Waterproof detection method and device based on fluorescence tracing and storage medium

Abstract

The invention discloses a waterproof detection method and device based on fluorescence tracing and a storage medium, and relates to the technical field of image processing, and the waterproof detection method based on fluorescence tracing comprises the steps: receiving a field picture frame collected by a detection device, IMU data uploaded by the detection device in real time, and positioning data; according to the IMU data and the positioning data, determining an acquisition coordinate corresponding to the field picture frame; determining a mapping area corresponding to the field picture frame in a building model according to the acquisition coordinate and a detection visual angle parameter of the detection device; determining a crack form and a seepage velocity of the mapping region according to the field picture frame; and rendering the crack in the building model based on the crack form and the seepage velocity. According to the invention, the technical effect of identifying the fine crack by determining the specific position and condition of the crack can be realized.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to a waterproof detection method, device and storage medium based on fluorescence tracing. Background Technology

[0002] In the field of industrial component and building structure inspection, early crack identification is crucial for ensuring structural safety. Currently, most mainstream crack detection solutions rely on the local insertion of electrodes. The technical principle is to determine the presence of cracks by monitoring changes in the conductivity between the electrodes. When the current path is blocked due to the appearance of a crack, the system will send back an abnormal signal.

[0003] However, since electrodes can only acquire conductivity data in local areas, it is difficult to locate the specific direction, length and distribution range of cracks through discrete signal points. Therefore, related technologies cannot determine the specific location and condition of cracks, making it difficult to distinguish minute cracks. Summary of the Invention

[0004] The main objective of this application is to provide a waterproof detection method, device, and storage medium based on fluorescence tracing, which aims to solve the technical problem that the specific location and condition of cracks cannot be determined, making it difficult to distinguish minute cracks.

[0005] To achieve the above objectives, this application provides a waterproof detection method based on fluorescence tracer, the method comprising:

[0006] Receives on-site image frames collected by the detection device, IMU data uploaded by the detection device in real time, and positioning data;

[0007] Based on the IMU data and the positioning data, determine the acquisition coordinates corresponding to the scene frame;

[0008] Based on the acquisition coordinates and the detection perspective parameters of the detection device, the mapping area corresponding to the on-site image frame in the building model is determined;

[0009] The crack morphology and seepage velocity in the mapped area are determined based on the on-site video frames.

[0010] The cracks are rendered in the building model based on the crack morphology and the seepage velocity.

[0011] In one embodiment, the step of determining the crack morphology and seepage velocity of the mapped area based on the on-site image frame includes:

[0012] The on-site image frames are input into a pre-trained image processing model, and the fluorescence morphology recognition results and flow dynamics calculation results are obtained through the image processing model.

[0013] The spatial location corresponding to the crack is determined based on the fluorescence morphology recognition results;

[0014] The seepage velocity at each location corresponding to the crack is determined based on the fluid dynamics calculation results.

[0015] In one embodiment, the step of rendering cracks in the building model based on the crack morphology and the seepage velocity includes:

[0016] Determine the starting point of the crack;

[0017] The second seepage velocity corresponding to each point along the crack propagation direction from the starting point is obtained, and the corresponding seepage velocity range of the starting point is determined, wherein the seepage velocity range is the second seepage velocity corresponding to the starting point ± a set value.

[0018] If the second seepage velocity is detected to exceed the seepage velocity range, the crack color between the starting point and the current point is determined based on the average seepage velocity between the starting point and the current point.

[0019] If the end point of the crack is not reached, the current point is taken as the starting point, and the process jumps to the step of determining the starting point of the crack.

[0020] Otherwise, the crack is rendered in the building model based on the crack color and the crack shape.

[0021] In one embodiment, the step of rendering cracks in the building model based on the crack morphology and the seepage velocity includes:

[0022] Determine the average seepage velocity between the start and end points of the crack;

[0023] The color of the crack between the starting point and the ending point is determined based on the average seepage velocity.

[0024] The crack is rendered in the building model based on the crack color and crack shape.

[0025] In one embodiment, after the step of rendering cracks in the building model based on crack morphology, the waterproofing detection method based on fluorescence tracing further includes:

[0026] Obtain the crack color-alarm information mapping relationship;

[0027] A water leakage alarm is generated based on the crack color and the crack color-alarm information mapping relationship.

[0028] In one embodiment, before the step of receiving the scene frame acquired by the detection device, the IMU data uploaded by the detection device in real time, and the positioning data, the following steps are included:

[0029] Calibrate the spraying device onto the architectural model;

[0030] In response to a jet command, determine the jet parameters and jet trajectory corresponding to the jet command;

[0031] The moving component of the spraying device is controlled to move according to the spraying trajectory, and the spraying component of the spraying device is controlled to spray the mixed material according to the spraying parameters, wherein the mixed material includes a first component and a second component.

[0032] In one embodiment, the step of calibrating the spraying device onto the building model includes:

[0033] Obtain the calibration position of the building model;

[0034] Control the spraying device to align with the calibrated position in the construction area;

[0035] The spray device is calibrated based on the calibration position.

[0036] In one embodiment, the step of determining the mapping area of ​​the on-site image frame in the building model based on the acquisition coordinates and the detection perspective parameters of the detection device includes:

[0037] The collected coordinates are transformed to the global coordinate system of the building model to obtain the spatial position coordinates of the detection device in the building model;

[0038] Based on the spatial position coordinates and the horizontal and vertical field of view parameters in the detection perspective parameters, a three-dimensional frustum model of the detection device is constructed.

[0039] Calculate the spatial intersection between the 3D view frustum model and the building model to obtain the initial mapping region;

[0040] Extract edge feature points from the scene frame and match the edge feature points with the surface feature points of the building model;

[0041] The initial mapping region is corrected based on the matching results to obtain the mapping region corresponding to the scene frame in the building model.

[0042] In addition, to achieve the above objectives, this application also provides a waterproof detection device based on fluorescence tracer, the waterproof detection device based on fluorescence tracer includes: a memory, a processor and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the waterproof detection method based on fluorescence tracer as described above.

[0043] In addition, to achieve the above objectives, this application also provides a storage medium, which is a computer-readable storage medium, on which a program for implementing a fluorescent tracer-based waterproof detection method is stored. The program for implementing the fluorescent tracer-based waterproof detection method is executed by a processor to implement the steps of the fluorescent tracer-based waterproof detection method as described above.

[0044] This application provides a waterproofing detection method based on fluorescence tracer. First, it receives on-site image frames acquired by a detection device, IMU data uploaded in real-time by the detection device, and positioning data. Based on the IMU data and the positioning data, it determines the acquisition coordinates corresponding to the on-site image frame. Based on the acquisition coordinates and the detection viewing angle parameters of the detection device, it determines the mapping area of ​​the on-site image frame in the building model. Based on the on-site image frame, it determines the crack morphology and seepage velocity in the mapping area. Finally, it renders cracks in the building model based on the crack morphology and seepage velocity. In other words, this application determines the acquisition coordinates through IMU and positioning data, combines viewing angle parameters to achieve precise mapping between the on-site image and the building model, and then uses fluorescence tracer characteristics to extract crack morphology and seepage velocity from the image, thereby clarifying the specific location and condition of cracks to identify minute cracks. Attached Figure Description

[0045] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0046] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a schematic flowchart of Embodiment 1 of the waterproof detection method based on fluorescence tracer of this application;

[0048] Figure 2 This is a flowchart illustrating steps S01-S03 in Example 5 of the waterproof detection method based on fluorescence tracing in this application.

[0049] Figure 3 This is a schematic diagram of the two-component injection process in Example 5 of the waterproof detection method based on fluorescence tracing in this application;

[0050] Figure 4 This is a simplified flowchart of the waterproof detection method based on fluorescence tracers proposed in this application;

[0051] Figure 5 This is a schematic diagram of the hardware structure involved in the waterproof detection device based on fluorescence tracing in this application.

[0052] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0053] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0054] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0055] Currently, in the field of industrial component and building structure inspection, early crack identification is crucial for ensuring structural safety. Most mainstream crack detection methods rely on locally inserted electrodes. The technical principle is to determine the presence of cracks by monitoring changes in conductivity between the electrodes. When the current path is blocked due to a crack, the system sends an abnormal signal. However, because electrodes can only acquire conductivity data for a localized area, it is difficult to pinpoint the specific direction, length, and distribution of cracks through discrete signal points. Therefore, these technologies cannot determine the exact location and condition of cracks, making it difficult to distinguish even minute cracks.

