Unmanned aerial vehicle intelligent detection system for bridge and tunnel diseases

Abstract

The invention relates to the technical field of bridge and tunnel disease detection, in particular to an unmanned aerial vehicle intelligent detection system for bridge and tunnel diseases. The bridge and tunnel disease unmanned aerial vehicle intelligent detection system comprises an unmanned aerial vehicle body and a ground base station, and the unmanned aerial vehicle body is internally provided with a flight control system, a holder system, a camera system, a Raspberry Pi system, a storage system and a power supply system. The 360-degree panoramic camera is used for data acquisition, comprehensive image data can be obtained, all-around monitoring and recording of bridge and tunnel inspection are achieved, no blind area exists, missing inspection is avoided, the efficiency, accuracy and safety of inspection work can be remarkably improved, and powerful technical support is provided for management and maintenance of bridges and tunnels.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge and tunnel disease detection, and specifically to an intelligent unmanned aerial vehicle (UAV) detection system for bridge and tunnel diseases. Background Art

[0002] Bridges and tunnels are important infrastructure for transportation, and their safety and stability are directly related to the safety of people's lives and property. During the use of bridges and tunnels, various diseases will occur under the influence of natural environment and human factors, such as cracks, water seepage, loosening, collapse, etc. These diseases will reduce the structural strength and service life of bridges and tunnels, and even cause serious safety accidents. Therefore, regular inspection and maintenance are required.

[0003] Currently, the inspection of bridges and tunnels mainly relies on manual labor, that is, professional personnel use instrument equipment to observe, measure and record the bridge and tunnel walls. This method has the following disadvantages:

[0004] 1. Manual inspection has low efficiency, long duration and high cost;

[0005] 2. The quality of manual inspection is unstable and is easily affected by factors such as personnel quality, experience, and eyesight;

[0006] 3. There are safety hazards in manual inspection. The bridge and tunnel structures are complex and may pose risks to the inspection personnel.

[0007] Currently, UAVs have been widely used in various fields, but the application of UAVs in bridge and tunnel diseases is still in the trial application stage, and there are many deficiencies in the inspection technology, which are mainly manifested in the following aspects:

[0008] 1. It is unable to collect panoramic images and can only perform detection operations from the first perspective. Data cannot be collected simultaneously for the bottom, side, top, and both sides of the bridge. The inspection efficiency is low and it is easy to miss inspections;

[0009] 2. It cannot automatically identify bridge and tunnel diseases. After the UAV obtains the bridge and tunnel inspection data, manual experience is still required to distinguish diseases, which is time-consuming and has a high labor cost;

[0010] 3. The bridge and tunnel inspection scenarios are complex, and it is difficult to fly automatically, and the safety and stability of the inspection cannot be guaranteed. Summary of the Invention

[0011] The purpose of the present invention is to provide an intelligent UAV detection system for bridge and tunnel diseases, so as to solve the problems of low inspection efficiency, long duration, high cost, unstable inspection quality, and safety hazards caused by the fact that the inspection of bridges and tunnels mainly relies on manual labor in the above-mentioned background art. At the same time, when only using UAVs for inspection, panoramic images cannot be collected, bridge and tunnel diseases cannot be automatically identified, and the safety and stability of UAV inspection cannot be guaranteed.

[0012] To achieve the above object, the present invention provides the following technical solutions: a UAV intelligent detection system for bridge and tunnel diseases, including a UAV body and a ground base station. The UAV body is internally provided with a flight control system, a pan-tilt system, a camera system, a Raspberry Pi system, a storage system, and a power supply system. The UAV bridge and tunnel inspection method includes the following operating steps:

[0013] S1. Input inspection task parameters, including inspection task location, length, width, and height information, into the ground base station control center and send them to the UAV;

[0014] S2. After receiving the task parameters, the UAV starts the lidar and infrared rangefinder to determine the initial position and attitude of the UAV, and plans the inspection flight trajectory of the UAV according to the task parameters;

