Tunnel Lining Subtle Crack Detection Method and System Based on Data Fusion

Abstract

The present invention discloses a method and system for detecting fine cracks in tunnel linings based on data fusion, which relates to the technical field of tunnel disease detection. The method includes the steps of: acquiring three-dimensional point cloud data and images of the tunnel lining and performing preprocessing; using a deep learning model to perform preliminary feature screening on the images to obtain suspected crack area images; performing densification processing on the three-dimensional point cloud data to obtain high-density depth information and obtaining a depth map; analyzing the spatial structure perception provided by the depth map, and using a data matching algorithm to fuse the suspected crack area images with the depth map to obtain fused data; calculating the fusion confidence of the fused edge data, and determining the crack detection result according to the fusion confidence. The present invention combines the high-precision depth information of the laser with the rich visual features of the camera to fully address the detection difficulties such as fine cracks, environmental noise interference, and variable crack morphologies.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel disease detection, and particularly to a method and system for detecting fine cracks in tunnel linings based on data fusion. Background Technique

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

[0003] The main diseases of tunnels include: lining damage, tunnel water damage, tunnel frost damage, seismic damage, fire, etc. Among them, lining damage is one of the typical and common diseases of tunnels, and its manifestation form is mostly lining cracks. Therefore, the detection of tunnel lining cracks is one of the most difficult projects in traffic engineering quality detection.

[0004] In the detection of tunnel lining cracks, the accurate identification and positioning of the target are crucial for ensuring structural safety and extending service life. However, the actual detection work faces multiple challenges. First, the cracks are usually extremely fine and have a low contrast with the overall color and texture of the lining, making it difficult to clearly distinguish their edges in the image; this problem is particularly prominent in traditional image processing algorithms. Second, the environmental conditions in the tunnel are complex, such as low light, high humidity, dust and impurities, etc., which may all lead to an increase in image noise collected by the camera and also have an adverse impact on the signal stability of auxiliary sensors such as lasers or ultrasonic waves. Finally, the crack morphology is variable, not only may it present fine and linear features, but also bifurcations, intersections or irregular edges may occur, and this complexity brings great difficulties to automated detection and model fitting.

[0005] Considering the above problems, how to achieve high-precision and robust crack detection through sensor data technology under the conditions of low contrast, harsh environment and complex crack morphology has become a key technical problem to be solved urgently at present. Summary of the Invention

[0006] Aiming at the deficiencies of the existing technology, the purpose of the present invention is to provide a method and system for detecting fine cracks in tunnel linings based on data fusion, which combines the high-precision depth information of lasers with the rich visual features of cameras to fully address the detection difficulties such as fine cracks, environmental noise interference and variable crack morphology.

[0007] To achieve the above purpose, the present invention is implemented through the following technical solutions:

[0008] The first aspect of the present invention provides a method for detecting fine cracks in tunnel linings based on data fusion, including the following steps:

[0009] Obtain the three-dimensional point cloud data and images of the tunnel lining and perform preprocessing;

[0010] Use a deep learning model to perform preliminary feature screening on the image to obtain an image of the suspected crack area;

[0011] Densify the 3D point cloud data to obtain high-density depth information and get a depth map;

[0012] Analyze the spatial structure perception provided by the depth map, and use a data matching algorithm to fuse the image of the suspected crack area with the depth map to obtain fused data;

[0013] Calculate the fusion confidence of the fused edge data, and determine the crack detection result according to the fusion confidence.

[0014] Furthermore, use a laser scanner to collect 3D point cloud data and use a camera to obtain images.

[0015] Even further, the specific steps of preprocessing include:

[0016] Obtain the internal parameter matrix of the camera and the external parameter matrix between the laser scanner and the camera through calibration;

[0017] Use the parameters obtained by calibration to project the 3D point cloud data collected by the laser scanner onto the camera image plane to achieve spatial alignment of different modality data.

