A spraying robot spraying quality evaluation method based on laser point cloud

Abstract

This invention relates to a method for evaluating the coating quality of a spraying robot based on laser point clouds, belonging to the field of coating quality inspection technology. The method includes the following steps: S1: Collecting baseline data before coating and target data after coating using a lidar sensor; S2: Performing time-series averaging on multiple consecutively collected point cloud frames before and after coating; S3: Constructing a triangular mesh model of the baseline surface before coating using the Delaunay triangulation method; S4: Calculating the minimum vertical distance from the point cloud to the triangular mesh to quantize the thickness; S5: Using a multi-filtering process to remove noise and outliers; S6: Generating a coating thickness heatmap based on thickness statistics and color mapping.

Description

Technical Field

[0001] This invention belongs to the field of spraying quality inspection technology, and relates to a method for evaluating the spraying quality of a spraying robot based on laser point cloud. Background Technology

[0002] Traditional methods for inspecting the quality of tunnel coatings primarily rely on manual observation and borehole sampling. Manual observation, often involving hammering and listening, heavily depends on the inspector's subjective experience, making it difficult to establish quantifiable evaluation standards and resulting in low efficiency. While borehole sampling can obtain specific thickness data, it is a destructive method that damages the existing support structure. Furthermore, the discrete distribution of sampling points limits the sample size, failing to accurately reflect the overall quality distribution and localized weak points within the tunnel. Therefore, traditional inspection methods are insufficient to meet the real-time, comprehensive quality inspection needs of coating robots.

[0003] The mixed reality-based in-situ visualization method for shotcrete thickness estimates the theoretical deposition thickness by tracking the pose of a robotic arm and using a collision model of virtual particles and environmental meshes. However, this method is essentially a virtual simulation based on process parameters, rather than a real physical measurement of the post-shotcrete surface, making it difficult to accurately obtain the true final surface morphology and construction quality. The fully automated intelligent tunnel wet shotcrete truck and method with machine vision utilizes line lasers and industrial cameras to detect shotcrete thickness in real time. It controls wet shotcreting through "S"-shaped path planning, extracts the center point of the laser strip, and uses the root mean square error of the fitted straight line to determine if the thickness meets the standard. This method only uses line structured light for single-line measurement and does not construct a complete three-dimensional surface model, failing to comprehensively reflect the overall thickness distribution and smoothness of the shotcrete layer. The intelligent shotcrete system utilizes a multi-degree-of-freedom 3D scanner group and dual robotic arms deployed on the shotcrete bridge to execute an automated workflow of "scanning-initial shotcreting-re-scanning-comparison-fine shotcreting correction." This method focuses on the automated closed-loop control of construction actions, and the evaluation process performs a simple numerical comparison between the scan results and a preset contour to determine whether it is qualified.

[0004] Three-dimensional laser scanning technology provides a non-contact measurement method for obtaining high-precision geometric information of tunnel surfaces. However, in the complex environment of high dust, low illumination and equipment vibration in actual underground mines, the collected point cloud data has significant noise, and the surface morphology before and after spraying is very different, which significantly increases the complexity of data processing and the uncertainty of analysis. Summary of the Invention

[0005] In view of this, the purpose of this invention is to meet the needs of spraying robots for real-time, comprehensive, and quantitative evaluation of spraying quality, and to solve the problems of slow detection speed, low coverage, and insufficient automation in existing tunnel spraying quality detection technologies. This invention provides a method that can achieve non-contact, full-coverage, automatic quantitative evaluation and visualization analysis of coating thickness and smoothness, aiming to improve detection efficiency and coverage, and generate quantitative results.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for evaluating the spraying quality of a spraying robot based on laser point clouds includes the following steps: S1: Use a lidar sensor to collect baseline data before spraying and target data after spraying by the spraying robot; S2: Perform time-series averaging on multiple point cloud frames continuously acquired before and after spraying; S3: The Delaunay triangulation method is used to construct a triangular mesh model of the reference surface before spraying; S4: Calculate the minimum vertical distance from the point cloud to the triangular mesh to quantize the thickness; S5: Employs a multi-filtering process to remove noise and outliers; S6: Generate a thermal map of spray coating thickness based on thickness statistics and color mapping.

