A method and system for unmanned aerial vehicle shield tunnel inspection

Abstract

This invention discloses a UAV-based shield tunnel inspection system and method, applied in the field of tunnel inspection technology. By manually specifying the three-dimensional coordinate axes of the shield tunnel profile and combining this with UAV automatic navigation and computational vision, the system automates UAV tunnel inspection. The invention utilizes an onboard microcomputer to reproduce the manually set coordinate system. Simultaneously, it uses the tunnel environment and the distance from the tunnel wall to calculate the coordinates of the UAV's position within the tunnel cross-section coordinate system. Furthermore, it uses the Sobel edge detection 3D method to identify the standard lining rings of the shield tunnel, thereby calculating the axial distance, establishing a 3D model of the tunnel, and determining the UAV's travel distance within the tunnel. Finally, it uses Gaussian filtering and computational vision algorithms such as the Sobel edge detection algorithm to analyze and identify tunnel damage in the image information. This invention has advantages such as low cost, high efficiency, and high automation, and can be widely applied to tunnel inspection.

Description

Technical Field

[0001] This invention relates to the field of tunnel inspection technology, and more specifically to a method and system for inspecting unmanned aerial vehicle (UAV) shield tunnels. Background Technology

[0002] In recent years, the mileage of tunnels has been increasing, which has led to a heavy workload for tunnel engineering quality inspection. Whether it's a newly built tunnel or a large-scale transportation project, high-precision testing is required to ensure its normal operation during construction. Manual tunnel exploration not only consumes a large amount of manpower and resources, but also involves risks and uncertainties during the process.

[0003] With the continuous development of unmanned aerial vehicle (UAV) technology, its exceptional maneuverability and controllability have garnered widespread attention. Various exploration companies are turning their attention to UAVs, and their application in tunnel exploration is gradually becoming mainstream. UAVs can be equipped with various sensors and devices to autonomously explore tunnel environments inaccessible to humans, significantly improving work efficiency and quality. Compared to traditional flaw detection vehicles, UAVs are much cheaper, requiring only a small number of operators and data processing personnel to conduct long-range exploration while acquiring more information at a lower cost—a key attraction for many companies. This approach reduces the workload of tunnel exploration while ensuring safety and minimizing accidents during the process.

[0004] However, the complex working environment inside tunnels presents significant challenges to the operation of unmanned aerial vehicles (UAVs). Once deep inside a tunnel, GPS and other satellite signals are blocked, making it impossible for most UAVs to complete missions within the tunnel. Furthermore, external interference from rail transit systems, such as electromagnetic interference, also affects their operation.

[0005] Therefore, how to provide an unmanned aerial vehicle (UAV) shield tunnel inspection system and method has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, the present invention provides an unmanned aerial vehicle (UAV) shield tunnel inspection system and method to solve the problems in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] On the one hand, this invention discloses a method for inspecting unmanned aerial vehicles (UAVs) tunnel boring machines, the specific steps of which are as follows:

[0009] The aircraft path is initially set by inputting a preset three-dimensional coordinate system, and the flight path is adjusted using environmental perception information, while the two-dimensional information of the tunnel cross-section is obtained.

[0010] Based on the standardized production characteristics of shield tunnel lining rings, each shield tunnel lining ring is identified. By combining the two-dimensional information of the tunnel cross-section with the axial flight mileage, the positioning information of each lining ring is superimposed to establish a preliminary three-dimensional model.

[0011] Collect image information inside the tunnel, combine it with surrounding environmental perception information, and improve the model using Gaussian filtering and Sobel edge detection algorithms;

[0012] Based on the principles of computer vision, various types of damage and their degree are identified.

[0013] Preferably, in the above-mentioned UAV shield tunnel inspection method, the flight path is adjusted using environmental perception information, and the specific steps are as follows:

[0014] The flight path is corrected by radial and tangential distortion coefficients;

[0015] The image is corrected using camera distortion parameters. The formula for radial distortion correction is as follows (1), and the formula for tangential distortion correction is as follows (2):

[0016] (1)

[0017] (2)

[0018] In the formula, Let x be the x-coordinate of any point. Let be the ordinate of any point. Let be the distance between any point and the origin. This is the radial distortion correction factor. This is the tangential distortion correction coefficient.