[0056] The main solution of this application is as follows: receiving on-site image frames collected by the detection device, IMU data uploaded by the detection device in real time, and positioning data; determining the acquisition coordinates corresponding to the on-site image frames based on the IMU data and the positioning data; determining the mapping area corresponding to the on-site image frames in the building model based on the acquisition coordinates and the detection perspective parameters of the detection device; determining the crack morphology and seepage velocity of the mapping area based on the on-site image frames; and rendering cracks in the building model based on the crack morphology and seepage velocity.

[0057] This application uses IMU and positioning data to determine the acquisition coordinates, combines the viewing parameters to achieve a precise mapping between the on-site image and the building model, and then uses the fluorescence tracer characteristics to extract the crack morphology and seepage velocity from the image, thereby clarifying the specific location and condition of the crack to identify minute cracks.

[0058] It should be noted that the executing entity in this embodiment can be a fluorescent tracer-based waterproof detection system, or a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or a fluorescent tracer-based waterproof detection device capable of performing the above functions. This embodiment does not specifically limit it in this way. The following uses a fluorescent tracer-based waterproof detection device as the executing entity as an example to describe this embodiment and the following embodiments.

[0059] Based on this, Embodiment 1 of this application proposes a waterproof detection method based on fluorescence tracing. Please refer to... Figure 1 The waterproof detection method based on fluorescence tracer includes steps S10 to S50:

[0060] Step S10: Receive the on-site image frames collected by the detection device, the IMU data uploaded by the detection device in real time, and the positioning data.

[0061] In this embodiment, the on-site image frame refers to the image of the construction area containing fluorescent tracer captured by the detection device, the IMU data is the acceleration and angular velocity information output by the inertial measurement unit, and the positioning data refers to the position coordinates obtained through GPS or an indoor positioning system.

[0062] As an alternative implementation, a drone equipped with an RGB camera is used to capture on-site image frames, obtain attitude data through the drone's built-in IMU, and simultaneously receive GPS positioning information.

[0063] As another alternative implementation, the detection device and the spraying device are integrated on a mobile platform, sharing a robotic arm or track system to achieve synchronous movement, and relative position data is obtained through an encoder.

[0064] As another optional implementation, the detection device employs a detection robot equipped with a 12-megapixel RGB-IR dual-camera system. The camera frame rate is set to 30fps, allowing simultaneous acquisition of on-site image frames in both visible and near-infrared bands. The near-infrared channel is used to enhance the signal recognition of the fluorescent tracer. The IMU uses a six-axis (three-axis accelerometer + three-axis gyroscope) MEMS sensor with a sampling rate of 200Hz. The output data includes acceleration (in m / s²) along the x / y / z axes. 2 The system provides angular velocity (in rad / s) and positioning data. Positioning data is provided by a dual-mode positioning module integrating RTK-GPS and UWB. In outdoor scenarios, centimeter-level WGS84 coordinates are obtained via RTK-GPS, while indoor scenarios automatically switch to UWB positioning. The positioning data update frequency is 10Hz. The system receives data via a 5G private network, uses the UDP protocol to ensure real-time performance, and enables a data verification mechanism to mark lost or corrupted frames and trigger retransmission.

[0065] Step S20: Determine the acquisition coordinates corresponding to the scene frame based on IMU data and positioning data.

[0066] In this embodiment, the acquisition coordinates refer to the spatial position of the scene image frame in the global coordinate system of the building model.

[0067] As an alternative implementation, the positioning error can be eliminated by fusing the inertial navigation data of the IMU and the GPS positioning data through the Kalman filter algorithm.

[0068] As another alternative implementation, a visual SLAM algorithm based on feature point matching is used, combined with IMU pre-integration technology, to achieve high-precision estimation of the acquired coordinates.

[0069] As a detailed implementation method, a sensor fusion algorithm based on Error State Kalman Filtering (ESKF) is employed. First, IMU data undergoes zero-bias calibration and temperature compensation. Allan variance analysis is used to determine the noise parameters of the gyroscope and accelerometer. Positioning data is used as the observed values, and IMU data as the predicted values ​​to construct state and observation equations: the state vector includes position (x, y, z), velocity (vx, vy, vz), attitude (quaternions q0-q3), and IMU zero bias; the observation equation is constructed using RTK-GPS / UWB position measurements. During the filtering process, an update is performed for each received frame of positioning data, while IMU data is used for continuous prediction. The final output is the acquired coordinates (3D coordinates in the global coordinate system of the building model, unit m) synchronized with the timestamp of the scene frame, with the positioning error controlled within ±3 cm.

[0070] Step S30: Determine the mapping area of ​​the scene frame in the building model based on the acquisition coordinates and the detection perspective parameters of the detection device.

[0071] In this embodiment, the detection viewpoint parameters include the camera's intrinsic and extrinsic parameter matrices, and the mapping area refers to the projection area of ​​the scene image onto the building model.

[0072] As an alternative implementation, a viewing cone with the optical center of the detection device as its vertex is constructed, and the mapping region is determined by calculating the intersection of the viewing cone and the building model.

[0073] As another alternative implementation, a spherical projection algorithm is used to project the scene onto the three-dimensional surface of the building model, and the corresponding area is determined by texture mapping.

[0074] As a detailed implementation method, the acquired coordinates (local coordinate system) are first transformed to the global coordinate system of the building model (such as the building coordinate system) using a coordinate transformation matrix. The transformation matrix is ​​pre-stored in the system based on the previously calibrated coordinate mapping relationship. In the detection viewpoint parameters, the horizontal field of view is set to 78°, and the vertical field of view is set to 45°. Based on the position of the acquired coordinates (x0, y0, z0) and the direction of the camera optical axis (derived from IMU attitude data), a three-dimensional frustum model is constructed with (x0, y0, z0) as the vertex and the optical axis as the central axis. The near-plane distance of the frustum is set to 0.5m, and the far-plane distance is set to 10m. A spatial geometric intersection algorithm is used to calculate the intersection area between the frustum and the building model (BIM model, containing a set of triangular facets of components such as walls and floors), obtaining an initial mapping area composed of multiple triangular facets. Subsequently, the Canny edge detection algorithm was used to extract edge feature points (such as crack edges and component corners) from the scene frames. Feature vectors were generated using SIFT feature descriptors and then KNN matching was performed with pre-extracted feature points on the surface of the building model (generated using MeshLab software), with a matching threshold set to 0.75. Finally, the RANSAC algorithm was used to remove mismatched points, and the homography matrix was calculated to geometrically correct the initial mapped area. The deviation between the corrected mapped area and the actual scene was less than ±5mm.

[0075] Step S40: Determine the crack morphology and seepage velocity in the mapped area based on the on-site image frame.

[0076] In this embodiment, the crack morphology includes the length, width, and orientation of the crack, and the seepage velocity refers to the flow rate of the fluorescent tracer in the crack.

[0077] As an alternative implementation, a convolutional neural network is used to perform semantic segmentation on the scene, identify crack areas, and extract morphological parameters.

[0078] As another alternative implementation, the optical flow method is used to analyze the movement trajectory of the fluorescent tracer in consecutive frames and calculate the seepage velocity.

[0079] As a detailed implementation method, for each frame of the live footage, preprocessing is first performed: Gaussian filtering (σ = 1.5) is applied to the RGB channels to remove noise, and adaptive thresholding is used to extract the fluorescence region in the IR channels. An improved U-Net++ network is used for crack morphology recognition. The network input is the preprocessed RGB-IR fused image, and the output is a pixel-level binary mask of the crack. The mask boundary is optimized through morphological operations (erosion-dilation), and the crack centerline is obtained using a skeleton extraction algorithm. The crack length (in mm) and orientation (angle with the horizontal direction, in °) are calculated based on the minimum bounding rectangle. The crack width (maximum distance perpendicular to the centerline) is calculated using a distance transformation algorithm. For seepage velocity calculation, a deep learning-based optical flow estimation model (such as RAFT) is used to process 10 consecutive frames of the live footage, outputting a two-dimensional velocity vector per pixel (in pixels / s). By combining the pixel-to-physical size conversion relationship calibrated by the camera (e.g., 1 pixel = 0.1 mm), the pixel velocity is converted into the actual seepage velocity (unit: mm / s), and outliers are removed by median filtering. Finally, the velocity distribution curve along the center line of the crack is output.