[0015] S3. The UAV executes the inspection task according to the flight trajectory, and at the same time starts the 360-degree panoramic camera to collect images and transmits the image data to the ground control center in real time;

[0016] S4. During the flight of the UAV, the lidar and infrared rangefinder are continuously used for scanning and measurement to obtain the surrounding distance, shape, temperature, and humidity data, and the flight speed, height, and direction parameters of the UAV are adjusted according to the data to achieve intelligent obstacle avoidance;

[0017] S5. After receiving the image data, the ground base station control center uses the deep learning intelligent inspection module to analyze and process the image data, identify the disease type, location, size, and degree information, store the disease information in the database, and generate an inspection report;

[0018] S6. After the UAV completes the inspection task, it returns to the starting position and ends the flight;

[0019] S7. The ground base station control center performs image processing on the photos. The specific steps are as follows:

[0020] S71. The ground base station control center receives the image data through the receiving module and loads it into the image processing module;

[0021] S72. The image processing module detects bridge and tunnel diseases through images, and combines the deep learning intelligent inspection module to build a deep learning neural network environment, and uses the YOLO V8 algorithm to optimize the disease recognition to improve the recognition of whether the disease is a crack, water leakage, or peeling.

[0022] S73. Extract the features of the identified disease types, including the location, length, width, area, and severity of the diseases.

[0023] S74. Combine the data of bridge-tunnel mileage, cross-section position, and leakage type, determine whether it affects train operation, and make corresponding selections to improve the disease information.

[0024] S75. Mark the diseases in combination with auxiliary information.

[0025] S76. Output the marked final information and store it in the storage system at the same time to classify the severity level of the lining, the causes of diseases, and the disease prevention measures to generate an inspection report.

[0026] Preferably, the ground base station includes a receiving module, an image processing module, a monitoring module, and a deep learning intelligent inspection module. The deep learning intelligent inspection module is mainly used to label the collected images, that is, to indicate normal and disease types, and use the custom labels as dataset samples for model training. The image processing module detects bridge-tunnel diseases and identifies diseases through the images obtained by the drone. The deep learning intelligent inspection module builds a deep learning neural network environment and optimizes the disease recognition through the YOLO V8 algorithm.

[0027] Preferably, the drone body is provided with arm rotors, an infrared rangefinder, and a lidar. The flight control system and the Raspberry Pi system are both electrically connected to the arm rotors and the infrared rangefinder on the drone body. The Raspberry Pi system is electrically connected to the lidar of the drone body.

[0028] Preferably, the drone body is provided with a binocular camera and a 360-degree panoramic camera. The camera system and the storage system are electrically connected to the binocular camera and the 360-degree panoramic camera of the drone.

[0029] Preferably, the drone body is provided with a communication module. The flight control system, the pan-tilt system, the camera system, the Raspberry Pi system, the storage system, and the power supply system are electrically connected to the ground base station through the communication module.

[0030] Preferably, the deep learning intelligent inspection module is electrically connected to the camera system, the Raspberry Pi system, and the storage system.

[0031] Preferably, the drone device used in the bridge-tunnel disease drone intelligent detection system includes a drone body, flight arms, an infrared rangefinder, a binocular camera, a frame, a wireless communication module, a 360-degree panoramic camera, a Raspberry Pi device, and a lidar.

[0032] Compared with the prior art, the beneficial effects of the present invention are:

[0033] 1. The present invention uses a 360-degree panoramic camera for data collection, which can obtain comprehensive image data, achieve all-round monitoring and recording of bridge and tunnel inspections, without blind spots and omissions, and can significantly improve the efficiency, accuracy, and safety of inspection work, providing strong technical support for the management and maintenance of bridges and tunnels.

[0034] 2. The invention conducts three-dimensional modeling of the bridge and tunnel inspection scene through a lidar system, perceives the complex inspection scene in real time, realizes automatic obstacle avoidance, reasonably plans the automatic inspection flight path, effectively reduces the risk of collision during UAV inspection, and achieves automatic flight and precise positioning and navigation in a complex bridge and tunnel inspection environment.