[0007] Furthermore, step S1 specifically includes the following steps: Before the spraying operation begins, the lidar sensor is exposed to the working environment and put into working condition; the lidar performs a rotational scan on the original surface of the tunnel to be sprayed, and obtains the three-dimensional geometric shape of the tunnel surface before spraying, as a reference data. After the spraying operation is completed, the lidar is used to perform a second scan of the sprayed tunnel surface at the same frame rate to obtain the three-dimensional shape of the tunnel surface covered by the sprayed coating. The resulting point cloud data is the target data used for quality assessment and thickness calculation.

[0008] Furthermore, step S2 specifically includes the following steps: Continuously acquire and cache the sequential data before and after spraying. Frame point cloud data, for the same pixel coordinates The spatial point sequence below is averaged in the time domain; let the first... k The coordinates of a point in a frame point cloud are The point cloud coordinates after mean processing Calculate using the following formula:

[0009] in, This represents the pixel coordinates of the point cloud in the k-th frame. The three-dimensional coordinate vector at the location; The point cloud data before and after spraying are subjected to the time-series averaging process described above to generate the baseline point cloud before spraying. With the target point cloud after spraying .

[0010] Furthermore, step S3 specifically includes the following steps: S31: Point Cloud Preprocessing is performed, including removing outlier noise points and smoothing the point cloud using moving least squares as needed to suppress high-frequency noise. S32: To perform triangulation, the 3D point cloud is projected onto an optimally fitted 2D plane, which is determined by principal component analysis of all points and spanned by the directions of the largest and second-largest principal components; the projected 2D point set is denoted as... ; S33: Point set on a two-dimensional projection plane Perform Delaunay triangulation to generate a two-dimensional triangular mesh. This process forms triangles by connecting adjacent points and strictly adheres to the Delaunay criterion, which requires each triangle to satisfy the conditions of maximizing the minimum angle and the empty circle property. S34: Map each vertex in the 2D triangular mesh back to 3D space based on its original 3D coordinates and projection relationships, reconstructing a 3D triangular mesh model that covers the original point cloud. .

[0011] Furthermore, step S4 specifically includes the following steps: S41: For All triangular faces are used to construct a spatial index structure to locate the target point. Candidate triangular facets with the closest spatial location For candidate face patches From the vertex Define the normal vector as Calculation points Signed distance to the plane containing the patch :

[0012] S42: Calculation point Projection point on the plane :

[0013] Determine the projection point using the centroid coordinate method. Does it fall on a triangular facet? Internal: If the conditions are met, then the thickness If the projection point is located outside the triangular facet, then the calculation point... The shortest distance to the three sides of the triangular face ; S43: Thickness value is... Traverse all target points to generate initial thickness point cloud data. The intensity value at each point stores the corresponding coating thickness. .

[0014] Furthermore, step S5 specifically includes the following steps: S51: Perform connected component analysis and small region suppression: Treat the thickness point cloud as a two-dimensional intensity image, and identify connected regions based on eight-neighbor search; for regions with areas smaller than a threshold... Isolated small regions are identified as noise, and their thickness values ​​are replaced with the average thickness of their effective neighboring points. S52: Adaptive median filtering: Assuming the point cloud is ordered, for each point's thickness value, the median of all valid thickness values ​​within its local window is taken as the output to smooth impulse noise. The process is described as follows:

[0015] in For point The neighborhood window; S53: Perform Gaussian smoothing filtering: Use a two-dimensional Gaussian kernel function to perform a weighted average of the thickness values ​​to eliminate high-frequency fluctuations. Defined as:

[0016] Through the above multi-stage filtering, the denoised thickness point cloud is obtained.

[0017] Furthermore, step S6 specifically includes the following steps: S61: Based on the preset thickness acceptable threshold For each point, determine its thickness: Then, a specific color representing qualification is assigned; for thickness... For defective points, a color gradient mapping is applied based on the degree of thickness deviation to normalize the thickness value. Interval:

[0018] in for The minimum thickness value in; S62: Normalization parameters are obtained through a predefined color mapping function. Converted to the corresponding RGB colors, the distribution and severity of insufficient thickness are clearly revealed; S63: Generate point clouds with color information The information is then displayed as a visual message and integrated into the human-machine interface of the painting robot to provide quality feedback and decision support for operators.