[0080] Step S50: Render the cracks in the building model based on the crack morphology and seepage velocity.

[0081] In this embodiment, rendering cracks refers to visually representing the location, shape, and seepage status of cracks on the building model.

[0082] As an alternative implementation, lines of different thicknesses are mapped according to the crack width, and gradient effects of different colors are mapped according to the seepage velocity.

[0083] As an alternative implementation, augmented reality technology is used to overlay crack information onto a real-world image of the building model.

[0084] As a detailed implementation, a WebGL-based 3D rendering engine (such as Three.js) is used to visualize the cracks. First, a 3D geometric model is generated in the mapped area of ​​the building model based on the crack morphology data (centerline, length, width): a cylindrical mesh is created with the centerline as the path and the width as the cross-sectional diameter. The number of segments in the cylinder is automatically adjusted according to the length (one segment every 10mm). Seepage velocity rendering uses heatmap mapping: the velocity values ​​(0-5mm / s) are mapped to the HSV color space (blue-green-red gradient). A vertex shader assigns a color value to each vertex of the cylindrical mesh, and a fragment shader achieves smooth color transitions. Simultaneously, dynamic effects are added: a blinking animation (frequency 2Hz) is enabled for areas with velocities >3mm / s, displaying the specific velocity value and coordinate information of that point when the mouse hovers over it. The rendered results are then fused with the building model through depth testing to ensure the crack model is correctly displayed on the building surface, supporting interactive operations such as scaling and rotation, and maintaining a frame rate above 30fps.

[0085] For example, in a waterproofing inspection scenario, a track-based inspection system integrated with a spraying device is employed. The system synchronously moves the spraying and inspection devices along a track on the building facade. First, a two-component mixture containing a fluorescent tracer is sprayed by the spraying device, followed immediately by the inspection device scanning the construction area. The inspection device uploads RGB images, IMU attitude data, and track encoder position data in real time, calculating the acquired coordinates using a visual SLAM algorithm. A view frustum is constructed based on the camera's intrinsic and extrinsic parameters to determine the mapped area of ​​the image in the building's BIM model. Fluorescent cracks are identified using a DeepLabv3+ network, crack morphology parameters are extracted, and seepage velocity is calculated using particle image velocimetry (PIV) technology. Finally, the cracks are rendered on the BIM model as a dynamic heatmap, with red areas representing high-speed seepage and yellow areas representing low-speed seepage.

[0086] For example, the above method is used in waterproofing inspection projects. The inspection robot cruises along a preset path, simultaneously acquiring on-site image frames (RGB-IR dual-band), IMU data (200Hz), and dual-mode positioning data (10Hz), which are then transmitted to the backend server via 5G. The server first fuses the IMU and positioning data using the ESKF algorithm to obtain the acquisition coordinates of each frame (error ±2cm). Subsequently, the coordinates are transformed to the global coordinate system of the building BIM model, and a view frustum is constructed by combining the camera's field of view (horizontal 78° / vertical 45°). The intersection of this frustum with the BIM model yields the initial mapping area. After SIFT feature matching correction, the mapping deviation is controlled within ±4mm. After preprocessing the image frames, the U-Net++ network accurately identifies minute cracks with a width of 0.2mm, extracting their length (5.2m), direction (30°), and other morphological parameters. The RAFT optical flow model calculates the seepage velocity distribution along the crack (0.5-4.2mm / s). Finally, the cracks are rendered as colored cylinders in the BIM model. Blue segments (0.5-1.5 mm / s) represent low-speed seepage, green segments (1.5-3 mm / s) represent medium-speed seepage, and red flashing segments (>3 mm / s) represent high-speed seepage. Inspectors can intuitively view the location, shape, and seepage status of the cracks through an interactive interface, providing accurate data support for the formulation of waterproofing and repair plans.

[0087] In this embodiment, during the data acquisition and coordinate determination stage, the system receives on-site image frames, IMU data, and positioning data collected by the detection device, and determines the acquisition coordinates accordingly. This links the on-site images with precise spatial locations, providing an accurate spatial reference for subsequent mapping of the images onto the building model. This solves the problem of ambiguous positional information in traditional local electrode detection, giving each on-site image frame a clear "spatial label." The mapping area is determined based on the acquisition coordinates and the detection perspective parameters of the detection device. By establishing a correspondence between the on-site images and the building model, a precise mapping from the real-world scene to the digital model is achieved. This mapping overcomes the limitations of local detection, allowing the detection range to be no longer restricted by the electrode insertion position, covering a larger construction area. Furthermore, the accuracy of the mapped area lays the foundation for subsequent crack information analysis and presentation, avoiding deviations in crack location judgment. Based on the on-site image frames, the crack morphology and seepage velocity in the mapped area are determined. Combined with the characteristics of fluorescence tracing technology, even minute cracks can be captured more clearly. Fluorescent tracers can develop within cracks, making them easier to identify in images. This allows for accurate analysis of crack morphology, such as length, width, and direction. Furthermore, the flow state of the fluorescent tracer can precisely determine the seepage rate. Compared to traditional methods, this significantly improves the sensitivity and accuracy of detecting minute cracks and seepage. By rendering cracks in a building model based on their morphology and seepage rate, abstract detection data is transformed into intuitive visualizations. This visualization allows staff to clearly see the specific location, shape, and seepage conditions of cracks within the building model, facilitating a comprehensive understanding of crack distribution and severity. This provides an intuitive and reliable basis for subsequent waterproofing treatment and construction decisions, improving work efficiency and decision-making accuracy.

[0088] Based on any of the above embodiments, in Embodiment 2 of this application, step S40 includes:

[0089] Step S41: Input the on-site image frame into the pre-trained image processing model, and obtain the fluorescence morphology recognition result and the fluid dynamics calculation result through the image processing model.

[0090] In this embodiment, the image processing model refers to a multi-task deep neural network that integrates semantic segmentation and optical flow estimation. The fluorescence morphology recognition result includes a binary mask image of the crack and a skeletonized centerline. The fluid dynamics calculation result is a velocity field image represented in the HSO color space.

[0091] As an alternative implementation, a multi-branch network based on DeepLabv3+ is constructed. The main branch uses the Xception backbone network for crack semantic segmentation, while the auxiliary branch calculates the optical flow field through the PWC-Net architecture. The two branches share the feature extraction layer.

[0092] As an alternative implementation, a spatiotemporal joint attention network is developed, which captures the temporal features between consecutive frames through 3D convolution, enhances the perception of fluorescence signals by utilizing channel attention mechanism, and outputs crack evolution sequence and velocity field tensor with time dimension.

[0093] Step S42: Determine the spatial location of the crack based on the fluorescence morphology recognition results.

[0094] In this embodiment, the spatial location corresponding to the crack refers to the set of homogeneous coordinate points that map the image pixel coordinates to the three-dimensional space of the building model.

[0095] As an optional implementation, the projection relationship between pixel coordinates and world coordinates is established through camera calibration parameters, the contour of the segmented crack mask is extracted, and the three-dimensional spatial curve of the crack centerline is fitted using the Random Sample Consensus (RANSAC) algorithm.

[0096] As another alternative implementation, a geometric reasoning model based on graph neural networks is constructed to transform the crack segmentation results into a node-edge graph structure. The three-dimensional coordinates of each node are calculated through a message passing mechanism, and finally, a Bézier curve is fitted to represent the crack trajectory.

[0097] Step S43: Determine the seepage velocity at each location corresponding to the crack based on the fluid dynamics calculation results.

[0098] In this embodiment, the seepage velocity refers to the apparent flow velocity of the fluid in the crack, which is obtained by calculating the displacement change rate of the fluorescent tracer.

[0099] As an optional implementation, the Lucas-Kanade optical flow algorithm based on feature point tracking is adopted to match fluorescent particles between consecutive frames, calculate the velocity vector through particle displacement and time interval, and convert it into the true seepage velocity by combining crack width information.