[0035] 3. The present invention uses deep learning-based intelligent disease detection to analyze and process image data, which can automatically identify the type, location, size, and degree of bridge and tunnel diseases, with high detection and recognition accuracy, quickly provide disease location information, strong timeliness, flexibility, and effectively improve the inspection efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is the flow chart of the disease detection proposed by the present invention;

[0037] Figure 2 is the two-dimensional navigation and positioning scene diagram of the lidar;

[0038] Figure 3 is the three-dimensional modeling diagram of the lidar inspection scene;

[0039] Figure 4 is the schematic diagram of the obstacle avoidance method of the lidar of the present invention;

[0040] Figure 5 is the reading schematic diagram of the lidar of the present invention;

[0041] Figure 6 is the schematic diagram of the inspection system process of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0043] Please refer to Figures 1-6 , an embodiment provided by the present invention:

[0044] Bridge and Tunnel Disease UAV Intelligent Detection System: It includes a UAV body and a ground base station. Inside the UAV body, there are a flight control system, a gimbal system, a camera system, a Raspberry Pi system, a storage system, and a power system. The method for the UAV to inspect bridges and tunnels includes the following operating steps:

[0045] S1. Input inspection task parameters, including the location, length, width, and height information of the inspection task, into the control center of the ground base station and send them to the UAV;

[0046] S2. After receiving the task parameters, the UAV starts the lidar and infrared rangefinder to determine its initial position and attitude, and plans the inspection flight trajectory of the UAV according to the task parameters;

[0047] S3. The UAV executes the inspection task according to the flight trajectory, and at the same time starts the 360-degree panoramic camera to collect images and transmits the image data to the ground control center in real time;

[0048] S4. During the flight, the UAV continuously uses the lidar and infrared rangefinder to scan and measure, obtains the surrounding distance, shape, temperature, and humidity data, and adjusts the flight speed, height, and direction parameters of the UAV according to the data to achieve intelligent obstacle avoidance;

[0049] S5. After receiving the image data, the ground base station control center uses the deep learning intelligent inspection module to analyze and process the image data, identifies the disease type, location, size, and degree information, stores the disease information in the database, and generates an inspection report;

[0050] S6. After the UAV completes the inspection task, it returns to the starting position and ends the flight;

[0051] S7. The ground base station control center performs image processing on the photos. The specific steps are as follows:

[0052] S71. The ground base station control center receives the image data through the receiving module and loads it into the image processing module;

[0053] S72. The image processing module detects bridge and tunnel diseases through images, combines with the deep learning intelligent inspection module to build a deep learning neural network environment, and uses the YOLO V8 algorithm to optimize the disease recognition to improve the recognition of whether the disease is a crack, water leakage, or spalling.

[0054] S73. Extract the features of the identified disease types, including the location, length, width, area, and severity of the diseases.

[0055] S74. Combine the data of bridge and tunnel mileage, cross-section position, and water leakage type to judge whether it affects train operation and make corresponding selections to improve the disease information.

[0056] S75. Mark the disease by combining auxiliary information.

[0057] S76. Output the marked final information and simultaneously store it in the storage system to classify the severity level of the lining, the causes of the disease, and the disease prevention measures, so as to generate an inspection report.

[0058] Furthermore, for the connection between the lidar and Pixhawk, the Pixhawk flight controller provides serial ports for custom ultrasonic rangefinders or sensors such as lidar. In the previous chapter, uORB was introduced. The biggest feature of this protocol is that it can read the data in the custom sensor from the serial port or I2C interface and connect it to the system. Therefore, the connection between the lidar and Pixhawk can select the TELEM1, TELEM2, SERIAL4, and SERIAL5 interfaces. In this system, the TELEM1 interface is selected for connection.

[0059] Furthermore, after connecting the lidar to Pixhawk, parameter settings are required to turn on the obstacle avoidance function. The debugging process will be described below.

