ), the distance between the nth drone and the three signal transmitters 3 is d nA d nB and d nC , the specific drone location information can be reflected by the following formula:

[0066]

[0067] According to the above formula, the specific position U of the drone in the tunnel n (x n ,y n ,z n ) is obtained through the triangulation positioning algorithm and provides accurate reference information for the completion of subsequent path planning and detection tasks.

[0068] In an exemplary embodiment, Figure 2 As shown in the figure, multiple drones are regarded as multiple particles with a certain safety range. After the drones have calibrated their position information in the tunnel, they start to perform flight inspection tasks. When they fly to the tunnel construction face, no one continues to transmit position information based on the calibrated spatial coordinates during the flight. When the drones fly to the tunnel construction face, the obstacle avoidance function of the drone detects obstacles based on the ultrasonic sensor 7 and dynamically adjusts the flight path through PSO. The specific process includes:

[0069] S210, the formula of multi-objective particle swarm optimization algorithm is as follows:

[0070]

[0071] Among them, d(p i ,p i+1 ) is the path distance, E(p i ) is energy consumption, O(p i) is the obstacle avoidance cost, λ1 is the energy consumption related weight coefficient, λ2 is the obstacle avoidance cost related weight coefficient, and the specific values ​​of λ1 and λ2 need to be tested and optimized according to the application scenario in order to achieve a balance in the path optimization between energy consumption and obstacle avoidance.

[0072] S211, such as Figure 4 The sensors shown can provide real-time feedback of obstacle information. When an obstacle is detected, the PSO can adjust the path and calculate a new flight direction. During the flight, the UAV starts the real-time detection and recording function, uses the high-definition visual sensor 4 to obtain tunnel wall images and videos, uses the high-resolution spectral analysis sensor 8 to obtain tunnel wall spectral data, and uses the infrared thermal imaging sensor 5 to obtain the temperature distribution of the tunnel wall. The images and videos are used to obtain the surface cracks and spalling of the tunnel wall; the spectral data are used to obtain the surface spalling and weathering of the tunnel wall; the temperature distribution is used to obtain the potential water seepage points, moisture distribution and internal defects of the material on the tunnel wall.

[0073] S212, a path detection database is established based on the data collected during the flight of the UAV, and is saved in a data storage module, so that historical data can be retrieved for comparison and analysis of abnormal changes during each flight detection.

[0074] In an exemplary embodiment, when the UAV arrives at the tunnel face area, in order to avoid collisions between multiple UAVs working in the tunnel face area, multi-objective particle swarm optimization (MOPSO) is used to optimize the paths of the UAVs. Figure 3 As shown, S3 specifically includes:

[0075] (1) Define the objective function: Minimize the total distance of the working face path to reduce energy consumption and time:

[0076]

[0077] Among them, d(p k ,p k+1 ) are two nodes p in the path k and p k+1 The distance between.

[0078] To avoid collisions between multiple drones, a certain safe distance should be maintained between drones. The collision cost function can be defined by the relative position between two drones:

[0079]

[0080] Among them, r i and r j is a vector function, representing the position coordinates of drones i and j in three-dimensional space, ||r i -rj || represents the distance between two drones.

[0081] Each drone's working path should cover new areas as much as possible and avoid repeated detection. To this end, the function for maximizing the detection area coverage is set as:

[0082]

[0083] Among them A i is the detection area of ​​UAV i, and the goal is to minimize the overlap of the monitoring area.

[0084] (2) Initialize the particle swarm: Each particle represents the path of a drone, and the path is represented by a point sequence p i =[p1,p2,...p n ] means that each point p i It is a detection point on the face of the tunnel. The initial position of each particle (path) is randomly generated, and the speed is also random. The speed indicates the adjustment direction and amplitude of the drone in path optimization.

[0085] Its position initialization formula is:

[0086] x i =x min +randΟ·(x max -x min ).

[0087] The speed initialization formula is:

[0088] v i =v min +randΟ·(v max -v min ).

[0089] Among them, x min and x max are the minimum and maximum boundaries of the palm face respectively, and randΟ is a random number between 0 and 1.

[0090] (3) Update particle position and velocity: In each iteration, the particle position and velocity are updated according to the particle's individual optimal position and global optimal position. The position update formula is:

[0091] x i (t+1)=x i (t)+v i (t+1).

[0092] Among them, x i (t+1) is the position of the particle (drone) after the iteration.

[0093] The speed update formula is:

[0094] v i (t+1)=ωv i (t)+c1r1(p i (t)-x i (t))+c2r2(g(t)-x i (t)).