[0100] As another alternative implementation, a physical constraint-based optical flow field optimization model is developed, incorporating the Navier-Stokes equations as regularization terms into the loss function, and solving the velocity field that satisfies the laws of fluid mechanics through variational methods to improve the measurement accuracy in low-speed seepage scenarios.

[0101] For example, in a waterproofing inspection scenario, the captured on-site video frames are input into a pre-trained multi-task network. The model first outputs pixel-level segmentation results of cracks through the DeepLabv3+ branch, identifying minute cracks with widths as low as 0.1 mm; simultaneously, it calculates the photoflow field of the fluorescent tracer through the PWC-Net branch. Using the camera's intrinsic and extrinsic parameter matrices, the crack pixel coordinates are projected onto the tunnel BIM model to obtain their three-dimensional spatial location. Spatiotemporal integration of the photoflow field is performed to calculate the seepage velocity at each point, and the velocity magnitude is represented by the arrow length and the flow direction by the arrow color on the BIM model.

[0102] This embodiment combines deep learning and computer vision technologies to achieve high-precision identification of crack morphology and quantitative analysis of seepage velocity, solving the technical problem that traditional methods cannot quantify seepage state and detect minute cracks.

[0103] Based on any of the above embodiments, in Embodiment 3 of this application, step S50 includes:

[0104] Step S51: Determine the starting point of the crack.

[0105] In this embodiment, the starting point refers to the crack feature point determined by curvature analysis and velocity gradient calculation.

[0106] As an optional implementation, the curvature of the crack skeleton line is calculated, and the local maxima of curvature are used as candidate starting points. The true starting point is then selected by detecting abrupt changes in seepage velocity. Specifically, the second derivative of each point on the skeleton line is calculated. When the absolute value of the derivative exceeds a set threshold and the rate of change of seepage velocity at the corresponding point is greater than 15%, it is determined to be a starting point.

[0107] As an alternative implementation, a spatiotemporal attention model is constructed. By analyzing the crack propagation process in consecutive frames, the source point of the seepage velocity field is identified as the starting point. A 3D convolutional neural network is used to extract spatiotemporal features, and the attention mechanism is used to enhance the perception of regions with abrupt velocity changes.

[0108] Step S52: Obtain the second seepage velocity corresponding to each point along the crack propagation direction from the starting point, and determine the corresponding seepage velocity range of the starting point.

[0109] In this embodiment, the seepage velocity range is a dynamic threshold range determined based on the statistical characteristics of the seepage velocity at the starting point.

[0110] As an optional implementation, a sliding window algorithm is used to sample along the crack centerline, and the seepage velocity at each point is predicted by Kalman filtering. Simultaneously, the standard deviation of historical velocities is calculated as a setpoint. The sliding window size is set to three times the average crack width, and the state transition matrix of the Kalman filter is constructed based on the fluid dynamics equations.

[0111] As an alternative implementation, a Gaussian mixture model is used to model the velocity distribution near the starting point, with the 95% confidence interval used as the seepage velocity interval. The model parameters are estimated using the Expectation-Maximization (EM) algorithm, and the interval width is adaptively adjusted.

[0112] Step S53: When the second seepage velocity is detected to exceed the seepage velocity range, the crack color between the starting point and the current point is determined based on the average seepage velocity between the starting point and the current point.

[0113] In this embodiment, the crack color refers to the HSV color space value mapped by the seepage velocity.

[0114] As an optional implementation, a piecewise linear mapping function is designed to map the average seepage velocity to a gradient color band from blue (low velocity) to red (high velocity), with uncertainty represented by a transparency channel. Specifically, the mapping relationships are: 0-0.2 m / s corresponds to blue (240°, 100%, 100%), 0.2-0.5 m / s corresponds to green (120°, 100%, 100%), 0.5-1.0 m / s corresponds to yellow (60°, 100%, 100%), and greater than 1.0 m / s corresponds to red (0°, 100%, 100%).

[0115] As an alternative implementation, a color mapping model based on a generative adversarial network is constructed. This model takes the seepage velocity sequence as input and outputs a heatmap color code that aligns with human intuition. Adversarial training further refines the generated color distribution to better match the visual perception of domain experts.

[0116] Step S54: If the end point of the crack has not been reached, take the current point as the starting point and jump to the step of determining the starting point of the crack.

[0117] In this embodiment, the termination point refers to the point where the seepage velocity approaches zero or the crack width is less than the detection threshold.

[0118] As an optional implementation, the fluid convergence point is determined by optical flow field divergence analysis, and the termination point is determined by combining the crack width change rate. When the divergence value is less than -0.1 and the width change rate exceeds 80%, it is determined to be the termination point.

[0119] As an alternative implementation, a reinforcement learning agent is used to roam within the crack network, and termination conditions are determined based on velocity decay and geometric features. When designing the reward function, a velocity approaching zero and a width less than 0.1 mm are used as positive reward signals.

[0120] Step S55: Render the cracks in the building model based on the crack color and crack shape.

[0121] In this embodiment, rendering refers to the process of mapping two-dimensional crack information onto the surface of a three-dimensional building model.

[0122] As an optional implementation, a vertex shader is used to implement the geometric deformation of the cracks, a fragment shader is used to render the color according to the seepage velocity, and a stencil buffer technique is used to achieve a three-dimensional protrusion effect of the cracks. The crack width is achieved by vertex offset, the protrusion height is proportional to the seepage velocity, and the maximum height is set to 5mm.

[0123] As an alternative implementation, a WebGL-based real-time rendering engine was developed to convert crack data into GLTF format, supporting interactive viewing and analysis within the BIM 3D scene. Multi-level detail (LOD) rendering was implemented, dynamically adjusting the crack rendering accuracy based on camera distance.

[0124] For example, in a waterproofing inspection scenario, the system first determines the crack initiation point through curvature analysis, calculates the seepage velocity at that point as 0.35 m / s, and sets a velocity range of ±0.1 m / s. Samples are taken every 5 mm along the crack propagation direction. When the seepage velocity at the 15th sampling point reaches 0.52 m / s, the average velocity from the initiation point to that point is calculated to be 0.43 m / s, mapped to yellow-green. This process continues until the crack termination point, segmenting the entire crack trajectory into blue-cyan-green-yellow gradient bands. The width is dynamically adjusted between 1-3 mm based on the crack morphology, creating an intuitive visualization of the seepage status in the BIM model. Users can observe the crack situation from multiple angles by rotating and zooming the model, and click on specific locations to view detailed seepage velocity data.

[0125] This embodiment achieves refined visualization of crack seepage status through dynamic threshold detection and segmented rendering technology, solving the technical problem that traditional static rendering methods cannot reflect seepage change trends, and providing a more intuitive and accurate basis for waterproofing repair decisions.

[0126] Based on any of the above embodiments, in Embodiment 4 of this application, step S50 includes:

[0127] Step S51: Determine the average seepage velocity between the start and end points of the crack.

[0128] In this embodiment, the mean seepage velocity is a scalar value obtained by integrating and averaging the seepage velocities at sampling points along the crack trajectory.

[0129] As an alternative implementation, a uniform sampling strategy is adopted, and more than 100 velocity sample points are collected at fixed intervals (e.g., 5 mm) along the center line of the crack, and the mean velocity is calculated by arithmetic mean.

[0130] As an alternative implementation, a density-based adaptive sampling algorithm is designed. The sampling point density is increased in areas with high crack curvature, and a weighted average method is used to calculate the average velocity, with the weight proportional to the crack area surrounding the sampling point. Step S52: The crack color between the starting and ending points is determined based on the average seepage velocity.

[0131] In this embodiment, the crack color is an RGB value determined by a predefined velocity-color mapping function.

[0132] As an alternative implementation, a linear mapping function is constructed to map the mean percolation velocity v to the hue value H in the HSV color space:

[0133]

[0134] Where S = 100% and V = 100%, HSV is then converted to RGB output.

[0135] As an alternative implementation, a color mapping model based on a conditional generative adversarial network (cGAN) is trained, using the average seepage velocity as a conditional input, to generate crack colors that conform to human intuition, while outputting color confidence scores. Step S53: Based on the crack color and crack morphology, the cracks are rendered in the building model.

[0136] In this embodiment, rendering refers to the process of projecting two-dimensional crack information onto the surface of the building model in a three-dimensional form.