[0018] Furthermore, the specific steps of using a deep learning model to perform preliminary feature screening on the image are:

[0019] Adopt a method of reducing the confidence threshold, extract the suspicious areas, establish a confidence distribution based on detection, analyze the suspicious areas according to the confidence values, and apply a dynamic filtering strategy to different categories of suspicious areas;

[0020] At the same time, obtain the 2D bounding box of each suspicious area to get an image of the suspected crack area.

[0021] Even further, the specific steps of densifying the 3D point cloud data are:

[0022] Construct a Poisson equation according to the 3D point cloud data;

[0023] Solve the Poisson equation for surface reconstruction to obtain a continuous depth map;

[0024] Based on the spatial alignment of different modality data, extract the area in the generated depth map corresponding to the suspicious area of the image to obtain high-density depth information.

[0025] Furthermore, the specific steps of analyzing the spatial structure perception provided by the depth map and using a data matching algorithm to fuse the image of the suspected crack area with the depth map are:

[0026] Use the edge detection algorithm to detect the image of the suspected crack area and obtain the first detection result;

[0027] By calculating the gradient depth between the depth image pixels, obtain the set of points with significantly discontinuous depth values as the second detection result;

[0028] Use the data matching algorithm to perform cross-modal data matching and fusion on the first detection result and the second detection result to obtain the fused data.

[0029] Furthermore, the specific steps of using the edge detection algorithm to detect the image of the suspected crack area are as follows:

[0030] Calculate the local contrast of the image of the suspected crack area;

[0031] Through the spatial structure perception provided by the depth map, combined with the detailed features of the image gray distribution, calculate the depth gradient using the depth map and normalize the depth gradient to obtain the enhancement factor;

[0032] Use the enhancement factor to weight the local contrast of the image to obtain the enhanced image;

[0033] Use the edge detection algorithm to perform edge detection on the enhanced image to obtain the first detection result.

[0034] The second aspect of the present invention provides a tunnel lining fine crack detection system based on data fusion, including:

[0035] A data acquisition module, configured to acquire the three-dimensional point cloud data and images of the tunnel lining and perform preprocessing;

[0036] A preliminary screening module, configured to use a deep learning model to perform preliminary feature screening on the image to obtain the image of the suspected crack area;

[0037] A depth information extraction module, configured to perform densification processing on the three-dimensional point cloud data to obtain high-density depth information and obtain a depth map;

[0038] A data fusion module, configured to analyze the spatial structure perception provided by the depth map and use the data matching algorithm to fuse the image of the suspected crack area with the depth map to obtain the fused data;

[0039] A crack detection module, configured to calculate the fusion confidence of the fused edge data and determine the crack detection result according to the fusion confidence.

[0040] The third aspect of the present invention provides a medium, on which a program is stored, and when the program is executed by a processor, it implements the steps in the method for detecting fine cracks in tunnel linings based on data fusion as described in the first aspect of the present invention.

[0041] A fourth aspect of the present invention provides a device, including a memory, a processor, and a program stored on the memory and executable on the processor. When the processor executes the program, the steps in the method for detecting minute cracks in tunnel linings based on data fusion as described in the first aspect of the present invention are implemented.

[0042] The above one or more technical solutions have the following beneficial effects:

[0043] The present invention discloses a method and system for detecting minute cracks in tunnel linings based on data fusion, including a process of rough extraction of the region of interest (ROI) and generation of a depth map. Data is extracted from two different modalities, namely three-dimensional point cloud data and camera images, and an edge and discontinuity point detection method is used to enhance crack features. Finally, through a multi-modal data fusion and confidence calculation strategy, accurate identification of cracks in tunnel linings is achieved, providing a scientific basis for subsequent structural health assessment and maintenance.

[0044] Advantages of additional aspects of the present invention will be given in part in the following description, become apparent in part from the following description, or be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings, which form a part of this specification, are used to provide a further understanding of the present invention. The schematic embodiments and descriptions thereof of the present invention are used to explain the present invention and do not constitute an improper limitation to the present invention.