[0019] Preferably, in the above-mentioned method for inspecting a UAV-based shield tunnel, the specific steps for obtaining two-dimensional information of the tunnel cross-section are as follows: A two-dimensional coordinate system is established on the shield cross-section with the center of the tunnel cross-section as the origin. Based on the distance between the UAV and the bottom of the tunnel and the distance between the UAV and the tunnel walls on both sides, the two-dimensional coordinates of the UAV are obtained through geometric calculation, as shown in the following formula:

[0020] ;

[0021] ;

[0022] In the formula, The horizontal coordinates of the cross section The vertical coordinates of the cross section are... The radius of the standard lining ring. , These represent the distances from the laser rangefinders at the left and right ends of the drone in the forward direction to the right and left walls, respectively. This is the distance from the lower laser rangefinder to the lower wall. (See attached image.) Figure 1

[0023] Preferably, in the above-mentioned UAV shield tunnel inspection method, the axial flight mileage is calculated as follows: the number of lining rings passed by the UAV is identified by the Sobel edge detection algorithm;

[0024] ;

[0025] In the formula, The flight distance representing the drone's axis. This represents the number of edges detected by the Sobel edge detection algorithm. This represents the standard width of each standard lining ring.

[0026] Preferably, in the above-mentioned UAV shield tunnel inspection method, the specific steps for improving the model using the Sobel edge detection algorithm are as follows:

[0027] The derivatives of the collected tunnel images are calculated in two directions, and a 3×3 convolution kernel is used to obtain an approximate differential.

[0028] A single convolutional kernel is used to detect horizontal pixel brightness abrupt changes, while a different convolutional kernel is used to detect vertical brightness abrupt changes. The source image is then convolved using these kernels to obtain the "Sobel edge image".

[0029] use and Represent x and y The gradient value in the direction is represented by A and B. x , y Convolution kernel in the direction:

[0030]

[0031] In the formula, Represents the convolution operator. I This represents the input image.

[0032] The final image gradient magnitude G is calculated using the following formula:

[0033]

[0034] The gradient direction is represented by the following formula: .

[0035] Preferably, in the above-mentioned UAV shield tunnel inspection method, the specific steps of improving the model using Gaussian filtering are as follows: Color information in the grayscale color space is removed to convert it into grayscale, while edgeless areas are turned black and edge areas are turned white or other saturated colors; a Gaussian smoothing filter is used, and in the two-dimensional spatial distribution, the Gaussian filtering equation is as follows:

[0036] ;

[0037] In the formula, σ is the standard deviation of the normal distribution.

[0038] Preferably, the above-mentioned UAV shield tunnel inspection method further includes: identifying various types of damage and damage degree based on the principle of computer vision; using color comparison between the damaged part and the tunnel surface to analyze the RGB color gamut to binarize the image and make the damaged part visible; calculating geometric parameters based on pixel information; and locating the damaged part based on the established three-dimensional model.

[0039] On the other hand, this invention discloses a UAV shield tunnel inspection system employing a UAV shield tunnel inspection method, comprising:

[0040] The drone body is equipped with a microcomputer, a lidar, a high-definition camera, and a laser rangefinder. The microcomputer is located inside the drone body, and the lidar is connected to the microcomputer. The lidar detects and senses the environment around the tunnel by emitting light pulses and reflecting electrical pulses in the tunnel, and at the same time provides preliminary feedback on the 3D modeling of the human end.

[0041] The high-definition camera is connected to a microcomputer and is used to take photos and videos during tunnel inspection, collect high-resolution images of the inner wall of the tunnel segments, and provide a preliminary understanding of the tunnel interior in real time for manual personnel, as well as for tunnel damage identification based on computer vision.

[0042] The laser rangefinders are mounted on the left and right ends of the UAV's fuselage and on the sensor brackets at the bottom. By emitting lasers and analyzing and identifying the lasers reflected back to the receiving unit, two-dimensional information of the tunnel cross-section is calculated. A two-dimensional cross-section coordinate system is established with the center of the cross-section as the origin, the horizontal direction of the cross-section as the X-axis, and the vertical direction of the cross-section as the Y-axis. The specific coordinates of the UAV in the cross-section coordinate system are calculated. Then, with the UAV's forward direction as the Z-axis, the Z-axis coordinate is set according to the flight mileage. The cross-section coordinate systems are merged to establish a three-dimensional positioning coordinate system.

[0043] Preferably, the above-mentioned UAV shield tunnel inspection system further includes a lighting device, which is installed on the UAV body.

[0044] As can be seen from the above technical solution, compared with the prior art, this invention discloses a method and system for inspecting shield tunnels using unmanned aerial vehicles (UAVs). Since the lining rings of shield tunnels are all standard uniform rings, a specified three-dimensional coordinate system can be manually input into the UAV, along with the corresponding coordinate parameters, to ensure it runs along a predetermined trajectory and meets the detection requirements. Simultaneously, the gaps between the lining rings are used to identify the UAV's boundaries, thereby determining its flight path along the axial direction. Using Gaussian filters and the Sobel edge detection algorithm, accurate positioning can be achieved in complex tunnels; this solves the problem that UAVs cannot perform complete inspections under external interference or signal shielding conditions. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0046] Figure 1 This is a schematic diagram of the two-dimensional coordinate structure calculation of the present invention.