[0060] (1) Modify the corresponding port function

[0061] In this system, the lidar selects the TELEM1 port in Pixhawk. This port can be used for different sensors. Among them, 11 corresponds to Lidar36. Therefore, we need to open the entire parameter table in configuration and debugging for adjustment, and change the value after SERIAL1_PROTOCOL to 11.

[0062] (2) Select the port baud rate

[0063] To achieve the transmission of information data, a transmission link needs to be established. During the process of data and other information being transmitted through the link, the signal unit carrying the data information is called a symbol. The number of symbols passing through the signal link per unit time is called the symbol transmission rate, abbreviated as the baud rate. Its unit is baud (Baud, symbol / s), and the baud rate can be regarded as the bandwidth index of the data link. Therefore, to enable Pixhawk to obtain the data of the lidar, the corresponding port baud rate needs to be selected, and the SERIAL1_BAUD parameter in the entire parameter table is modified to 115 corresponding to the 115200 baud rate.

[0064] (3) Select the sensor model

[0065] This step is mainly a selection that can be made after the Mission Planner ground station is updated. However, the lidar model used in this system is not available in this option. Instead, selecting the product RPLidarA2 of the same company can ensure normal operation. Changing PRX_TYPE to 5 can correspond to the required option.

[0066] (4) Select the lidar installation location

[0067] This step is to enable the flight controller to know the installation location of the lidar, so as to determine whether the acquired data is the forward data of the upright lidar or the reverse data generated by the inverted lidar. Modify the parameter list PRX_ORIENT data. 0 means the lidar is at the top of the UAV, and 1 means the lidar is inverted at the bottom of the UAV. In this system, the lidar is placed at the top of the UAV, so this parameter needs to be modified to 0.

[0068] (5) Disable the flow control function

[0069] When using the TELEM1 / TELEM2 ports of Pixhawk, there will be a data stream between the hardware and Pixhawk. Data loss often occurs during transmission, especially when the storage space of the data buffer of the receiving-end Pixhawk is insufficient during transmission. The transmitting-end sensor still transmits the collected data, resulting in the loss of the transmitted data. Flow control can solve this problem. The function of flow control is to send a stop command to the transmitting end when the data processing ability of the receiving-end Pixhawk processor is insufficient. However, the ports of the Pixhawk flight controller used in this system do not have hardware flow processing pins, so this function needs to be disabled. Change the BRD_SER1_RTSCTS parameter to 0 in the entire parameter list to disable the flow control function.

[0070] (6) Restart Pixhawk

[0071] After parameter adjustment, the Pixhawk flight controller needs to be restarted. When powering on again, make sure the wiring between the lidar and Pixhawk is correct before reconnecting the battery. After power-on, select the correct serial port and baud rate in the ground station Mission Planner to connect to the flight controller. At this time, you can use the shortcut key Crtr+F to open the temp window in the ground station interface, and click Proximity to open the radar window to visualize the surrounding obstacles;

[0072] (7) Enable the obstacle avoidance function

[0073] At this time, only the most important part of obstacle information perception in the aircraft obstacle avoidance process is completed. At this time, the built-in obstacle avoidance function in the used firmware can be turned on. At this time, it is necessary to open the entire parameter table again and modify the parameter AVOID_ENABLE representing obstacle avoidance. Since this lidar cannot identify the obstacle category, the obstacle avoidance mode needs to be modified to avoid all obstacles. Among them, 7 represents turning on the obstacle avoidance function in all directions. Therefore, this parameter needs to be changed to 7.

[0074] (8) Select the channel as the switch for the obstacle avoidance function

[0075] After adding the lidar to enable the UAV to have the obstacle avoidance function, the obstacle avoidance function has no effect when flying in a normal empty airspace. Therefore, it can be selected to be turned on in some special environments that require obstacle avoidance. For example, in bridge and tunnel inspections, there will be many visual blind spots in the field of view of the UAV. It is difficult to manually avoid obstacles in the visual blind spots. At this time, the obstacle avoidance function can be actively turned on at the remote control end so that the aircraft can perform autonomous obstacle avoidance. Therefore, it is necessary to select the currently vacant channel in the 8 available channels for setting. First, it is necessary to find the CH8_OPT parameter corresponding to controlling the 8th channel in the entire parameter table and modify this parameter. In this parameter list, 40 represents the control of the obstacle avoidance function. At this time, this parameter can be modified to 40.