[0137] As an optional implementation, a template buffering technique is used, in which a binary mask of the crack is used as a template to draw a three-dimensional crack with thickness and shadow on the surface of the building model. The crack color is determined by step S52.

[0138] As an alternative implementation, a physically based rendering (PBR) material system was developed. This system adjusts the surface roughness and metallicity of the crack based on the average seepage velocity to simulate the reflective properties of liquid under different seepage conditions. For example, in a bridge box girder waterproofing inspection scenario, the system detected a 3.2-meter-long crack. 120 velocity sample points were uniformly collected along the crack's centerline, and the average seepage velocity was calculated to be 0.45 m / s. The corresponding HSV value was calculated using a linear mapping function as (120°, 100%, 100%), which was converted to an RGB value of (0, 255, 0) (green). Using template buffering technology, a 2mm wide and 1mm thick green 3D crack was rendered on the surface of the box girder's BIM model. A 0.5mm shadow effect was added to the crack edges to enhance the 3D effect. Users can adjust the viewing angle through an interactive interface to clearly see the extension of the crack inside the box girder.

[0139] This embodiment significantly reduces computational complexity while ensuring visualization effects by simplifying the calculation of the average velocity and using a single-color rendering strategy. It enables a rapid and intuitive presentation of the crack seepage status, providing waterproofing inspectors with an efficient decision support tool.

[0140] Furthermore, after rendering the cracks based on the crack shape in the building model, the crack color-alarm information mapping relationship is obtained; based on the crack color and the crack color-alarm information mapping relationship, a water leakage alarm is generated.

[0141] In this embodiment, the crack color-alarm information mapping relationship refers to the correspondence rules between different crack colors and the corresponding leakage severity and handling suggestions.

[0142] As an optional implementation, an expert system can be used to pre-define mapping relationships, with blue corresponding to minor leaks (no emergency treatment required), green to moderate leaks (requiring regular observation), and red to severe leaks (requiring immediate repair).

[0143] As another optional implementation, a decision tree algorithm is used to train a mapping model based on historical detection data and maintenance records, automatically generating the correspondence between colors and alarm levels, and dynamically updating it as data accumulates.

[0144] Step S70: Generate a water leakage alarm based on the crack color and the crack color-alarm information mapping relationship.

[0145] In this embodiment, a water leakage alarm refers to a prompt message that includes the location of the leak, its severity, and handling suggestions.

[0146] As an optional implementation, when a crack is detected to be red, the system automatically triggers an audible and visual alarm and pushes an emergency repair notification containing a BIM model location link to the manager's mobile APP.

[0147] As another alternative implementation, the alarm level corresponding to a single color is weighted and adjusted by combining the average crack length and seepage velocity. For example, a green crack longer than 5 meters is automatically upgraded to a yellow alarm and simultaneously flashed on the building model.

[0148] For example, during waterproofing testing, the system obtains a preset mapping relationship: blue (seepage velocity ≤ 0.2 m / s) corresponds to a Level 1 alarm (routine inspection), yellow (0.2 m / s < seepage velocity ≤ 0.5 m / s) corresponds to a Level 2 alarm (arrange maintenance), and red (seepage velocity > 0.5 m / s) corresponds to a Level 3 alarm (emergency repair). Upon detecting a red crack, the system immediately generates a Level 3 alarm message, including the crack's precise coordinates in the garage BIM model, a seepage velocity of 0.6 m / s, and a suggestion to "immediately close the nearby water supply and drainage valves and organize grouting repair," while simultaneously triggering the audible and visual alarm device in the monitoring center.

[0149] This embodiment establishes a direct link between color and alarm, enabling rapid early warning of water leakage risks. This allows managers to prioritize handling emergencies based on alarm levels, improving the response efficiency of waterproofing repairs.

[0150] Based on any of the above embodiments, in Embodiment 5 of this application, referring to Figure 2 Before step S10, the following are included:

[0151] Step S01: Calibrate the spraying device onto the building model.

[0152] In this embodiment, calibration refers to establishing the transformation relationship between the physical coordinates of the spraying device and the global coordinate system of the building model to ensure accurate mapping of the spraying position.

[0153] As an optional implementation, a hand-eye calibration method based on a checkerboard calibration board is adopted. A high-precision checkerboard is placed at a preset position on the building model, and the end of the robotic arm of the spraying device is controlled to carry a camera to capture images of the calibration board. The camera intrinsic parameters are calculated using the Zhang Zhengyou calibration algorithm, and the transformation matrix between the robotic arm base coordinate system and the building model coordinate system is solved using the PnP algorithm.

[0154] As another optional implementation, a laser tracker is used to collect the coordinates of multiple feature points of the end effector of the spraying device in the coordinate system of the building model. Combined with the kinematic model of the spraying device itself, the coordinate transformation relationship is obtained by least squares fitting, and the calibration error is controlled within ±0.5mm.

[0155] Step S02: In response to the injection command, determine the injection parameters and injection trajectory corresponding to the injection command.

[0156] In this embodiment, the spraying parameters include the spraying pressure, flow rate, mixing ratio, and spraying distance of the mixed material, and the spraying trajectory refers to the movement path of the spraying device on the building model.

[0157] As an optional implementation method, a smooth spraying trajectory is generated by B-spline curve interpolation based on the three-dimensional contour of the area to be waterproofed in the building model. At the same time, the spraying parameters are preset according to the material of the area: the spraying pressure is set to 0.8MPa, the flow rate is 1.2L / min, the mixing ratio of component A to component B is 1:1.2, and the spraying distance is 300mm for concrete surfaces.

[0158] As another optional implementation, the system receives the spray path drawn interactively by the user through the BIM model, and automatically adjusts the spray angle based on the normal vectors of each point on the path. It also uses a fuzzy control algorithm to dynamically correct the spray parameters according to the ambient temperature and humidity. For example, the spray pressure is reduced by 0.05 MPa for every 5°C increase in ambient temperature.

[0159] Step S03: Control the moving component of the spraying device to move according to the spraying trajectory, and control the spraying component of the spraying device to spray the mixed material according to the spraying parameters.

[0160] In this embodiment, the moving component includes a track slider, a robotic arm, or an AGV moving platform, and the spraying component includes a two-component metering pump, a static mixer, and a nozzle. The first component may be epoxy resin, and the second component may be a curing agent. The two components are mixed to form a polymer material with waterproof properties.

[0161] As an optional implementation, the servo motor of the moving component is driven by a PID control algorithm. The position information fed back by the encoder is collected in real time and compared with the theoretical position of the injection trajectory. The position deviation is eliminated by proportional-integral-derivative adjustment, and the control accuracy reaches ±1mm. At the same time, the speed ratio of the two-component metering pump is controlled by PLC to ensure that the mixing ratio error does not exceed ±2%. The static mixer adopts a 16-section spiral structure to ensure the uniformity of mixing.

[0162] As another optional implementation, the model predictive control (MPC) algorithm is used to plan the acceleration and velocity curves of the moving component to avoid impact and vibration during the movement; the injection component is equipped with a mass flow meter to monitor the instantaneous flow of the two components in real time, and the frequency of the metering pump is adjusted through closed-loop control. When the mixing ratio deviation exceeds the threshold, an audible and visual alarm is triggered and the machine is automatically shut down.

[0163] Furthermore, in this embodiment, the moving component includes a track slider, a robotic arm, or an AGV moving platform, and the spraying component includes a two-component metering pump, a static mixer, and a nozzle. The first component is a quantum dot hydrogel (an aqueous gel system containing CdSe / ZnS quantum dots, exhibiting fluorescent tracer properties), and the second component is an aerosol propellant (compressed gas containing HFC-134a, boiling point -26.1°C). The two components are mixed and then atomized under high pressure to form a spray coating that combines waterproof sealing and fluorescent marking functions. (Refer to...) Figure 3The first component is quantum dot hydrogel, and the second component is aerosol propellant. The injection device detects the injection of the two components and injects them into the construction area through high-pressure mixing. The mixed material penetrates through the capillary pores of the concrete. After crack detection is completed, the mixed material can be naturally degraded in 28 days.