[0095] Among them, ω is the inertia weight, which controls the ability of the particle to maintain its current motion, and c1 and c2 are acceleration factors, which control the particle to its individual optimal position p. i (t) and the ability to move to the global optimal position g(t). r1 and r2 are random numbers to ensure the diversity of particles.

[0096] (4) Evaluate particle fitness: For each particle, calculate its fitness under multiple objective functions, based on the following:

[0097] a) Path distance f1(p i ) as small as possible.

[0098] b) The distance between drones should be greater than the set safety threshold to avoid collisions. The safety distance should be set based on the flight speed of the drone, the complexity of the flight environment, and other factors. If the distance between drones is less than the safety threshold, the system will increase the penalty, and the optimization algorithm will tend to increase the distance between them.

[0099] c) Detection area coverage f3(A i ) should be maximized to ensure that the area covered by each drone does not overlap with other drones as much as possible.

[0100] Each objective function is normalized using the following formula:

[0101]

[0102] Among them, f' k (x) is the normalized objective function value, f k (x) is the original objective function value, f k max and f k min are the maximum and minimum values ​​of the objective function respectively.

[0103] The multi-objective optimization goal of fitness is to find the Pareto optimal solution set, that is, the solution set that cannot be dominated by other solutions in multiple objectives. If solution x1 is better than solution x2 in all objective functions, then x1 is considered to dominate x2, that is:

[0104]

[0105] (5) In each iteration of the multi-objective particle swarm optimization algorithm, each particle updates its individual optimal position and the global optimal position according to its fitness value.

[0106] a) Individual optimal position selection: For each particle i, its individual optimal position p i (t) represents the optimal solution reached by the particle in the historical iteration. That is, each particle will retain the solution with the highest fitness in all previous iterations as the individual optimal position. In each iteration, by comparing the fitness of the current solution with the fitness value of the individual optimal position, if the current solution performs better than the previously recorded optimal position under all objective functions, the individual optimal position is updated.

[0107]

[0108] Among them, x i (t+1) represents the current solution of particle i at iteration t+1, p i (t) represents the historical optimal solution, f(x) is the fitness function, and the comprehensive performance of the solution on multiple objective functions is compared.

[0109] b) The global optimal position g(t) represents the position of the particle that performs best in the current iteration among all particles. In each iteration, the algorithm evaluates the fitness values ​​of all particles on multiple objective functions and selects a solution in the Pareto optimal solution set as the global optimal position.

[0110]

[0111] Among them, g(t) is the current global optimal position, and f(g(t)) is its fitness value.

[0112] During each iteration, the system updates the individual optimal and global optimal solutions for all particles to ensure that each particle moves towards the global optimal solution and gradually converges to the multi-objective optimal solution set.

[0113] (6) The algorithm continues to iterate, updating the particle's position and velocity until one of the following termination conditions is met:

[0114] 1) The predetermined maximum number of iterations is reached.

[0115] 2) The change of the particle swarm is less than a certain threshold, indicating that it has converged.

[0116] (7) After the iteration is completed, a set of Pareto optimal solutions is obtained, representing the second optimized path. The system selects a set of solutions from the solution set and assigns them to each drone for execution. Specifically, the path of each drone is assigned to the corresponding sub-area to ensure that the maximum area of ​​the face is covered and to avoid overlapping and collision of drone paths. The drone starts the inspection task according to the second optimized path.

[0117] In an exemplary embodiment, when the drone arrives at the tunnel face area, the multi-drone collaborative path planning and scanning detection function is further enabled, and multiple drones are controlled to scan along the edge of the rock wall at the tunnel face according to the pre-set flight instructions to determine the position of the working area of ​​the tunnel face to be detected, such as Figure 3 As shown in the figure, when multiple drones arrive at the tunnel face and scan the tunnel face contour range, the drones turn on special light sources to assist in shooting, that is, real-time monitoring of lighting conditions, dynamic adjustment of light source intensity and angle, so as to enable efficient shooting in insufficient light conditions. S4 specifically includes:

[0118] Using Formula I new =I old +α·(L desired -L current ) to adjust the light source.

[0119] Among them, I new is the adjusted light intensity, I old is the light source intensity before adjustment, L desired For ideal brightness, L current is the current brightness, α is the weight parameter for adjusting the intensity and amplitude of the light source, which controls the speed and amplitude of the light source intensity change. The larger the α value, the faster the light source intensity is adjusted, and vice versa, the smaller α is, the slower the adjustment speed is.