[0076] (9) Select the start obstacle avoidance distance and the obstacle avoidance method

[0077] The start obstacle avoidance distance is the distance at which the aircraft identifies that the obstacle can threaten the safety of the aircraft during the approaching process between the obstacle and the aircraft, so as to ensure the flight safety of the aircraft. The obstacle avoidance action is the action form taken by the aircraft when performing obstacle avoidance, which can be divided into braking obstacle avoidance and bypassing sliding obstacle avoidance. As Figure 4 shown, the so-called braking obstacle avoidance is to stop the UAV from approaching when it reaches the start obstacle avoidance distance. The bypassing sliding obstacle avoidance is that the UAV bypasses the obstacle by sliding around from the area without obstacles on the side after identifying the obstacle.

[0078] The lidar used in this system has a detection radius of up to 12 meters. Therefore, in this system, an obstacle avoidance range of 0 - 12 meters centered on the lidar can be set. To ensure the flight safety and smoothness of the aircraft, when setting this value, generally, the maximum distance between the rotation ranges of the diagonal propellers needs to be considered as shown in the figure. On this basis, an appropriate additional distance can be added. In this system, the safety distance is set to 3 meters. To set 3 meters, the parameter AVOID_MARGIN representing the starting obstacle avoidance distance needs to be found in the parameter list. Setting this parameter to 3 means that the obstacle avoidance function is triggered when the distance from the obstacle is 3 meters. After setting the safety distance, the obstacle avoidance method also needs to be set by modifying the parameter AVOID_BEHAVE representing the obstacle avoidance method. In this parameter, 0 represents bypass sliding obstacle avoidance and 1 represents braking obstacle avoidance. Since this system is considered to be used in bridge tunnels, bypass sliding obstacle avoidance is selected and this parameter is changed to 0. It can also be modified at any time according to different application environments.

[0079] Furthermore, in essence, the Raspberry Pi is a microcomputer. Different from ordinary computers, the Raspberry Pi uses an SD / MicroSD card as its hard disk. In the current 4B version, it uses an ARM Cortex-A7 2.15GHz as the CPU and can be used for relatively complex development applications. It is widely used in intelligent robots. The configurations for Raspberry Pi development include:

[0080]

[0081]

[0082] The lidar is used to perform 3D scanning on the bridge-tunnel structure to obtain distance and shape data, and algorithm deployment is carried out on the Raspberry Pi system to achieve automatic planning of the inspection path of the drone and flight obstacle avoidance control. The infrared rangefinder is used for infrared ranging to obtain temperature and humidity data for the flight control of the drone. To complete mapping, it is necessary to build a mapping environment for the Raspberry Pi, including the installation of the Raspberry Pi system, the configuration of the ROS (Robot Operating System) system under the Ubuntu system, the installation of the software development kit SDK (Software Development Kit) used by the radar, and the compilation of the mapping algorithm, etc. Before starting the system setup, some equipment and software need to be prepared. Due to the particularity of the Raspberry Pi, the SD card needs to be used as a hard disk, and the flashed system and files need to be stored in it. In this system, a 32G capacity SD card, a card reader, a 5V voltage power cord, and the Windows system software balenaEtcher are used to burn the system for the Raspberry Pi. This system uses the Ubuntu18.04 server system. After confirming that the SD card is recognized by the computer, we need to format it to ensure that there are no other files on the card. Then we need to open balenaEtcher, select the downloaded system and then burn it. After completing the system burn, the SD card can be inserted back into the Raspberry Pi, and at this time, the power cord can be connected to turn it on. After turning it on, first, we need to connect to the network to install the ROS system. First, we need to open the console and enter the following code in the console to complete the update of the source list.