[0047] Figure 2 This is a schematic diagram of the three-dimensional coordinate system described in this invention.

[0048] Figure 3 This is a flowchart of the method described in this invention. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] This invention discloses a method for inspecting shield tunnels using unmanned aerial vehicles (UAVs), the specific steps of which are as follows:

[0051] The aircraft path is initially set by inputting a preset three-dimensional coordinate system, and the flight path is adjusted using environmental perception information, while the two-dimensional information of the tunnel cross-section is obtained.

[0052] Based on the standardized production characteristics of shield tunnel lining rings, each shield tunnel lining ring is identified. By combining the two-dimensional information of the tunnel cross-section with the axial flight mileage, the positioning information of each lining ring is superimposed to establish a preliminary three-dimensional model.

[0053] Collect image information inside the tunnel, combine it with surrounding environmental perception information, and improve the model using Gaussian filtering and Sobel edge detection algorithms;

[0054] Based on the principles of computer vision, various types of damage and their degree are identified.

[0055] It also includes: constructing a three-dimensional model of the tunnel based on the identified types and degrees of damage, and visualizing the damage.

[0056] Because the camera's shooting angle changes during flight, and there is also shaking, there will be some difference between the virtual and real images. Distortion parameters can be used to correct the image, improving image accuracy and reproducing the true damage as accurately as possible. Therefore, environmental perception information is used to adjust the flight path. The specific steps are as follows:

[0057] The flight path is corrected by radial and tangential distortion coefficients;

[0058] The image is corrected using camera distortion parameters. The formula for radial distortion correction is as follows (1), and the formula for tangential distortion correction is as follows (2):

[0059] (1)

[0060] (2)

[0061] In the formula, Let x be the x-coordinate of any point. Let be the ordinate of any point. Let be the distance between any point and the origin. This is the radial distortion correction factor. This is the tangential distortion correction coefficient.

[0062] To address the issue of lacking GPS positioning signals, this invention employs a three-dimensional coordinate system and the Sobel edge detection algorithm to determine the drone's position and flight distance within the tunnel. Before or during flight, a two-dimensional coordinate system is manually established on the shield tunnel section, with the center of the tunnel section as the origin. Laser rangefinders mounted on the left, right, and bottom of the drone continuously emit lasers during flight. These lasers, after colliding with the tunnel walls, are reflected back to the drone's receiver, providing the distances between the drone and the tunnel bottom and sides. The two-dimensional coordinates of the drone are then obtained through geometric calculations, as shown in the following formula:

[0063] ;

[0064] ;

[0065] In the formula, The horizontal coordinates of the cross section The vertical coordinates of the cross section are... The radius of the standard lining ring. , These represent the distances from the laser rangefinders at the left and right ends of the drone in the forward direction to the right and left walls, respectively. This is the distance from the lower laser rangefinder to the lower wall.

[0066] To further optimize the above technical solution, the axial flight distance is calculated as follows: the number of lining rings passed by the UAV is identified by the Sobel edge detection algorithm;

[0067] ;

[0068] In the formula, The flight distance representing the drone's axis. This represents the number of edges detected by the Sobel edge detection algorithm. This represents the standard width of each standard lining ring.

[0069] To further optimize the above technical solution, the specific steps for improving the model using the Sobel edge detection algorithm are as follows:

[0070] For damage identification and analysis, the Sobel edge detection algorithm was adopted. This algorithm demonstrates strong optimization capabilities when handling images with gradual grayscale changes and high noise levels. The Sobel operator provides accurate edge localization and is computationally fast. Because the image is two-dimensional, differentiation is only required in two directions: vertical edges have larger gradient magnitudes in the horizontal direction, and horizontal edges have larger gradient magnitudes in the vertical direction. Therefore, a 3×3 convolution kernel can be used to obtain approximate derivatives. Abrupt changes in pixel brightness in the horizontal direction can be detected using one convolution kernel, while abrupt changes in brightness in the vertical direction can be detected using another. Convolving the source image with these kernels yields the "Sobel edge image."

[0071] use and Represent x and y The gradient value in the direction is represented by A and B. x , y Convolution kernel in the direction:

[0072]

[0073] In the formula, Represents the convolution operator. I This represents the input image.

[0074] The final image gradient magnitude G is calculated using the following formula:

[0075]

[0076] The gradient direction is represented by the following formula: .