[0046] Figure 1 It is a flowchart of the method for detecting minute cracks in tunnel linings based on data fusion in Embodiment 1 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0048] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless otherwise clearly specified in the context, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof;

[0049] Embodiment 1:

[0050] Embodiment 1 of the present invention provides a method for detecting minute cracks in tunnel linings based on data fusion, as Figure 1 shown, including the following steps:

[0051] Step 1: Obtain the 3D point cloud data and images of the tunnel lining and perform preprocessing.

[0052] Step 1.1: In a specific embodiment, first, a laser scanner and a camera that have completed sensor calibration are configured on the tunnel inspection vehicle. The 3D point cloud data is collected using the laser scanner, and the images are obtained using the camera.

[0053] Step 1.2: Preprocess the acquired data.

[0054] Step 1.2.1: Obtain the internal parameter matrix K of the camera and the external parameter matrix T between the laser scanner and the camera through calibration. The external parameter matrix includes a rotation matrix and a translation vector. Among them, the calibration method uses the existing calibration methods in the prior art.

[0055] Step 1.2.2: Use the parameters obtained through calibration to project the 3D point cloud data collected by the laser scanner onto the camera image plane to achieve spatial alignment of different modality data and lay a foundation for subsequent data fusion.

[0056] In a specific embodiment, assume that the 3D points collected by the laser scanner are (homogeneous coordinates), and the internal parameter matrix of the camera is .

[0057] Among them, represents the focal length in the horizontal direction of the image plane (in pixels), represents the focal length in the vertical direction of the image plane (in pixels), represents the horizontal coordinate of the image center point, represents the vertical coordinate of the principal point image center point.

[0058] The external parameter matrix between the laser scanner and the camera is , then the formula for projecting the point cloud data onto the image plane is:

[0059] .

[0060] Get , and after conversion, it is normalized to pixel coordinates .

[0061] Among them, represents the horizontal coordinate of the pixel in the image, usually representing the position in the width direction of the image, represents the vertical coordinate of the pixel in the image, usually representing the position in the height direction of the image, represents the scale factor in the homogeneous coordinates, which is used for homogeneous coordinate representation. When When it is 1, the pixel coordinates are 2D coordinates in the actual sense .

[0062] Step 2: Use a deep learning model to perform preliminary feature screening on the image to obtain an image of the suspected crack area.

[0063] Step 2.1: Adopt the method of reducing the confidence threshold, extract the suspicious areas, establish a confidence distribution based on detection, analyze the suspicious areas according to the confidence values, and apply a dynamic filtering strategy to different categories of suspected areas.

[0064] In a specific implementation, the original image for tunnel lining crack detection faces multiple difficulties. Cracks usually have low contrast and small differences from the tunnel background, especially being easily masked in uneven lighting or low-light environments. Secondly, complex background interferences (such as stains, seams, and light spots) are similar in shape to cracks, easily leading to false detections. In addition, crack features are usually subtle and discontinuous, with widths possibly only a few pixels, making it difficult for traditional methods to effectively capture them. At the same time, non-uniform lighting and dynamic range limitations inside the tunnel further weaken image details.

[0065] To address these difficulties, in this embodiment, based on obtaining sensor data, first, the method of non-local mean filtering is used to remove surface texture noise while retaining crack edge features to reduce complex background interferences. Secondly, the high dynamic range imaging algorithm (HDR) is adopted to adjust the contrast of light and dark areas and enhance the visibility of cracks. After the image is preprocessed, it is passed through existing deep learning models such as the YOLO or Faster R-CNN model for preliminary feature screening.