[0164] As an optional implementation, an ultra-high pressure plunger pump is used as the metering device. The delivery pressure of the first component is set to 15-20 MPa, and the pressure of the second component propellant is controlled to be stable at 8-10 MPa by a pressure reducing valve. The two components are mixed in a specially designed impact-type mixing chamber to form a supersonic jet. The mixing chamber is made of tungsten carbide material to withstand high pressure impact. The moving component adopts a multi-axis linkage robotic arm, which transmits trajectory commands in real time through EtherCAT bus. The motion accuracy of the end nozzle is controlled within ±0.3 mm to ensure that the alignment deviation between the spray area and the crack does not exceed 2 mm.

[0165] As an alternative implementation, the injection assembly is equipped with a twin-screw metering pump, which controls the delivery volume of quantum dot hydrogel (accuracy ±0.5%) through a closed-loop control of the servo motor speed. The propellant storage cylinder is equipped with a pressure sensor for real-time monitoring, and automatically switches to a backup cylinder when the pressure is below 6MPa. The mixed material is accelerated to Mach 1.2 through a Laval nozzle, forming an atomized cone with a diameter of 5-10mm on the construction surface. The moving assembly uses a magnetic navigation AGV in conjunction with a lifting column, and uses laser radar to avoid obstacles while moving along a preset trajectory, ensuring that the distance to the construction surface is kept stable at 300±5mm.

[0166] For example, in a waterproofing construction scenario, the track-mounted spraying device is first calibrated onto the tunnel BIM model using a laser tracker. After calibration, the system receives spraying instructions for the cracked area in the tunnel arch. Based on the three-dimensional coordinates of the crack in the BIM model, a rectangular spraying area with a width of 500mm centered on the crack is generated. A spiral spraying trajectory along the crack is generated using a B-spline curve. The spraying pressure is set to 1.0MPa, the flow rate to 1.5L / min, the mixing ratio of component A (polyurethane prepolymer) to component B (catalyst) to be 1:1, and the spraying distance to be 250mm. The spraying device is controlled to move along the track component at a speed of 0.5m / s. Simultaneously, a two-component metering pump delivers materials according to the set ratio. After mixing by a static mixer, the materials are evenly sprayed out through a fan-shaped nozzle, covering the crack and surrounding area to form a continuous waterproof coating.

[0167] For example, in a crack waterproofing scenario, a magnetically guided AGV is used in conjunction with a lifting column as the moving component, equipped with a two-component high-pressure injection system. The first component is an aqueous hydrogel containing CdSe / ZnS quantum dots (quantum dot concentration 0.5 mmol / L), and the second component is HFC-134a aerosol propellant (purity 99.9%). The system first scans the surface of the containment vessel using lidar to generate a 3D point cloud model and aligns it with a preset BIM model, planning a spiral injection trajectory (pitch 5 mm) along the crack direction. The injection parameters are set as follows: quantum dot hydrogel delivery pressure 18 MPa, propellant pressure 9 MPa, two-component flow ratio 1:0.8, and the speed is controlled by a closed-loop servo motor of a twin-screw metering pump (speed fluctuation ≤ ±1 rpm). After injection is started, the AGV moves along the magnetically guided track, and the lifting column adjusts the nozzle height in real time to ensure a stable distance of 300 mm from the concrete surface. The mixture is accelerated to Mach 1.2 by the Laval nozzle, forming an 8 mm diameter atomized cone, creating a continuous fluorescent waterproof coating in the crack area. When the propellant tank pressure drops to 6 MPa, the system automatically switches to the backup tank (switching time < 0.5 s) to avoid interruption of injection. After construction, the coating thickness uniformity error is < ±0.3 mm, and the detection sensitivity of quantum dot fluorescence signals under 365 nm ultraviolet light reaches 0.1 mm crack resolution, providing a clear marking basis for subsequent fluorescence tracer detection.

[0168] This embodiment ensures that the mixed material can cover the area to be waterproofed according to the design requirements through precise calibration, reasonable parameter and trajectory planning, and accurate control. This provides a reliable material basis for subsequent crack detection and improves the quality and efficiency of waterproofing construction.

[0169] Further, step S01 includes:

[0170] Step S011: Obtain the calibration position of the building model.

[0171] In this embodiment, the calibration location refers to the feature point or feature area preset in the building model for calibrating the spraying device. Typically, a location with obvious geometric features and not easily deformed is selected within the construction area, such as a corner of a wall or the center point of an embedded part.

[0172] As an optional implementation, preset three-dimensional coordinate points are extracted from the Building Information Model (BIM) as calibration locations. These coordinate points have been marked as "calibration references" in the model, and their global coordinate system adopts the National Geodetic Coordinate System 2000.

[0173] As another alternative implementation method, a point cloud model is generated by scanning the construction area with laser, and a point cloud segmentation algorithm is used to identify highly recognizable feature structures (such as circular embedded parts with a diameter of 50mm), and the coordinates of their center points are used as the calibration positions.

[0174] Step S012: Control the spraying device to align with the calibrated position in the construction area.

[0175] In this embodiment, alignment refers to adjusting the position and attitude of the spraying device so that the positioning reference of the spraying device coincides with the calibrated position of the construction area in space.

[0176] As an optional implementation, the construction area image is captured by the vision sensor mounted on the spraying device. The visual features corresponding to the calibration position are identified by the template matching algorithm. The moving components of the spraying device (such as AGV or robotic arm) are driven by the PID control algorithm to perform translation and rotation adjustment until the visual features are located in the center of the image and are consistent with the preset size ratio.

[0177] As another optional implementation, the distance and angle deviation from the calibration position is measured by a laser rangefinder on the spraying device, and the deviation data is transmitted to the motion controller to control the servo motor of the moving component to perform closed-loop motion, so that the deviation value is controlled within ±0.5mm.

[0178] Step S013: Calibrate the spray device based on the calibration position.

[0179] In this embodiment, calibrating the spraying device based on the calibration position refers to establishing the transformation relationship between the coordinate system of the spraying device itself and the global coordinate system of the building model, so as to achieve accurate mapping of the spraying trajectory.

[0180] As an optional implementation, after the spraying device is aligned to the calibration position, the coordinates of the end effector of the spraying device in its own coordinate system are recorded. Combined with the coordinates of the calibration position in the global coordinate system of the building model, the transformation matrix is ​​calculated using the coordinate transformation formula to complete the calibration.

[0181] As another optional implementation method, a multi-point calibration method is adopted to obtain the coordinates of multiple calibration positions in the coordinate system of the spraying device and the global coordinate system of the building model. The optimal transformation matrix is ​​fitted by the least squares method to reduce the error that may be generated by single-position calibration and make the calibration accuracy reach ±0.3mm.

[0182] This embodiment ensures a high-precision match between the spraying device and the building model through clear calibration location acquisition, precise alignment control, and scientific calibration calculation, providing a strong guarantee for the accuracy of subsequent spraying construction.

[0183] Based on any of the above embodiments, in Embodiment Six of this application, step S30 includes:

[0184] Step S31: Transform the collected coordinates to the global coordinate system of the building model to obtain the spatial position coordinates of the detection device in the building model.

[0185] In this embodiment, the acquired coordinates are the coordinates of the detection device under its own positioning system, the global coordinate system of the building model is a pre-set coordinate system used to uniformly describe the spatial position of the building, and the spatial position coordinates are the three-dimensional coordinate representation of the detection device in the global coordinate system.

[0186] As an optional implementation, the acquired coordinates are transformed to the global coordinate system of the building model through a coordinate transformation matrix. This transformation matrix is ​​pre-calculated based on the calibration results of the detection device and the building model. During the transformation process, the rotation parameters are processed using the quaternion method to reduce the cumulative error of the coordinate transformation.

[0187] As another alternative implementation, the Kalman filter algorithm is used to fuse the positioning data of the detection device and the coordinates of known feature points of the building model, and the coordinate transformation parameters are optimized in real time to dynamically transform the collected coordinates to the global coordinate system, ensuring that the accuracy of the spatial position coordinates is within ±1mm.

[0188] Step S32: Based on the spatial position coordinates and the horizontal and vertical field of view parameters of the detection viewpoint, construct a three-dimensional frustum model of the detection device.

[0189] In this embodiment, the horizontal field of view is the maximum angle range that the detection device can capture in the horizontal direction, the vertical field of view is the maximum angle range in the vertical direction, and the three-dimensional field of view model is a spatial geometric model that simulates the shooting range of the detection device. Its vertex is the optical center of the lens of the detection device, and the bottom edge is determined by the horizontal field of view and the vertical field of view.