[0077] Because the Sobel algorithm directly uses thresholds to judge edge points, it can lead to excessive false positives due to noise. To maximize image fidelity, this invention also employs a Gaussian filtering method for image denoising, gradient calculation, and non-maximum suppression (NMS) on edges. Double thresholding is then used on detected edges to remove false positives. Finally, all edges and their connections are analyzed to preserve true edges and eliminate insignificant ones. In computer vision, color information in the grayscale color space is removed, converting it to grayscale, turning edgeless areas black, and edge areas white or other saturated colors. This method uses a Gaussian smoothing filter, which effectively suppresses noise under a normal distribution. Since the Sobel algorithm follows a normal distribution for images under varying brightness, this method exhibits strong robustness.

[0078] Using a Gaussian smoothing filter, the Gaussian filtering equation for a two-dimensional spatial distribution is as follows:

[0079] ;

[0080] In the formula, σ is the standard deviation of the normal distribution.

[0081] On two planes, the contour lines obtained using this formula form a standard concentric circle, expanding outwards from a single point. A convolutional array of non-zero pixels is then used to transform the image into its original form. The value of each pixel is a weighted average of its neighboring pixel values. Because the Gaussian distribution value of the initial point is the largest, its weight is the largest, while the weights of adjacent pixels decrease with distance from the initial point. Compared to other equalized blur filtering algorithms, the algorithm proposed in this embodiment can better preserve edge effects in the image.

[0082] After the 3D model is completed, the drone uses algorithms to identify and detect damage such as gaps, records the coordinates of the gaps in the model in a microcomputer, and then the microcomputer reconstructs them into the 3D tunnel model to intuitively reflect the damaged area and realize the function of reporting the location and condition of damage.

[0083] To further optimize the above technical solution, the method also includes: identifying various types of damage and their degree based on the principles of computer vision; using color comparison between the damaged area and the tunnel surface to analyze the RGB color gamut to binarize the image and make the damaged area visible; calculating geometric parameters based on pixel information; and locating the damaged area based on the established 3D model.

[0084] Another embodiment of the present invention discloses a UAV shield tunnel inspection system employing a UAV shield tunnel inspection method, comprising:

[0085] The drone body is equipped with a microcomputer, a lidar, a high-definition camera, and a laser rangefinder. The microcomputer is located inside the drone body, and the lidar is connected to the microcomputer. The lidar detects and senses the environment around the tunnel by emitting light pulses and reflecting electrical pulses in the tunnel, and at the same time provides preliminary feedback on the 3D modeling of the human end.

[0086] The high-definition camera is connected to a microcomputer and is used to take photos and videos during tunnel inspection, collect high-resolution images of the inner wall of the tunnel segments, and provide a preliminary understanding of the tunnel interior in real time for manual personnel, as well as for tunnel damage identification based on computer vision.

[0087] The laser rangefinders are mounted on the left and right ends of the UAV's fuselage and on the sensor brackets at the bottom. By emitting lasers and analyzing and identifying the lasers reflected back to the receiving unit, two-dimensional information of the tunnel cross-section is calculated. A two-dimensional cross-section coordinate system is established with the center of the cross-section as the origin, the horizontal direction of the cross-section as the X-axis, and the vertical direction of the cross-section as the Y-axis. The specific coordinates of the UAV in the cross-section coordinate system are calculated. Then, with the UAV's forward direction as the Z-axis, the Z-axis coordinate is set according to the flight mileage. The cross-section coordinate systems are merged to establish a three-dimensional positioning coordinate system.

[0088] To further optimize the above technical solution, a lighting device is also included, which is installed on the drone body.

[0089] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0090] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for inspecting unmanned aerial vehicles (UAVs) tunnel boring machines, characterized in that, The specific steps are as follows: The aircraft path is initially set by inputting a preset three-dimensional coordinate system, and the flight path is adjusted using environmental perception information, while the two-dimensional information of the tunnel cross-section is obtained. Based on the standardized production characteristics of shield tunnel lining rings, each shield tunnel lining ring is identified. By combining the two-dimensional information of the tunnel cross-section with the axial flight mileage, the positioning information of each lining ring is superimposed to establish a preliminary three-dimensional model. Collect image information inside the tunnel, combine it with surrounding environmental perception information, and improve the model using Gaussian filtering and Sobel edge detection algorithms; Based on the principles of computer vision, various types of damage and their degree are identified; The specific steps for obtaining two-dimensional information of the tunnel cross-section are as follows: Establish a two-dimensional coordinate system on the shield tunnel cross-section with the center of the tunnel cross-section as the origin. Based on the distance between the drone and the bottom of the tunnel and the distance between the drone and the tunnel walls on both sides, obtain the two-dimensional coordinates of the drone through geometric calculation, as shown in the following formula: ； ; In the formula, The horizontal coordinates of the cross-section The vertical coordinates of the cross-section The radius of the standard lining ring. , These represent the distances from the laser rangefinders at the left and right ends of the drone in the forward direction to the right and left walls, respectively. This is the distance from the lower laser rangefinder to the lower wall; The number of lining rings traversed by the drone was identified using the Sobel edge detection algorithm; ; In the formula, The flight distance representing the drone's axis. This represents the number of edges detected by the Sobel edge detection algorithm. This represents the standard width of each standard lining ring.