[0083] Furthermore, to complete mapping, it is necessary to build a mapping environment for the Raspberry Pi, including the installation of the Raspberry Pi system, the configuration of the ROS (Robot Operating System) system under the Ubuntu system, the installation of the software development kit SDK (Software Development Kit) used by the radar, and the compilation of the mapping algorithm, etc. The compilation of the mapping environment will be introduced in detail below.

[0084] (1) Burning of the Ubuntu system

[0085] Before starting the system setup, some equipment and software need to be prepared. Due to the particularity of the Raspberry Pi, the SD card needs to be used as a hard disk, and the flashed system and files need to be stored in it. Therefore, in this system, a 32G capacity SD card, a card reader, a 5V voltage power cord, and the Windows system software balenaEtcher are used to burn the system for the Raspberry Pi. This system uses the Ubuntu18.04 server system. After confirming that the SD card is recognized by the computer, we need to format it to ensure that there are no other files on the card. Then we need to open balenaEtcher, select the downloaded system and then burn it.

[0086] (2) ROS System Setup

[0087] After completing the system flashing, insert the SD card back into the Raspberry Pi. Then, connect the power supply cable to turn it on. After powering on, we first need to connect to the network to install the ROS system. First, open the console and enter the following code in the console to complete the update of the source list.

[0088]

[0089] Start installing ROS.

[0090]

[0091] Set the environment variables.

[0092]

[0093] Download other functional components.

[0094]

[0095] Perform initialization

[0096]

[0097] Perform update

[0098]

[0099] Finally, enter roscore to check if the installation is successful.

[0100] (3) SDK Installation and Testing

[0101] The official SDK currently provides two types: C++ and ROS nodes. The ROS node form is used in this system, and the C++ version is included in this version. First, we need to compile the working environment.

[0102]

[0103] After completing the compilation, you can start connecting the lidar for testing. At this time, you can open rviz to display the measurement data, and the following picture will appear, indicating that the SDK installation is successful.

[0104]

[0105] Mapping System Lidar Test

[0106] (4) SLAM Mapping Algorithm HectorSLAM

[0107] In this system, the HectorSLAM algorithm is used for mapping. This algorithm is a SLAM algorithm without an odometer. It uses the Gauss_Newton method to solve the scan-matching problem and obtains the rigid body transformation of the laser point set mapped to the existing map. Therefore, the advantage of this algorithm is that it can optimize the laser beam lattice using the map created from the already acquired data to estimate the representation of the laser points in the map and the probability of occupying the grid. So it is applicable to the construction of 2D maps in uneven areas and is theoretically feasible for mapping drones. However, this algorithm has high hardware requirements, requires a lidar with a high refresh rate, and needs to control the speed of the payload platform during mapping to ensure an ideal mapping effect.

[0108] The most core part of this algorithm is to construct an error function for the current frame by combining the existing map, and use the Gauss_Newton method to obtain the optimal and deviation amounts to achieve the conversion of laser points to the grid map.

[0109] When deploying on a Raspberry Pi, it is first necessary to install the dependent libraries.

[0110]

[0111] After completing the installation of the dependent libraries, the HectorSLAM source code can be downloaded.

[0112]

[0113] When the source code download is completed, compilation can be carried out.

[0114]

[0115] After successful compilation, the mapping system can be tested by turning on the node to activate the A1 lidar.

[0116]

[0117] Start mapping.

[0118]

[0119] Furthermore, the ground base station includes a receiving module, an image processing module, a monitoring module, and a deep learning intelligent inspection module. The deep learning intelligent inspection module is mainly used to label the collected images, that is, to mark normal and disease types, and use the custom labels as dataset samples for model training. The image processing module detects bridge and tunnel diseases, identifies diseases through the images obtained by the drone for the ontology. The deep learning intelligent inspection module builds a deep learning neural network environment and optimizes disease recognition through the YOLO V8 algorithm. The receiving module is used to receive the inspection image data sent by the transmission module; the image processing module analyzes and discriminates potential diseases based on the returned inspection data; the monitoring module is used to monitor the flight status of the inspection drone in real time. The drone is equipped with a 360-degree panoramic camera to conduct inspections to obtain on-site image data or video data, and synchronously transmit the image data for processing to determine whether there are disease fault points in the image. If there are diseases, the disease locations are located, and the disease patterns are stored in the disease detection database. If there are no diseases, the inspection operation continues.