[0190] As an optional implementation, the spatial coordinates of the detection device are used as the vertices. The direction vectors of the four edges of the frustum are calculated based on the horizontal and vertical field of view. The depth range of the frustum is determined by combining the lens focal length of the detection device, thereby constructing a complete three-dimensional frustum model.

[0191] As another alternative implementation, the view frustum construction function in the OpenGL graphics library is used. By inputting the spatial position coordinates, horizontal field of view, vertical field of view, and near and far clipping plane parameters, a three-dimensional view frustum model is quickly generated, and the relative positional relationship between the view frustum and the building model is displayed in real time through a visualization interface.

[0192] Step S33: Calculate the spatial intersection of the 3D view frustum model and the building model to obtain the initial mapping region.

[0193] In this embodiment, the spatial intersection is the part where the 3D view frustum model and the building model overlap, and the initial mapping area is the area corresponding to this intersection on the building model, which is the initial range of the scene frame mapped in the building model.

[0194] As an optional implementation, the axis-aligned bounding box (AABB) collision detection algorithm is adopted. First, the bounding boxes of each component of the 3D view frustum model and the building model are calculated. The parts that may have intersections are filtered out by the intersection detection of the bounding boxes. Then, the intersection calculation at the triangular facet level is performed on these parts to determine the spatial intersection and obtain the initial mapping region.

[0195] As another alternative implementation, a ray casting algorithm is used to emit rays from the vertices of the 3D view frustum model to the bottom edge, calculate the intersection points of the rays with the surface of the building model, and connect these intersection points to form a polygonal region as the initial mapping region. This method is suitable for scenarios where the building model is a mesh model.

[0196] Step S34: Extract edge feature points from the scene frame and match the edge feature points with the surface feature points of the building model.

[0197] In this embodiment, edge feature points are key pixels on the edges of objects in the scene frame, which have obvious grayscale changes or gradient features; surface feature points of the building model are points on the surface of the building model that have unique geometric shapes or textures, such as corner vertices, door and window frame corners, etc.

[0198] As an optional implementation, the Canny edge detection algorithm is used to extract edge feature points from the scene frame, and the SIFT algorithm is used to calculate the descriptors of these feature points. At the same time, surface feature points are extracted from the building model and their SIFT descriptors are calculated. The feature points of the two are matched using the nearest neighbor matching algorithm, and the matching threshold is set to 0.7.

[0199] As another alternative implementation, a deep learning model (such as CNN) is used to extract features from the scene frames to obtain high-dimensional feature vectors of edge feature points. A point cloud registration algorithm is then used to match these vectors with the point cloud data of feature points on the surface of the building model. False matching points are then removed using the RANSAC algorithm to improve matching accuracy.

[0200] Step S35: Correct the initial mapping area based on the matching result to obtain the mapping area corresponding to the scene frame in the building model.

[0201] In this embodiment, the matching results include the correspondence between edge feature points and surface feature points and the matching error. The correction process is to adjust the boundary of the initial mapping area based on these results so that the mapping area more accurately reflects the building model area captured by the on-site frame.

[0202] As an optional implementation, the correction amount of the mapping area is calculated based on the coordinate deviation of the matching point pair, and the boundary of the initial mapping area is translated and rotated to make the projection of the edge feature points on the building model coincide with the corresponding surface feature points. The boundary error of the corrected mapping area is controlled within ±0.5mm.

[0203] As another alternative implementation, an optimization function based on the matching results is constructed. The initial mapping region is used as the initial value, and the spatial distance error of the matching point pairs is minimized through the gradient descent algorithm. The range and shape of the mapping region are iteratively optimized to finally obtain the accurate mapping region.

[0204] For example, in waterproofing inspection, the inspection device acquires a set of on-site image frames. The acquired coordinates are transformed to the global coordinate system of the building model using a coordinate transformation matrix, resulting in spatial coordinates of (30.5m, 15.2m, 8.3m). The horizontal field of view of the inspection device is 60°, and the vertical field of view is 45°, based on which a three-dimensional view frustum model is constructed. The spatial intersection of this view frustum and the office building BIM model is calculated, yielding an initial mapped area of ​​a rectangular region on the exterior wall. Edge feature points from the on-site image frames are extracted and SIFT-matched with surface feature points such as wall corners and window frames in the BIM model of this region. The matching results reveal a slight offset in the initial mapped area. Correction amounts are calculated and adjusted to obtain an accurate mapped area that perfectly corresponds to the actual area captured in the on-site image frames.

[0205] This embodiment achieves high-precision determination of the mapping area of ​​the on-site image frame in the building model through precise coordinate transformation, view frustum construction, spatial intersection calculation, feature point matching, and region correction, providing a reliable spatial reference for the accurate analysis and presentation of subsequent crack information.

[0206] For example, to help understand the technical concept or principle of the fluorescent tracer-based waterproof detection method combined with Embodiments 1 and 2 above, please refer to... Figure 4 , Figure 4 A simplified flowchart of a waterproof detection method based on fluorescence tracers is provided below:

[0207] First, quantum dot injection: A solution or material containing quantum dots (such as CdSe / ZnS quantum dots) is injected into the waterproof structure to be detected. Quantum dots have unique fluorescence properties and can emit fluorescence under specific conditions, facilitating subsequent detection. Then, UV excitation: The injected quantum dot area is irradiated with an ultraviolet (UV) light source. The quantum dots emit fluorescence under UV excitation, making potential leakage locations visible. Second, fluorescence capture: High-sensitivity optical detection equipment (such as high-resolution cameras and fluorescence imagers) is used to acquire fluorescence images generated under UV excitation, recording information such as fluorescence distribution and intensity. Third, image preprocessing: The captured fluorescence images undergo preprocessing operations such as noise reduction, contrast enhancement, and grayscale conversion to remove interference factors, improve image quality, and make fluorescence features clearer and more obvious, facilitating subsequent analysis. Finally, AI analysis: Artificial intelligence algorithms (such as convolutional neural networks in deep learning) are used to analyze the preprocessed images, identifying features in the fluorescence images that may represent leakage, such as abnormal fluorescence areas and changes in fluorescence intensity, to determine whether leakage exists and its location and extent. This leads to 3D reconstruction: Combining positioning data from the detection device (such as location information obtained through IMU data and positioning modules) and image analysis results, information such as the location of the leak is reconstructed in 3D within the building model (such as a BIM model), visually presenting the specific location and shape of the leak in space. Finally, a leak report is generated: Based on the 3D reconstruction results and analysis data, a detailed leak report is generated, including the location of the leak, the degree of leakage, and recommended repair measures, providing a basis for subsequent waterproofing repair work.

[0208] For example, in an indoor waterproofing inspection scenario in a residential building, before conducting the inspection, for areas prone to leakage, such as bathrooms and kitchens, an aqueous gel material containing CdSe / ZnS quantum dots is injected using an injection device into the tile grout lines on the walls and floors, as well as areas where the waterproofing coating may have weak points. The quantum dots are evenly dispersed in the material, allowing them to penetrate into the waterproofing structure. Inspectors then enter the building with a portable ultraviolet lamp and, following a predetermined inspection route, thoroughly irradiate the areas where the quantum dots have been injected. The ultraviolet lamp has a power of 30W to ensure irradiation intensity and uniformity, allowing the quantum dots to be fully excited and produce fluorescence. A fluorescence imaging camera equipped with a high-sensitivity CMOS image sensor, with a resolution of 4000×3000 pixels, is used to photograph the UV-excited areas. The camera is set with automatic exposure and white balance functions to accurately capture fluorescence images and record the fluorescence status of each inspection area. The captured fluorescence images are transmitted to the computer in the inspection system. Gaussian filtering is used to reduce noise in the images, histogram equalization is used to enhance image contrast, and the color images are converted to grayscale to highlight the difference between the fluorescence and the background. A convolutional neural network model trained on a large amount of fluorescence image data of leaks was used to analyze the preprocessed images. The model identified an area of ​​abnormally enhanced fluorescence near the corner of the bathroom floor, indicating a possible leak at this location. Based on the fluorescence intensity and distribution, the initial assessment was that the leak was minor. During the inspection, the inspectors used an IMU and UWB positioning module installed on the imaging camera to acquire location information in real time. Combined with the AI ​​analysis results, the leak location information was mapped onto the BIM model of the building complex, visually presenting the specific coordinates and extent of the leak on the bathroom floor in three-dimensional space. The leak area was shown to be irregularly shaped, with an area of ​​approximately 0.1 square meters. Based on the 3D reconstruction and analysis data, the inspection system automatically generated a leak report. The report clearly stated that the leak was located at the corner of the bathroom floor, the leak was minor, and recommended resealing the tile gaps with sealant and local repair of the waterproof coating in the area. Images of the leak location and screenshots of the 3D model were also included to facilitate the repair work.