[0066] Using a deep learning model to perform preliminary crack detection on the tunnel lining in the camera image, as the preliminary feature screening, this embodiment adopts the method of reducing the confidence threshold, thereby extracting more suspicious regions (ROIs). This can cover as many areas as possible where subtle cracks may exist and avoid missing detections due to low target contrast. The deep learning model can adopt existing models such as YOLO or Faster R-CNN. At the same time, a confidence distribution based on detection is established, and the suspicious regions are analyzed according to the confidence values, and a dynamic filtering strategy is applied to different categories of suspected regions. For example: High confidence : Directly retain the target without strict filtering. Medium confidence : Use strict area and shape filtering rules to eliminate possible false positives. Low confidence : Only retain the targets that meet the aspect ratio or shape factor and further process them. In addition, in this embodiment, according to the detection confidence , the threshold of the filtering rule is dynamically adjusted.

[0067] Step 2.2: Simultaneously obtain the two-dimensional bounding boxes of each suspicious area and perform image processing to obtain the suspected crack area image.

[0068] In a specific implementation, the two-dimensional bounding box is:

[0069] 。

[0070] where B is the two-dimensional bounding box, 、 、 and are the minimum abscissa, minimum ordinate, maximum abscissa, and maximum ordinate, respectively.

[0071] According to the position of the two-dimensional bounding box B, crop the corresponding region of interest (ROI) from the original image.

[0072] Convert the cropped ROI into a grayscale image I(x, y). The conversion formula is:

[0073]

[0074] where 、 、 are the pixel values of the red, green, and blue channels, respectively.

[0075] Step 3: Densify the three-dimensional point cloud data to obtain high-density depth information and get a depth map.

[0076] Step 3.1: Construct a Poisson equation based on the three-dimensional point cloud data.

[0077] In a specific implementation, the laser scanner usually outputs sparse point cloud data. To efficiently fuse it with the camera image, the point cloud data needs to be densified. The Poisson reconstruction depth continuity algorithm is used to convert the discrete point cloud into a continuous depth image.

[0078] Specifically, find an indicator function χ such that its gradient is close to the vector field given by the point cloud normal vector field V, and solve the following minimization problem:

[0079] 。

[0080] where represents the gradient of the scalar field χ, which is a vector field indicating the rate of change of χ in all directions.

[0081] Step 3.2: Solve the Poisson equation for surface reconstruction to obtain a continuous depth map.

[0082] The above minimization problem is equivalent to solving the Poisson equation:

[0083] 。

[0084] Among them, represents the Laplacian operator of the scalar field, which is the divergence of the gradient and reflects the "curvature degree" of the scalar field locally. represents the divergence of the vector field V, which is a scalar value and describes the "divergence" or "convergence" degree of V at a certain point. By taking the isosurface χ = 0.5, the reconstructed surface can be obtained, thus forming a continuous depth map D.

[0085] Through densification processing, the point cloud data is resampled and interpolated to form a regularized two-dimensional depth map. This depth map is a two-dimensional matrix, and the value of each pixel in the matrix represents the distance (depth value) from a certain point to the sensor under the corresponding viewing angle, with the unit of centimeter. The depth map is usually the same size as the camera image, and each pixel precisely corresponds to a pixel point on the camera image, having a per-pixel depth resolution.

[0086] Step 3.3: Based on the spatial alignment of different modality data, extract the region in the generated depth map corresponding to the suspicious region of the image to obtain high-density depth information.

[0087] In order to map the two-dimensional bounding box B in the picture to the BBOX in the depth map, the internal parameters (camera projection parameters) and external parameters (rotation and translation matrices) between the camera and the laser scanner obtained in the previous steps are used to establish the mapping relationship between the two coordinate systems. The two-dimensional pixel coordinates of B in the picture are converted to three-dimensional points in the camera coordinate system through the camera internal parameters and depth back-projection. Then, the external parameter matrix is used to map the three-dimensional points in the camera coordinate system to the laser scanner coordinate system to obtain three-dimensional points consistent with the viewing angle of the laser scanner. Next, the three-dimensional points in the laser scanner coordinate system are projected onto the two-dimensional plane of the depth map to obtain the corresponding BBOX region on the depth map. Finally, according to the projection result, a new BBOX is generated in the depth map to represent the position and range of the corresponding region on the depth map. Among them, the BBOX region refers to a rectangular region, usually used to define the boundary of a rectangular region, which is represented by the coordinates of the lower left corner and the upper right corner.