[0019] The beneficial effects of this invention are as follows: (1) This invention effectively suppresses random noise in single-frame measurement under complex downhole environments by performing time-series averaging on multiple frames of point cloud data collected continuously before and after spraying, significantly improving the signal-to-noise ratio and geometric accuracy of the original point cloud data, providing a high-quality and highly reliable data foundation for subsequent analysis, and overcoming the shortcomings of low accuracy and susceptibility to interference in traditional manual observation and borehole sampling methods. (2) The present invention uses the Delaunay triangulation method to construct triangular mesh models of the reference surface before spraying. The generated mesh is uniform and of high quality, and can accurately and completely represent the complex three-dimensional geometric shape of the tunnel surface. Compared with the existing mixed reality-based in-situ visualization simulation method of spraying thickness, it provides a more realistic and robust geometric reference for thickness calculation, ensuring the comprehensiveness and accuracy of the evaluation. (3) The present invention defines and quantifies the coating thickness by calculating the minimum vertical distance from the target point to the reference triangular mesh. This method is based on accurate scanning data and geometric model, combined with nearest neighbor search and accurate point-to-surface / point-to-edge distance calculation, and realizes full-field, continuous and automated measurement of coating thickness on complex curved surfaces with millimeter-level accuracy. It solves the problem that the existing technology cannot truly and comprehensively quantify the final coating thickness. (4) The present invention adopts a multi-filtering process combining connected component analysis, median filtering and Gaussian smoothing for the initial thickness calculation results, which can effectively identify and remove noise points and outliers, smooth data fluctuations, and thus extract real and continuous thickness distribution information from noisy data, thereby enhancing the robustness of the evaluation algorithm under harsh working conditions and the credibility of the results. (5) Based on the filtered thickness data, the present invention automatically calculates key statistical indicators and generates an intuitive thickness distribution heat map by color mapping according to the preset qualified threshold. This transforms the abstract massive three-dimensional measurement data into easily understandable visual information, enabling operators or control systems to quickly and accurately grasp the overall quality status and local details. It realizes the instantaneous and visual feedback of the detection results, and improves the efficiency of quality monitoring and decision support capabilities.

[0020] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0021] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is a flowchart of a laser scanning-based method for detecting the quality of roadway spraying, according to an embodiment of the present invention. Figure 2 This is a schematic diagram showing the installation position of the lidar on the painting robot according to an embodiment of the present invention; Figure 3 This is a flowchart of the spraying quality inspection process according to an embodiment of the present invention; Figure 4 (a) and (b) are schematic diagrams of the reference grid and reference model of the roadway before spraying, respectively, according to an embodiment of the present invention. Figure 5 This is a schematic diagram of the coating thickness thermal curve according to an embodiment of the present invention; Figure 6 This is a web-based visualization platform according to an embodiment of the present invention. Detailed Implementation

[0022] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0023] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0024] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0025] Example 1: like Figure 1As shown, this embodiment of the invention is a method for evaluating the spraying quality of a spraying robot based on laser point clouds. This method uses a laser radar sensor mounted on the spraying robot to collect three-dimensional point cloud data of the tunnel surface before and after the spraying operation. It then sequentially performs point cloud mean processing, triangular mesh construction, thickness calculation and mapping, noise filtering, and statistical analysis steps to achieve automated evaluation and visualization of key quality indicators such as coating thickness distribution and smoothness. This embodiment describes the quality evaluation process executed within a single spraying operation cycle, which can be repeated in subsequent operation cycles. The specific evaluation process includes the following steps: (1) Data collection A protective device is installed in front of the robotic arm of the painting robot. Inside this device is a fixed 16-line mechanical lidar sensor. Figure 2 As shown, the data acquisition process is divided into two stages: baseline data acquisition before spraying and target data acquisition after spraying.

[0026] Before the spraying operation begins, the control system first sends a command to activate the lidar protection device, exposing the lidar sensor to the working environment and putting it into operation. The lidar performs a rotating scan of the original surface of the tunnel to be sprayed at a fixed frequency of 10 frames per second, acquiring the three-dimensional geometric shape of the tunnel surface before spraying. The continuous point cloud data collected in this stage, after subsequent processing, will be used for two purposes: first, as the original spatial reference for the spraying robot to perform path planning and trajectory generation; second, as the geometric benchmark for calculating the spray thickness in subsequent spraying quality assessment, i.e., the initial benchmark point cloud data.

[0027] After the spraying operation is completed, to assess the quality of the sprayed layer, the control system activates the protection device again and commands the lidar to perform a second scan of the sprayed tunnel surface at the same frame rate. This scan aims to acquire the three-dimensional topography of the tunnel surface covered by the sprayed layer; the resulting point cloud data serves as the target point cloud data for quality assessment and thickness calculation. The overall process is as follows: Figure 3 As shown. After all data acquisition is complete, the control system shuts down the protection device to protect the sensor from subsequent operations or environmental factors.