2. The method for inspecting a shield tunnel using an unmanned aerial vehicle (UAV) according to claim 1, characterized in that, The specific steps for adjusting the flight path using environmental perception information are as follows: The flight path is corrected by radial and tangential distortion coefficients; The image is corrected using camera distortion parameters. The formula for radial distortion correction is as follows (1), and the formula for tangential distortion correction is as follows (2): （1） （2） In the formula, Let x be the x-coordinate of any point. Let be the ordinate of any point. Let be the distance between any point and the origin. This is the radial distortion correction factor. This is the tangential distortion correction coefficient.

3. The method for inspecting a shield tunnel using an unmanned aerial vehicle (UAV) according to claim 1, characterized in that, The specific steps for improving the model using the Sobel edge detection algorithm are as follows: The derivatives of the collected tunnel images are calculated in two directions, and a 3×3 convolution kernel is used to obtain an approximate differential. A single convolutional kernel is used to detect horizontal pixel brightness abrupt changes, while a separate convolutional kernel is used to detect vertical brightness abrupt changes. The source image is then convolved using these kernels to obtain the Sobel edge image. use and Represent x and y The gradient value in the direction is represented by A and B. x , y Convolution kernel in the direction: In the formula, Represents the convolution operator. I Indicates the input image; The final image gradient magnitude G is calculated using the following formula: The gradient direction is represented by the following formula: .

4. The method for inspecting a shield tunnel using an unmanned aerial vehicle (UAV) according to claim 3, characterized in that, The specific steps for improving the model using Gaussian filtering are as follows: Color information in the grayscale color space is removed, thus converting it to grayscale, while edgeless regions are turned black, and edge regions are turned white or other saturated colors; a Gaussian smoothing filter is used, and in a two-dimensional spatial distribution, the Gaussian filtering equation is shown below: ; In the formula, σ is the standard deviation of the normal distribution.

5. The method for inspecting a shield tunnel using an unmanned aerial vehicle (UAV) according to claim 1, characterized in that, Also includes: Based on the principles of computer vision, various types of damage and their degree are identified. By comparing the colors of the damaged area with those of the tunnel surface, the RGB color gamut is analyzed to binarize the image and make the damaged area visible. Geometric parameters are calculated based on pixel information. The damaged area is located based on the established 3D model.

6. A UAV shield tunnel inspection system employing the UAV shield tunnel inspection method as described in any one of claims 1-5, characterized in that, include: The drone body is equipped with a microcomputer, a lidar, a high-definition camera, and a laser rangefinder. The microcomputer is located inside the drone body, and the lidar is connected to the microcomputer. The lidar detects and senses the environment around the tunnel by emitting light pulses and reflecting electrical pulses in the tunnel, and at the same time provides preliminary feedback on the 3D modeling of the human end. The high-definition camera is connected to a microcomputer and is used to take photos and videos during tunnel inspection, collect high-resolution images of the inner wall of the tunnel segments, and provide a preliminary understanding of the tunnel interior in real time for manual personnel, as well as for tunnel damage identification based on computer vision. The laser rangefinders are mounted on the left and right ends of the UAV's fuselage and on the sensor brackets at the bottom. By emitting lasers and analyzing and identifying the lasers reflected back to the receiving unit, two-dimensional information of the tunnel cross-section is calculated. A two-dimensional cross-section coordinate system is established with the center of the cross-section as the origin, the horizontal direction of the cross-section as the X-axis, and the vertical direction of the cross-section as the Y-axis. The specific coordinates of the UAV in the cross-section coordinate system are calculated. Then, with the UAV's forward direction as the Z-axis, the Z-axis coordinate is set according to the flight mileage. The cross-section coordinate systems are merged to establish a three-dimensional positioning coordinate system.

7. The UAV shield tunnel inspection system according to claim 6, characterized in that, Also includes: A lighting device is mounted on the drone body.

Images