[0209] This application provides a fluorescent tracer-based waterproof detection device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the fluorescent tracer-based waterproof detection method in Embodiment 1 above.

[0210] The following is for reference. Figure 5The diagram illustrates a structural schematic of a fluorescent tracer-based waterproof detection device suitable for implementing embodiments of this application. The fluorescent tracer-based waterproof detection device in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, personal digital assistants (PDAs), tablets, and in-vehicle terminals, as well as fixed terminals such as digital TVs and desktop computers. Figure 5 The illustrated fluorescent tracer-based waterproof detection device is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this application.

[0211] like Figure 5 As shown, the fluorescent tracer-based waterproof testing device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the fluorescent tracer-based waterproof testing device. The processing unit 1001, the ROM 1002, and the RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the fluorescence tracer-based waterproof detection device to exchange data wirelessly or via wired communication with other devices. Although fluorescence tracer-based waterproof detection devices with various systems are shown in the figures, it should be understood that it is not required to implement or possess all of the systems shown. More or fewer systems can be implemented alternatively.

[0212] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0213] The fluorescent tracer-based waterproof detection device provided in this application, employing the fluorescent tracer-based waterproof detection method described in the above embodiments, can solve the technical problem of difficulty in distinguishing minute cracks due to the inability to determine the specific location and condition of cracks. Compared with the prior art, the beneficial effects of the fluorescent tracer-based waterproof detection device provided in this application are the same as those of the fluorescent tracer-based waterproof detection device provided in the above embodiments, and other technical features of this fluorescent tracer-based waterproof detection device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0214] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0215] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0216] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the fluorescent tracer-based waterproof detection method in the above embodiments.

[0217] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, radio frequency (RF), or any suitable combination thereof.

[0218] The aforementioned computer-readable storage medium may be included in a fluorescence-based waterproof testing device; or it may exist independently and not assembled into a fluorescence-based waterproof testing device.

[0219] The aforementioned computer-readable storage medium carries one or more programs that, when executed by the fluorescence-based waterproofing detection device, cause the fluorescence-based waterproofing detection device to: receive on-site image frames acquired by the detection device, IMU data uploaded in real time by the detection device, and positioning data; determine the acquisition coordinates corresponding to the on-site image frames based on the IMU data and the positioning data; determine the mapping area corresponding to the on-site image frames in the building model based on the acquisition coordinates and the detection perspective parameters of the detection device; determine the crack morphology and seepage velocity of the mapping area based on the on-site image frames; and render cracks in the building model based on the crack morphology and seepage velocity.

[0220] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0221] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0222] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0223] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described fluorescent tracer-based waterproof detection method. This solves the technical problem of difficulty in distinguishing minute cracks due to the inability to determine their specific location and condition. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the fluorescent tracer-based waterproof detection method provided in the above embodiments, and will not be repeated here.

[0224] This application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described fluorescent tracer-based waterproof detection method.

[0225] The computer program product provided in this application can solve the technical problem of difficulty in distinguishing minute cracks due to the inability to determine the specific location and condition of cracks. Compared with the prior art, the beneficial effects of the computer program product provided in the embodiments of this application are the same as the beneficial effects of the fluorescent tracer-based waterproof detection method provided in the above embodiments, and will not be repeated here.

[0226] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent scope of this application.

Claims

1. A waterproof detection method based on fluorescence tracer, characterized in that, The waterproof detection method based on fluorescence tracing includes: Receives on-site image frames collected by the detection device, IMU data uploaded by the detection device in real time, and positioning data; Based on the IMU data and the positioning data, determine the acquisition coordinates corresponding to the scene frame; Based on the acquisition coordinates and the detection perspective parameters of the detection device, the mapping area corresponding to the on-site image frame in the building model is determined; The crack morphology and seepage velocity in the mapped area are determined based on the on-site video frames. The cracks are rendered in the building model based on the crack morphology and the seepage velocity.

2. The waterproof detection method based on fluorescence tracer as described in claim 1, characterized in that, The step of determining the crack morphology and seepage velocity of the mapped area based on the on-site image frame includes: The on-site image frames are input into a pre-trained image processing model, and the fluorescence morphology recognition results and flow dynamics calculation results are obtained through the image processing model. The spatial location corresponding to the crack is determined based on the fluorescence morphology recognition results; The seepage velocity at each location corresponding to the crack is determined based on the fluid dynamics calculation results.

3. The waterproof detection method based on fluorescence tracer as described in claim 1, characterized in that, The step of rendering cracks in the building model based on the crack morphology and the seepage velocity includes: Determine the starting point of the crack; The second seepage velocity corresponding to each point along the crack propagation direction from the starting point is obtained, and the corresponding seepage velocity range of the starting point is determined, wherein the seepage velocity range is the second seepage velocity corresponding to the starting point ± a set value. If the second seepage velocity is detected to exceed the seepage velocity range, the crack color between the starting point and the current point is determined based on the average seepage velocity between the starting point and the current point. If the end point of the crack is not reached, the current point is taken as the starting point, and the process jumps to the step of determining the starting point of the crack. Otherwise, the crack is rendered in the building model based on the crack color and the crack shape.

4. The waterproof detection method based on fluorescence tracer as described in claim 1, characterized in that, The step of rendering cracks in the building model based on the crack morphology and the seepage velocity includes: Determine the average seepage velocity between the start and end points of the crack; The color of the crack between the starting point and the ending point is determined based on the average seepage velocity. The crack is rendered in the building model based on the crack color and crack shape.

5. The waterproof detection method based on fluorescence tracer as described in claim 3 or 4, characterized in that, Following the step of rendering cracks based on crack morphology in the building model, the waterproofing detection method based on fluorescence tracers further includes: Obtain the crack color-alarm information mapping relationship; A water leakage alarm is generated based on the crack color and the crack color-alarm information mapping relationship.

6. The waterproof detection method based on fluorescence tracer as described in claim 1, characterized in that, Before the step of receiving the on-site image frames collected by the detection device, the IMU data uploaded by the detection device in real time, and the positioning data, the following steps are included: Calibrate the spraying device onto the architectural model; In response to a jet command, determine the jet parameters and jet trajectory corresponding to the jet command; The moving component of the spraying device is controlled to move according to the spraying trajectory, and the spraying component of the spraying device is controlled to spray the mixed material according to the spraying parameters, wherein the mixed material includes a first component and a second component.

7. The waterproof detection method based on fluorescence tracer as described in claim 6, characterized in that, The step of calibrating the spraying device onto the building model includes: Obtain the calibration position of the building model; Control the spraying device to align with the calibrated position in the construction area; The spray device is calibrated based on the calibration position.

8. The waterproof detection method based on fluorescence tracer as described in claim 1, characterized in that, The step of determining the mapping area of ​​the on-site image frame in the building model based on the acquired coordinates and the detection perspective parameters of the detection device includes: The collected coordinates are transformed to the global coordinate system of the building model to obtain the spatial position coordinates of the detection device in the building model; Based on the spatial position coordinates and the horizontal and vertical field of view parameters in the detection perspective parameters, a three-dimensional frustum model of the detection device is constructed. Calculate the spatial intersection between the 3D view frustum model and the building model to obtain the initial mapping region; Extract edge feature points from the scene frame and match the edge feature points with the surface feature points of the building model; The initial mapping region is corrected based on the matching results to obtain the mapping region corresponding to the scene frame in the building model.

9. A waterproof detection device based on fluorescence tracing, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the fluorescent tracer-based waterproof detection method as described in any one of claims 1 to 8.

10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the waterproof detection method based on fluorescence tracing as described in any one of claims 1 to 8.

Images