[0028] (2) Perform time-series averaging on multiple point clouds collected continuously before and after spraying.

[0029] In the downhole environment, equipment vibration and inherent sensor noise can cause random errors in single-frame laser point cloud data, affecting its geometric accuracy. To improve data quality, this embodiment performs time-series averaging on multiple consecutive point cloud frames acquired before and after spraying. This process aims to suppress random noise through statistical averaging in the time domain, obtaining point clouds with higher signal-to-noise ratios for subsequent modeling and calculation.

[0030] In practice, the system allocates separate data buffers for the pre-coating and post-coating stages. In each stage, the lidar continuously acquires N frames (e.g., N=10) of raw point cloud data at a fixed frame rate (e.g., 10 FPS) and stores them in the corresponding buffer. Because a mechanical rotating lidar is used, its scanning mode is fixed, ensuring that the pixel coordinate structure (i.e., height and width dimensions) of the point cloud remains stable across consecutive frames. For all frames of point cloud data within the buffer, points at the same pixel coordinate position... 3D points Calculate their arithmetic mean to generate the merged point coordinates for that location. The calculation method is as follows:

[0031] in, This represents the pixel coordinates of the point cloud in the k-th frame. The system calculates the three-dimensional coordinate vector at a given location. Before performing the summation and averaging, the system first performs a preliminary validity check on each frame of point cloud, discarding obviously invalid data points (such as those with coordinates NaN or infinity) to ensure that all data points participating in the averaging calculation are valid measurements.

[0032] Through the above time-series averaging process, two high-quality point cloud datasets were generated: one representing the original surface of the alleyway before spraying, denoted as the baseline point cloud. Another set represents the surface of the tunnel after spraying, denoted as the target point cloud. This step significantly reduces the uncertainty of a single measurement, providing a reliable data foundation for subsequent accurate geometric modeling and thickness calculation.

[0033] (3) The Delaunay triangulation method was used to construct the triangular mesh model of the reference surface before spraying.

[0034] Based on the high-quality reference point cloud generated in step 2) before spraying This embodiment constructs its corresponding continuous triangular mesh surface model. This provides an accurate geometric reference for thickness calculation. To accurately reconstruct the surface geometry of the tunnel, this embodiment uses the Delaunay triangulation method to mesh the point cloud. This method can generate a set of triangles that satisfy the empty circle property (i.e., the circumcircle of any triangle does not contain other input points), thus mathematically ensuring the uniformity of the mesh and avoiding the generation of extremely long and narrow low-quality triangles, which is beneficial to the numerical stability of subsequent geometric calculations. The specific mesh generation process includes the following steps: First, analyze the baseline point cloud. Preprocessing is performed. A statistical outlier removal algorithm is used to identify and remove discrete noise points whose average distance to their nearest neighbors exceeds a certain multiple of the standard deviation. Subsequently, moving least squares can be selectively applied to smooth the remaining point cloud to fit local surfaces and suppress high-frequency noise, providing a cleaner and more regular set of input points for triangulation.

[0035] To perform standard 2D Delaunay triangulation, the 3D point cloud needs to be projected onto a 2D plane. This projection plane is determined through principal component analysis: the covariance matrix of all points in the point cloud is calculated, eigenvalue decomposition is performed, and the plane spanned by the eigenvectors corresponding to the largest and second-largest eigenvalues ​​is selected as the optimal fitting plane. All 3D points are orthogonally projected onto this plane to obtain the 2D point set. This step transforms the problem into a two-dimensional spatial processing method while preserving the main geometric and topological relationships of the point cloud.

[0036] In the two-dimensional plane obtained by projection, for the set of points Perform standard Delaunay triangulation. Specifically, the Bowyer-Watson algorithm can be used. This process automatically connects adjacent points to form a triangular mesh and ensures that each generated triangle satisfies the Delaunay criterion, which states that in a two-dimensional plane, the circumcircle of any triangle does not contain a set of points. Other points in the array. This generates a two-dimensional mesh composed of triangles.