[0120] In an exemplary embodiment, Figure 4 As shown in the figure, the drone sensor configuration includes:

[0121] The high-definition visual sensor 4 is used to obtain images and videos of the tunnel wall, and the surface defects of the tunnel wall, such as cracks, peeling, deformation, etc., are obtained based on the images and videos of the tunnel wall. During the flight of the drone, the high-definition visual sensor 4 is used to take real-time photos of the tunnel wall, and the edge detection algorithm (such as Canny edge detection) is used to identify the starting point, end point and length of the crack.

[0122] The temperature distribution of the tunnel wall is acquired by using the infrared thermal imaging sensor 5, and the potential water seepage points, moisture distribution, thermal stress and voids inside the material, etc. on the tunnel wall are obtained based on the temperature distribution.

[0123] The humidity sensor 6 is used to monitor the humidity in the tunnel air, assisting the infrared thermal imaging sensor 5 to more accurately locate the water seepage area.

[0124] The ultrasonic sensor 7 is used to detect obstacles in the tunnel, and combined with the particle swarm optimization algorithm, it ensures that the UAV can automatically find its way in the complex tunnel environment; in addition, the ultrasonic sensor 7 continuously emits sound waves when the UAV is flying, and determines the integrity of the tunnel wall through the reflected sound wave signals.

[0125] The high-resolution spectral analysis sensor 8 is used to analyze the surface changes of the material, such as surface peeling, weathering, etc., through reflected light of different wavelengths.

[0126] In an exemplary embodiment, S5 specifically includes:

[0127] S510, in order to improve the accuracy of disease identification, the image data and spectral data collected by the drone are transmitted in real time to the control center at the drone base station through high-speed wireless communication transmission equipment for data fusion. Error control coding technology is used during the transmission process to ensure the accuracy and integrity of the data.

[0128] Data transmission uses Huffman coding to reduce the redundancy of transmitted data and improve transmission efficiency. The specific coding includes the following formula:

[0129] D compressed =H(D original ).

[0130] Among them, H() is the Huffman compression function, D original is the original data or information, D compressed is the data after Huffman compression.

[0131] S511: A construction face recognition model is constructed based on a convolutional neural network (CNN) and spectral analysis technology. The construction face recognition model identifies potential diseases on the tunnel construction face, thereby more comprehensively analyzing the health status of the tunnel construction face. The data fusion formula is:

[0132]

[0133] Among them, F image is the image data collected by the high-definition visual sensor; F spectrum is the spectral data collected by a high-resolution spectral analysis sensor; It represents the joint operation of image data and spectral data, and maps different types of data into the same feature space through the fusion of feature layers.

[0134] Among them, the convolutional neural network (CNN) can process image data and spectral data at the same time, and integrate the feature vectors of two different paths through the fusion layer to generate a joint feature map. Its CNN structure layer includes: input layer, the input layer can input image and spectral data into the two branches of CNN respectively. Convolution layer, each branch is through multiple convolution layers to extract features of different scales. The image branch mainly identifies structural features such as cracks and edges, and the spectral branch mainly identifies material change features. Pooling layer, the pooling layer can downsample the features to reduce the amount of calculation while maintaining important feature information. Fusion layer, the fusion layer can splice the image and spectrum to generate a multi-modal joint feature representation.

[0135] To further improve the accuracy of disease identification, the convolutional neural network uses multi-scale feature extraction technology to identify structural anomalies of different sizes, depths and locations; the convolutional neural network can automatically detect the edges and widths of cracks, and spectral data analysis can detect the physical changes of materials at the cracks, based on which crack identification, spalling detection and structural anomaly detection can be performed on the tunnel construction face. In addition, the detection data is fed back and stored in a timely manner. This method can not only identify current construction diseases, but also combine historical data and multiple test results to conduct a comprehensive analysis and prediction of the development trend of tunnel construction face diseases.

[0136] In an exemplary embodiment, a continuous and stable power supply can be provided for the entire drone, and a redundant design and a dynamic power distribution strategy are used to adjust the power supply of each drone to ensure the normal operation of the drone under a high-load environment; the power supply of each drone is adjusted through the dynamic power distribution strategy, which specifically includes the following formula:

[0137]

[0138] Among them, p i is the i-th detection point, p i+1 is the i+1th detection point, d(p i ,p i+1 ) is the distance between the detection points, R(p i ,p i+1 ) is the environmental risk function, n is the total number of detection points, and β is the risk coefficient adjusted in real time according to environmental changes.

[0139] In an exemplary embodiment, after S5, the following steps are further included:

[0140] After each operation, the UAV saves the first optimized path and the second optimized path as well as the environmental parameters during the operation, and generates a path log L(p i ,E i ), where E i Denotes the detection point p iThe environmental parameters include: spectral data, image data and obstacle information.

[0141] When the next operation is performed, the path is adjusted based on the first optimized path, the second optimized path and the real-time environmental parameters.