[0088] Then, using the aforementioned calibration parameter process, based on the spatial alignment of different modality data, the region in the generated depth map corresponding to the image ROI is extracted, thereby obtaining the high-density depth information that matches the visual ROI, namely the depth map D. Tunnel lining cracks are usually accompanied by certain depth changes (such as depressions at the crack edges). Through the depth map, these characteristic regions can be accurately located. In low-contrast images, the texture information of the cracks may be blurred, but the depth information can provide supplementation, thereby enhancing the robustness of detection. The high-density depth map can intuitively display the three-dimensional spatial relationship of the scene, providing basic support for subsequent geometric feature extraction (such as normal vector calculation, surface reconstruction) and object recognition. The high-density depth information map plays a key role in the crack detection method by enhancing the detection accuracy, reducing interference, providing supplementary features, and supporting subsequent analysis. It provides information in more dimensions for the detection process, enabling the system to more robustly complete the detection task in complex environments.

[0089] Step 4: Analyze the spatial structure perception provided by the depth map, and use the data matching algorithm to fuse the image of the suspected crack region with the depth map to obtain the fused data. Combining the high-density depth map with the two-dimensional image, the depth map can supplement the deficiencies of the two-dimensional information and effectively handle the target detection and segmentation problems in complex environments.

[0090] On the image, an edge detection algorithm is used to extract the edge features of the image. Since the tunnel crack recognition pictures are usually in low-contrast situations, the algorithm combines the depth information and the local features of the image. The depth information is fused with the gray distribution of the image to enhance the visibility of the crack edges.

[0091] Step 4.1: Use the edge detection algorithm to detect the image of the suspected crack region to obtain the first detection result.

[0092] Step 4.1.1: Calculate the local contrast of the image of the suspected crack region.

[0093] In a specific implementation, the input grayscale image and the corresponding depth map . In the grayscale image, calculate the local contrast , and commonly used Laplace transform or local standard deviation. Taking the local standard deviation as an example:

[0094] .

[0095] Among them, is the local window centered on , is the mean value of the pixels in the window, and N represents the total number of pixel points contained in the local window , that is, the size of the window.

[0096] ​Step 4.1.2: Based on the spatial structure perception provided by the depth map and the detailed features of the image grayscale distribution, the depth gradient is calculated using the depth map and the depth gradient is normalized to obtain the enhancement factor.

[0097] In a specific implementation, the depth gradient is calculated using the depth map. :

[0098] .

[0099] Normalize the depth gradient to get the enhancement factor :

[0100] .

[0101] Step 4.1.3: Use the enhancement factor to weight the local contrast of the image to obtain the enhanced image.

[0102] The fusion process uses the spatial structure perception provided by the depth map and the detailed characteristics of the image grayscale distribution to locally enhance the low-contrast areas, thereby improving the visibility and coherence of the crack edges.

[0103] In a specific implementation, the local contrast of the image is weighted using the enhancement factor to obtain an enhanced image. :

[0104] .

[0105] in, It is a parameter to adjust the depth enhancement strength, and the default value can be set to 0.5.

[0106] In a tunnel environment, cracks usually appear as a sudden change in local depth or a break in continuity, with more dramatic depth changes compared to the surrounding areas. The depth gradient can highlight this change characteristic, thereby making the cracks stand out from other smooth areas. Furthermore, there may be pseudo-edges in the image caused by texture, illumination changes, etc., but these pseudo-edges are usually not accompanied by significant depth changes. By analyzing the depth gradient, these interferences can be effectively eliminated, and only the areas with significant depth changes are retained as candidate cracks.