[0120] Furthermore, the drone body is provided with arm rotors, an infrared rangefinder, and a lidar. The flight control system and the Raspberry Pi system are both electrically connected to the arm rotors and the infrared rangefinder on the drone body. The Raspberry Pi system is electrically connected to the lidar of the drone body. The receiving module is used to receive the inspection image data sent by the transmission module; the image processing module analyzes and discriminates potential diseases based on the returned inspection data; the monitoring module is used to monitor the flight status of the inspection drone in real time. The drone is equipped with a 360-degree panoramic camera to conduct inspections to obtain on-site image data or video data, and synchronously transmit the image data for processing to determine whether there are disease fault points in the image. If there are diseases, the disease locations are located, and the disease patterns are stored in the disease detection database. If there are no diseases, the inspection operation continues. The 360-degree panoramic camera and the lidar are transmitted through the Raspberry Pi system of the drone body, so as to better fuse the data and improve the comprehensive understanding of the bridge and tunnel environment. The whole composed of multiple groups of monitoring data can improve the accuracy and integrity of the data through the integration of the Raspberry Pi system.

[0121] Furthermore, the drone body is provided with a binocular camera and a 360-degree panoramic camera. The camera system and the storage system are electrically connected to the binocular camera and the 360-degree panoramic camera of the drone. The 360-degree panoramic camera is used for 360-degree panoramic image acquisition during inspection operations, obtains image data, and transmits the data to the ground control center.

[0122] Furthermore, a communication module is provided on the UAV body. The flight control system, the gimbal system, the camera system, the Raspberry Pi system, the storage system, and the power system are electrically connected to the ground base station through the communication module. The wireless communication module is used for two-way wireless communication with the ground station, receiving instructions from the ground control center, and sending the status and data of the UAV to the ground station.

[0123] Furthermore, the deep learning intelligent inspection module is electrically connected to the camera system, the Raspberry Pi system, and the storage system

[0124] Furthermore, the UAV device used in the intelligent inspection system for bridge and tunnel diseases includes a UAV body, flight arms, an infrared rangefinder, a binocular camera, a frame, a wireless communication module, a 360-degree panoramic camera, a Raspberry Pi device, and a lidar. This UAV device cooperates with the intelligent inspection system for bridge and tunnel diseases to achieve the inspection task of bridges and tunnels.

[0125] Working principle: Through the lidar and the Raspberry Pi system, the risk of collision during UAV inspection can be effectively reduced. The lidar is an active sensor. When working, it emits a laser beam, and then receives the reflected or scattered spectral signals to determine the distance between surrounding objects and the sensor, as well as the contours of objects in the surrounding environment, etc. The lidar system conducts three-dimensional modeling of the bridge and tunnel inspection scene, perceives the complex inspection scene in real time, and realizes automatic obstacle avoidance and reasonably plans the automatic inspection flight path through the Raspberry Pi system; Secondly, the 360-degree panoramic image acquisition can record the structure, hidden danger status, etc. of the bridge and tunnel inspection scene in real time, high definition, and all-round, which is convenient for management, maintenance, and safety inspection. It can also generate a panoramic model to provide visual reference and assistance for the design, renovation, and repair of bridges and tunnels; Finally, based on the deep learning intelligent inspection module, the combination of panoramic image acquisition and disease detection intelligent algorithm realizes intelligent disease detection, improves the efficiency and accuracy of inspection, reduces labor costs and risks, and effectively reduces losses caused by facility failures.

[0126] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference numerals in the claims should not be regarded as limiting the claims involved.