[0142] Using the formula p new =f(p old ,E real-time ,L history ) to determine the adjusted path points.

[0143] Among them, p new is the adjusted path point, p old is the historical path point when the task was last executed, E real-time is the real-time environmental parameter in the current task process, L history is the historical path data of the drone when it last performed a mission, and f() is the path adjustment function.

[0144] Through the above formula, the present application can measure data and historical path information, dynamically optimize the flight route, and then timely adjust and update the path and environmental data values ​​based on environmental changes and measured data, to achieve dynamic path adjustment of multi-UAV collaborative operations to cope with environmental changes and real-time updates of detection needs.

[0145] This application realizes the path planning and task allocation of drones through intelligent algorithms, solving the problems of limited coverage and low detection efficiency of traditional inspection methods. Through the collaborative work of multiple drones, not only the detection coverage of the tunnel construction area is expanded, but also the efficiency of inspection is improved. Especially in the complex construction environment in the tunnel, this system can ensure effective collaboration and collision avoidance between drones through path optimization algorithms. Therefore, it is necessary to protect the collaborative control mechanism, task allocation and path planning optimization technology of drones.

[0146] Secondly, the lighting conditions inside the tunnel are complex, and conventional camera equipment is difficult to work effectively. This application can provide stable lighting in low-light environments to ensure the clarity and reliability of the captured images. At the same time, the automatic adjustment function of the light source allows it to be flexibly adjusted according to the real-time environment.

[0147] In addition, the multimodal data processing and fusion technology adopted in this application uses convolutional neural network (CNN) to jointly analyze the image data collected by drones and spectral data, thereby enhancing the accuracy of disease identification. This multimodal data processing can not only detect structural problems on the tunnel surface, but also deeply analyze the internal defects of the material.

[0148] In addition, this application introduces dynamic optimization technology in the path planning of drones, which can be dynamically adjusted according to historical path information and real-time environmental data. With this technology, drones can flexibly avoid obstacles in complex tunnel construction environments and optimize flight paths in real time to ensure efficient execution of inspection tasks.

[0149] To sum up, this application can combine multiple fields such as multi-UAV collaborative operation, special light source assistance, high-resolution spectral analysis sensors, multimodal data fusion, dynamic path optimization, etc., to ensure the comprehensiveness and efficiency of the tunnel construction intelligent inspection system, and significantly improve the accuracy and safety of the inspection.

[0150] In an exemplary embodiment, a tunnel construction inspection system based on multi-UAV collaborative operation is provided, which specifically includes:

[0151] Information acquisition module: used to obtain the location information of each drone.

[0152] The first UAV collaborative control module is used to determine the first optimized path corresponding to each UAV based on the position information and obstacle information of the UAV when each UAV performs a flight inspection mission and flies to the tunnel construction face, and to reach the tunnel construction face according to the corresponding first optimized path by using the particle swarm optimization algorithm and the dynamic obstacle avoidance mechanism.

[0153] The second UAV collaborative control module is used to determine the second optimization path corresponding to each UAV by using a multi-objective particle swarm optimization algorithm according to the position information corresponding to the UAV after arriving at the tunnel construction face.

[0154] Special light source auxiliary shooting module: used to scan on the second optimized path according to the pre-set flight instructions, and adjust the light source according to the light conditions to obtain image data and spectral data.

[0155] Intelligent data processing and analysis module: used to identify the construction face recognition model based on the image data and the spectral data; the construction face recognition model is constructed based on a convolutional neural network.

[0156] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0157] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A tunnel construction inspection method based on multi-UAV collaborative operation, characterized in that: The tunnel construction inspection method based on multi-UAV collaborative operation includes: Get the location information of each drone; When each UAV performs a flight inspection mission and flies to the tunnel construction face, a first optimization path corresponding to each UAV is determined by using a particle swarm optimization algorithm and a dynamic obstacle avoidance mechanism according to the position information and obstacle information corresponding to the UAV; and the UAV reaches the tunnel construction face according to the corresponding first optimization path; After arriving at the tunnel construction face, a multi-objective particle swarm optimization algorithm is used to determine a second optimization path corresponding to each UAV according to the position information corresponding to the UAV; Scanning is performed on the second optimized path according to the pre-set flight instructions; and the light source is adjusted according to the light conditions to obtain image data and spectral data; According to the image data and the spectral data, a construction face recognition model is used for recognition; the construction face recognition model is constructed based on a convolutional neural network.