[0107] In addition, crack features in grayscale images may become blurred when the contrast is low or there is a complex background. Depth gradients can also provide additional boundary information for these areas, complementing the deficiencies in grayscale images and enhancing the overall robustness of crack features. Cracks are usually linear or curvilinear structures, which appear as continuous mutations in depth values on depth maps. Depth gradients can reveal these linear features, thus significantly enhancing the cracks. After normalizing the depth gradients, the enhancement factors can be further used to fuse feature information of different modalities, making the crack areas more prominent in multimodal data and facilitating subsequent precise detection and analysis.

[0108] Step 4.1.4: Use an edge detection algorithm to perform edge detection on the enhanced image to obtain a first detection result. In this embodiment, the edge detection algorithm uses the Canny or Sobel algorithm.

[0109] Step 4.2: By calculating the gradient depth between depth image pixels, obtain a set of points with significantly discontinuous depth values as the second detection result.

[0110] In a specific implementation, on the densified depth map, by calculating the depth gradient values between pixels, a set of points with significantly discontinuous depth values is detected. These discontinuous points usually reflect the positions of crack edges or faults.

[0111] For the densified depth map calculate the gradient between pixels :

[0112] .

[0113] When , mark this pixel as a depth discontinuous point, denoted as , where is the set threshold. The G(i,j) calculated by this formula is the gradient value calculation of the specific pixel position, used for the analysis and processing of per-pixel operations, for the detection of gradient changes between specific points, and directly participates in the discontinuity judgment.

[0114] Step 4.3: Use a data matching algorithm to perform cross-modal data matching and fusion on the first detection result and the second detection result to obtain the fused data.

[0115] In a specific implementation, to suppress noise interference, non-maximum suppression and filtering processing are respectively performed on the two parts of the detection results, namely the first detection result and the second detection result, and then cross-modal matching data fusion is performed through a certain matching algorithm. In this embodiment, the matching algorithm uses Chamfer matching or matching based on the nearest neighbor distance. Taking Chamfer as an example:

[0116] Let the set of image edge points be , and the set of depth discontinuity points be . The Chamfer matching can be used to calculate the matching distance as follows:

[0117] .

[0118] The smaller the matching distance, the more consistent the two-modal edges are, thus improving the credibility of the existence of cracks.

[0119] Step 5: Calculate the fusion confidence of the fused edge data, and determine the crack detection result according to the fusion confidence.

[0120] In a specific implementation, based on the fused edge data, a confidence index is designed, which may include the following elements:

[0121] 1. The coincidence degree (matching accuracy) between the image edge and the depth discontinuity edge.

[0122] 2. The span of the discontinuity points in the depth map (representing the crack width or depth change).

[0123] 3. Edge continuity (cracks are usually continuous linear).

[0124] These indexes are combined into a multi-factor confidence calculation formula. Finally, a preset threshold is obtained through comprehensive analysis and calculation of multiple features. After feature consistency check, when the fusion confidence exceeds the preset threshold, the system determines that there is a crack in the ROI area. The calculation of the fusion confidence is based on the weighted results of the gray distribution, depth features, and edge features. The threshold is set through experimental optimization to balance false detection and missed detection. At the same time, the morphological features of the cracks and the verification of multi-modal data further enhance the reliability of the judgment, laying a foundation for subsequent precise analysis.

[0125] Feature consistency refers to the regular coordination or matching of various features (such as gray level, depth, edge, etc.) in terms of space and nature within the same ROI area. This consistency is a key index used to verify whether the area conforms to the target features (such as cracks). Feature consistency check is a conventional technical means in this field and will not be elaborated here.

[0126] Specifically, combining the information of the image and depth parts, the comprehensive confidence is defined as a weighted combination of multiple indexes, for example:

[0127] 1. Edge coincidence degree : The inverse ratio of the Chamfer matching can be taken and denoted as after normalization.

[0128] 2. Depth discontinuity span : For example, calculate the average gradient magnitude of depth discontinuity points within the ROI, and after normalization, denote it as .

[0129] 3. Edge continuity : It can be represented by the detected edge length or continuity index, and after normalization, denote it as .