Claims

1. The intelligent UAV detection system for bridge and tunnel defects includes a UAV body and a ground base station, and is characterized by: The drone body is internally provided with a flight control system, a pan / tilt system, a camera system, a Raspberry Pi system, a storage system and a power supply system. The drone body bridge and tunnel inspection method includes the following operation steps: S1. Input the inspection task parameters, including the inspection task location, length, width, and height information, at the ground base station control center and send them to the drone; S2. After receiving the mission parameters, the UAV starts the laser radar and infrared rangefinder to determine the initial position and attitude of the UAV, and plans the inspection flight trajectory of the UAV according to the mission parameters; S3. The drone performs the inspection task according to the flight trajectory, and at the same time starts the 360-degree panoramic camera to collect images and transmit the image data to the ground control center in real time; S4. During the flight, the drone continuously uses laser radar and infrared rangefinder to scan and measure, obtain the surrounding distance, shape, temperature and humidity data, and adjust the flight speed, altitude and direction parameters of the drone according to the data to achieve intelligent obstacle avoidance; S5. After receiving the image data, the ground base station control center uses the deep learning intelligent inspection module to analyze and process the image data, identify the type, location, size, and degree of the disease, store the disease information in the database, and generate an inspection report; S6: After the drone completes the inspection mission, it returns to the starting position and ends the flight; S7, the ground base station control center performs image processing on the photo, the specific steps are: S71, the ground base station control center receives the image data through the receiving module and loads it into the image processing module; S72, the image processing module detects bridge and tunnel defects through images, and builds a deep learning neural network environment in combination with the deep learning intelligent inspection module. It uses the YOLO V8 algorithm to optimize defect recognition and improve the ability to distinguish whether the defect is cracks, leaks or peeling; S73, extracting features of the identified disease types, including the location, length, width, area, and severity of the disease; S74. Combine the data of bridge and tunnel mileage, section location, and water leakage type to determine whether it affects driving and make corresponding choices to improve the damage information; S75. Mark the disease in combination with the auxiliary information; S76. Output the marked final information and store it in the storage system at the same time to classify the lining severity level, the cause of the disease, and the disease prevention and control measures to generate an inspection report.

2. The intelligent drone detection system for bridge and tunnel defects according to claim 1 is characterized by: The ground base station includes a receiving module, an image processing module, a monitoring module and a deep learning intelligent inspection module. The deep learning intelligent inspection module is mainly used to annotate the collected images, that is, to mark the normal and disease types, and use the customized labels as data set samples for model training. The image processing module detects bridge and tunnel diseases and identifies diseases through images obtained by drones. The deep learning intelligent inspection module builds a deep learning neural network environment, and the YOLO V8 algorithm optimizes disease identification.

3. The intelligent bridge and tunnel defect detection system using unmanned aerial vehicles according to claim 1 is characterized by: The drone body is provided with an arm rotor, an infrared rangefinder and a laser radar. The flight control system and the Raspberry Pi system are electrically connected to the arm rotor and the infrared rangefinder on the drone body. The Raspberry Pi system is electrically connected to the laser radar of the drone body.

4. The intelligent drone detection system for bridge and tunnel defects according to claim 1 is characterized by: The drone body is provided with a binocular camera and a 360-degree panoramic camera, and the camera system and the storage system are electrically connected to the binocular camera and the 360-degree panoramic camera of the drone.

5. The intelligent drone detection system for bridge and tunnel defects according to claim 1 is characterized by: The drone body is provided with a communication module, and the flight control system, the gimbal system, the camera system, the Raspberry Pi system, the storage system and the power supply system communicate with the ground base station through the communication module.

6. The intelligent drone detection system for bridge and tunnel defects according to claim 1 is characterized by: The deep learning intelligent inspection module is electrically connected to the camera system, Raspberry Pi system, and storage system.

7. The intelligent drone detection system for bridge and tunnel defects according to claim 1 is characterized by: The drone device used in the bridge and tunnel disease drone intelligent detection system includes a drone body, a flight arm, an infrared rangefinder, a binocular camera, a frame, a wireless communication module, a 360-degree panoramic camera, a Raspberry Pi device and a laser radar.

Images