2. The tunnel construction inspection method based on multi-UAV collaborative operation according to claim 1 is characterized in that: The obtaining of the location information of each drone specifically includes: Three signal transmitters are set up in the tunnel; the positions of the three signal transmitters are not in the same straight line; According to the signals transmitted by the three signal transmitters, the triangulation positioning algorithm is used to obtain the corresponding position information of each drone.

3. The tunnel construction inspection method based on multi-UAV collaborative operation according to claim 1 is characterized in that: The method of reaching the tunnel construction face according to the corresponding first optimized path specifically includes: During each flight of the UAV, path detection data in the tunnel is obtained; the path detection data includes: tunnel wall images and videos obtained by high-definition visual sensors, spectral data of the tunnel wall obtained by high-resolution spectral analysis sensors, and temperature distribution of the tunnel wall obtained by infrared thermal imaging sensors; images and videos are used to obtain surface cracks and peeling of the tunnel wall; spectral data are used to obtain surface peeling and weathering of the tunnel wall; temperature distribution is used to obtain potential water seepage points, moisture distribution and internal defects of materials on the tunnel wall; The path detection data is stored.

4. The tunnel construction inspection method based on multi-UAV collaborative operation according to claim 1 is characterized in that: The step of adjusting the light source according to the light conditions to obtain image data and spectral data specifically includes: Using Formula I new =I old +α·(L desired -L current ) Adjust the light source; Among them, I new is the adjusted light intensity, I old is the light source intensity before adjustment, α is the weight parameter for adjusting the light source intensity and amplitude, L desired For ideal brightness, L current is the current brightness.

5. The tunnel construction inspection method based on multi-UAV collaborative operation according to claim 1 is characterized in that: The identifying using the construction face identification model according to the image data and the spectral data specifically includes: Preprocessing the image data and the spectral data; the preprocessing includes: error control coding; The pre-processed data is transmitted to the control center at the drone base station in real time; The control center uses the construction face recognition model to perform recognition based on the preprocessed data.

6. The tunnel construction inspection method based on multi-UAV collaborative operation according to claim 1 is characterized in that: The tunnel construction inspection method based on multi-UAV collaborative operation also includes: Redundant design and dynamic power allocation strategy are used to adjust the power supply of each drone.

7. The tunnel construction inspection method based on multi-UAV collaborative operation according to claim 6 is characterized in that: The redundant design and dynamic power distribution strategy are used to adjust the power supply of each drone, specifically including the following formula: Among them, p i is the i-th detection point, p i+1 is the i+1th detection point, d(p i ,p i+1 ) is the distance between the detection points, R(p i ,p i+1 ) is the environmental risk function, n is the total number of detection points, and β is the risk coefficient adjusted in real time according to environmental changes.

8. The tunnel construction inspection method based on multi-UAV collaborative operation according to claim 1 is characterized in that: The method further comprises: identifying the construction face using a construction face identification model according to the image data and the spectral data; and then: The first optimized path and the second optimized path and the environmental parameters during operation are saved, and a path log is generated; the environmental parameters include: spectral data, image data and obstacle information; When the next operation is performed, the path is adjusted based on the saved path and the real-time environmental parameters; the saved path includes: a first optimized path and a second optimized path.

9. The tunnel construction inspection method based on multi-UAV collaborative operation according to claim 8 is characterized in that: When the next operation is performed, the path is adjusted based on the saved path and the real-time environmental parameters, specifically including: Using the formula p new =f(p old ,E real-time ,L history ) determine the adjusted path points; Among them, p new is the adjusted path point, p old is the historical path point when the task was last executed, E real-time is the real-time environmental parameter in the current task process, L history is the historical path data of the drone when it last performed a mission, and f() is the path adjustment function.

10. A tunnel construction inspection system based on multi-UAV collaborative operation, characterized in that: The tunnel construction inspection system based on multi-UAV collaborative operation includes: An information acquisition module is used to obtain the location information of each drone; The first UAV collaborative control module is used to determine the first optimized path corresponding to each UAV by using a particle swarm optimization algorithm and a dynamic obstacle avoidance mechanism according to the position information and obstacle information corresponding to the UAV when each UAV performs a flight inspection mission and flies to the tunnel construction face; and arrive at the tunnel construction face according to the corresponding first optimized path; The second UAV cooperative control module is used to determine the second optimization path corresponding to each UAV by using a multi-objective particle swarm optimization algorithm according to the position information corresponding to the UAV after arriving at the tunnel construction face; A special light source auxiliary shooting module is used to scan on the second optimized path according to a preset flight instruction, and adjust the light source according to the light conditions to obtain image data and spectral data; An intelligent data processing and analysis module is used to identify the construction face according to the image data and the spectral data using a construction face recognition model; the construction face recognition model is constructed based on a convolutional neural network.

Images