[0130] The comprehensive confidence formula can be expressed as:

[0131] .

[0132] Among them, , , are weight coefficients, satisfying . When exceeds the preset threshold , it is determined that there is a crack in the ROI area.

[0133] In the prior art, high-precision sensors are often used to improve the accuracy of data acquisition. Although high-precision sensors can theoretically provide more accurate single-modal data, their effects are often limited by various factors in practical applications. High-precision cameras are easily affected by reflections, shadows, and contrast changes under complex lighting conditions, while high-precision laser scanners can provide accurate depth information, but their ability to capture texture details is limited. Especially when the crack width is small, their resolution may be insufficient. The method of this embodiment compensates for the limitations of a single sensor under specific conditions through multi-modal fusion, making full use of the texture features of the camera and the geometric information of the laser scanner. By fusing the information of the two, this embodiment uses depth information to provide geometric constraints for visual features, eliminates errors caused by lighting changes, and at the same time complements the sparsity of depth data in the laser scanner through visual information, forming complementary advantages, thereby significantly improving the robustness and accuracy of detection. On this basis, the uniqueness of the method of this embodiment lies in fully integrating the advantages of the two sensors through a spatial enhancement strategy based on depth gradients and cross-modal discontinuity matching, and significantly enhancing the significance of crack detection through multi-stage screening and refinement, generating a confidence evaluation model, and improving the adaptability and detection effect of the system in real scenarios from multiple levels. This fusion can not only provide higher robustness in crack detection, but also further improve the overall accuracy through cross-modal matching and confidence evaluation, thus being superior to a single high-precision sensor in real scenarios.

[0134] Embodiment 2:

[0135] Embodiment 2 of the present invention provides a tunnel lining micro-crack detection system based on data fusion, including:

[0136] A data acquisition module, configured to acquire three-dimensional point cloud data and images of a tunnel lining and perform preprocessing;

[0137] A preliminary screening module, configured to use a deep learning model to perform preliminary feature screening on the images to obtain suspected crack area images;

[0138] A depth information extraction module, configured to perform densification processing on the three-dimensional point cloud data to obtain high-density depth information and obtain a depth map;

[0139] A data fusion module, configured to analyze the spatial structure perception provided by the depth map and use a data matching algorithm to fuse the suspected crack area images with the depth map to obtain fused data;

[0140] A crack detection module, configured to calculate the fusion confidence of the fused edge data and determine the crack detection result according to the fusion confidence.

[0141] Embodiment III:

[0142] Embodiment III of the present invention provides a medium on which a program is stored, and when the program is executed by a processor, it implements the steps in the method for detecting fine cracks in a tunnel lining based on data fusion as described in Embodiment I of the present invention.

[0143] Embodiment IV:

[0144] Embodiment IV of the present invention provides a device, including a memory, a processor, and a program stored on the memory and executable on the processor. When the processor executes the program, it implements the steps in the method for detecting fine cracks in a tunnel lining based on data fusion as described in Embodiment I of the present invention.

[0145] The steps involved in the above Embodiments II, III, and IV correspond to those in Method Embodiment I. For specific implementation manners, reference may be made to the relevant description part of Embodiment I.

[0146] Those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computer device. Optionally, they can be implemented by program codes executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0147] Although the specific implementation manners of the present invention have been described above in conjunction with the accompanying drawings, it is not a limitation to the protection scope of the present invention. Those skilled in the art should understand that various modifications or deformations that can be made without creative efforts on the basis of the technical solution of the present invention are still within the protection scope of the present invention.