[0037] For each vertex in the 2D triangular mesh generated in the previous step, based on its original 2D projected coordinates and known projection plane parameters, its original 3D coordinates are recovered through inverse transformation. The corresponding three-dimensional coordinates are then used. In this way, the topological connectivity of the two-dimensional mesh is directly applied to the three-dimensional points, thereby reconstructing a three-dimensional triangular mesh model that fully covers the spatial extent of the original point cloud. This model can accurately reflect the macroscopic undulations and local details of the original surface of the tunnel. Figure 4 Figures (a) and (b) show schematic diagrams of the triangular mesh and model generated based on point clouds.

[0038] (4) Calculate the minimum vertical distance quantization thickness from the point cloud to the triangular mesh.

[0039] Target point cloud Points in the model relative to the reference mesh The offset of the surface in the normal direction is defined in this method as the thickness of the coating at that point. This is for accurate calculation of the target point cloud. any point in the middle To the baseline mesh model To determine the minimum vertical distance to the surface, this embodiment employs a precise distance calculation method combining spatial index search and geometric projection. The specific steps are as follows: ① First, a baseline triangular mesh model is established. Construct a spatial index data structure, namely a KD-Tree based on the center points of all triangular facets. For the target point cloud... Each point in Use this index to quickly retrieve One or more candidate triangular faces that are spatially closest to it Assuming Composed of three vertices Defined, its unit normal vector is (Depend on (Obtained through normalization). Calculation points. to the dough Sign distance of the infinitely extending plane :

[0040] Where the distance symbol represents a point Relative to the plane normal The location.

[0041] ② Next, calculate the points Vertical projection point on the plane :

[0042] Then, the projection point is determined using the centroid coordinate method. Is it located in the candidate triangle? Inside the boundary. Calculation with A triangle with vertex A is relative to point B barycentric coordinates If satisfied and Then determine In triangle Inside. At this moment, point to the dough The vertical distance is the effective thickness. .

[0043] ③If the projection point Located in triangle Externally, the minimum distance exists at the point. Find the shortest distance to the three sides of the triangle. Calculate the points respectively. to line segment , , The minimum value among the nearest points is denoted as . At this point, the thickness... Take the smaller value between the vertical distance and the minimum edge distance: .

[0044] For target point cloud Each point in Repeat steps ① to ③ above to calculate the corresponding coating thickness. The thickness value As intensity information, it is assigned to the coordinate data of that point, ultimately generating an initial thickness point cloud dataset. Each data point in this point cloud contains its spatial location. and a scalar value characterizing the coating thickness at that location. .

[0045] (5) Use a multi-filtering process to remove noise and outliers.

[0046] Due to scanning errors in the complex downhole environment, residual errors in point cloud registration, and potential geometric singularities in thickness calculation, the initial thickness point cloud generated in step 4) The data typically contains discrete noise points, outliers, and local high-frequency fluctuations. To extract information that accurately reflects the continuous distribution of the coating thickness, this embodiment employs a multi-stage filtering process to remove noise data.

[0047] The specific multi-stage filtering process includes the following three steps in sequence: ① First, using the regular two-dimensional grid structure after reprojection, the three-dimensional thickness point cloud is... Each point in thickness value The extracted data is filled into a 2D matrix of the same size as the point cloud image, thus transforming the 3D point cloud processing problem into a 2D image processing problem. On this 2D thickness image, a connected component labeling algorithm based on eight neighborhoods (top, bottom, left, right, and four diagonal directions) is performed to identify all spatially continuous pixel regions with similar thickness values ​​within a preset tolerance range. The pixel area of ​​each labeled connected region is calculated. An area threshold is set. (In this example, the area is 0.01 square meters), the area is smaller than Connected components are identified as isolated small regions formed by noise or anomalies. For these small regions identified as noise, instead of directly deleting their data points, the thickness values ​​of all pixels within them are replaced with the arithmetic mean of the thickness values ​​of all valid adjacent pixels on the boundary of that region. This step aims to eliminate physically unreasonable local thickness abrupt changes caused by a few erroneous measurement points.

[0048] ② After small-region suppression, median filtering is applied to the two-dimensional thickness image to smooth salt-and-pepper noise. Median filtering is a non-linear filtering technique. Its principle is to scan each pixel in the image using a sliding window, sort the thickness values ​​of all pixels within the window, and take the median as the output value of the center pixel. This process is performed on each pixel... The filtering operation can be formally described as follows:

[0049] in, Represented in pixels A local neighborhood window centered on the center, typically using a size of A rectangular window (in this example) Invalid values ​​are ignored during median calculation. This step effectively filters out isolated noise points while preserving edge information of the thickness distribution.