Claims

1. A method for detecting fine cracks in tunnel lining based on data fusion, characterized in that: The following steps are involved: Obtain 3D point cloud data and images of tunnel lining and perform preprocessing; Use the deep learning model to perform preliminary feature screening on the image to obtain images of suspected crack areas; Densify the 3D point cloud data to obtain high-density depth information and obtain a depth map; Analyze the spatial structure perception provided by the depth map, and use the data matching algorithm to fuse the image of the suspected crack area with the depth map to obtain fused data; The edge detection algorithm is used to detect the image of the suspected crack area and obtain the first detection result: Calculate the local contrast of the image of the suspected crack area; Through the spatial structure perception provided by the depth map, combined with the detailed features of the image grayscale distribution, the depth gradient is calculated using the depth map, and the depth gradient is normalized to obtain the enhancement factor; Calculate depth gradient using depth map : ； Normalize the depth gradient to get the enhancement factor : ； The local contrast of the image is weighted by the enhancement factor to obtain the enhanced image: ； in, is a parameter that adjusts the strength of depth enhancement; Performing edge detection on the enhanced image using an edge detection algorithm to obtain a first detection result; By calculating the depth gradient between the pixels of the depth map, a set of points with significant depth discontinuities is obtained as the second detection result: Densified depth map Calculate the gradient between pixels : ； when When , the pixel point is marked as a depth discontinuity point, where is the threshold value set; Using a data matching algorithm to perform cross-modal data matching and fusion on the first detection result and the second detection result to obtain fused data; The fusion confidence of the fused edge data is calculated, and the crack detection result is determined according to the fusion confidence.

2. The method for detecting fine cracks in tunnel lining based on data fusion according to claim 1, characterized in that: A laser scanner is used to collect 3D point cloud data, and a camera is used to acquire images.

3. The method for detecting fine cracks in tunnel lining based on data fusion according to claim 1, characterized in that: The specific steps of preprocessing include: The intrinsic parameter matrix of the camera and the extrinsic parameter matrix between the laser scanner and the camera are obtained by calibration; The calibrated parameters are used to project the 3D point cloud data collected by the laser scanner onto the camera image plane to achieve spatial alignment of data from different modalities.

4. The method for detecting fine cracks in tunnel lining based on data fusion according to claim 1, characterized in that: The specific steps for using the deep learning model to perform preliminary feature screening on images are: The confidence threshold reduction method is adopted to extract the suspicious areas, establish the confidence distribution based on the detection, analyze the suspicious areas according to the confidence values, and apply dynamic filtering strategies to different categories of suspicious areas; At the same time, a two-dimensional bounding box of each suspicious area is obtained to obtain an image of the suspected crack area.

5. The method for detecting fine cracks in tunnel lining based on data fusion according to claim 3, characterized in that: The specific steps for densifying 3D point cloud data are as follows: Construct Poisson's equation based on 3D point cloud data; Solve Poisson's equation to reconstruct the surface and obtain a continuous depth map; Based on the spatial alignment of data from different modalities, the areas corresponding to the questionable areas of the image in the generated depth map are extracted to obtain high-density depth information.

6. A tunnel lining fine crack detection system based on data fusion, using the tunnel lining fine crack detection method based on data fusion as claimed in any one of claims 1 to 5, characterized in that: include: A data acquisition module is configured to acquire three-dimensional point cloud data and images of the tunnel lining and perform preprocessing; A preliminary screening module is configured to perform preliminary feature screening on the image using a deep learning model to obtain an image of a suspected crack area; A depth information extraction module is configured to perform densification processing on the three-dimensional point cloud data to obtain high-density depth information and obtain a depth map; A data fusion module is configured to analyze the spatial structure perception provided by the depth map, and fuse the image of the suspected crack area with the depth map using a data matching algorithm to obtain fused data; The crack detection module is configured to calculate the fusion confidence of the fused edge data and determine the crack detection result according to the fusion confidence.

7. A computer-readable storage medium, characterized in that: A plurality of instructions are stored therein, and the instructions are suitable for being loaded by a processor of a terminal device and executing the method for detecting fine cracks in tunnel lining based on data fusion as described in any one of claims 1 to 5.

8. A terminal device, characterized in that: It includes a processor and a computer-readable storage medium, the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded by the processor and executing the method for detecting fine cracks in tunnel lining based on data fusion as described in any one of claims 1-5.

Images