[0050] ③ To further enhance the spatial continuity of the thickness data and eliminate potential high-frequency random fluctuations remaining after median filtering, a Gaussian smoothing filter is finally applied to the thickness image. Gaussian filtering is a linear smoothing filter that achieves smoothing by weighted averaging of the neighborhood of each pixel; the weights are determined by a two-dimensional Gaussian function. (Two-dimensional Gaussian kernel function) Defined as:

[0051] in, It is the standard deviation of the Gaussian distribution, which controls the degree of smoothness (kernel width). In implementation, firstly, based on the selected... The value generates a discrete value of size. A Gaussian convolution kernel is applied, and then normalized so that the sum of all weights is 1. This Gaussian kernel is then convolved with the previously processed 2D thickness image in a 2D convolution operation. For each location in the image... Its smoothed thickness value The calculation is as follows:

[0052] After undergoing multiple connected component analysis, median filtering, and Gaussian smoothing processes, optimized two-dimensional thickness distribution data is obtained. This two-dimensional thickness distribution data is then remapped back to its corresponding three-dimensional spatial coordinates, generating a high-quality, high-reliability final thickness assessment point cloud dataset. This provides high-quality input data for subsequent statistical analysis and visualization.

[0053] (6) Generate a thermal map of spray coating thickness based on thickness statistics and color mapping.

[0054] Based on the final thickness point cloud dataset obtained in step 5) This embodiment first automatically calculates a set of global statistical indicators to quantitatively evaluate the overall coating quality. Key indicators calculated include: minimum coating thickness across the entire area. Maximum coating thickness in the entire venue Average coating thickness across the entire field, and standard deviation of the thickness. Among them, the thickness difference It is used as a core quantitative indicator to measure the surface smoothness of the sprayed coating; the smaller the range, the more uniform the thickness distribution.

[0055] To achieve an intuitive and efficient presentation of the evaluation results, this embodiment converts the numerical thickness data into a color-coded visual heatmap. The specific conversion and rendering process is as follows: ① Set a qualified coating thickness threshold according to engineering requirements or design specifications. Traversal For each data point in the dataset, determine its thickness value. Does it meet the requirements? For all "qualified" points that meet the criteria, a specific identification color, such as pure blue, is uniformly assigned to them, clearly marking the areas where the thickness meets the standard on the heat map.

[0056] ②Regarding thickness For the "non-conforming" points, the severity of their deficiencies needs to be further reflected using color. First, the thickness values ​​of all non-conforming points are normalized to a range. Internal, normalized parameters Calculate using the following formula:

[0057] in for The minimum thickness value in; ② Parameters It characterizes the relative degree of insufficient thickness. This indicates that the thickness is exactly the threshold. , This indicates the thickness is at its minimum value. Then, a predefined continuous color mapping function is used to... The value is mapped to the corresponding RGB color. This embodiment uses a gradient scheme that transitions from green through yellow to red: when When it is close to 0 (slightly insufficient), it is mapped to green; as... As it increases, the color gradually changes to yellow; when When the value is close to 1 (severely insufficient), it is mapped to red. This mapping relationship can be achieved through linear interpolation in the RGB color space.

[0058] After assigning color values ​​to all points, a new point cloud dataset is generated. This point cloud not only preserves the original spatial coordinate information, but each point also contains an RGB color value calculated based on its thickness and threshold relationship, thus forming a three-dimensional thermal map of the coating thickness, as illustrated in the diagram below. Figure 5 As shown. Finally, the Robot Operating System (ROS) will... The data is published as a topic message. This visualized message can be subscribed to in real time and rendered on the interactive interface accompanying the painting robot, such as... Figure 6 As shown, this allows operators to clearly understand the thickness distribution, quickly locate defective areas, and provide a direct basis for decision-making regarding whether respraying or process parameter adjustment is necessary.

[0059] Example 2: An electronic device, comprising a memory and a processor; The memory is used to store computer programs; The processor is configured to implement the method described in Embodiment 1 when executing the computer program.

[0060] Example 3: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in Embodiment 1.

[0061] Example 4: A computer program product includes a computer program that, when executed by a processor, implements the method described in Example 1.

[0062] In the above embodiments, the reference to "this embodiment" in the specification indicates that a specific feature, structure, or characteristic described in connection with the embodiment is included in at least some embodiments, but not necessarily all embodiments. Multiple appearances of "this embodiment" do not necessarily refer to the same embodiment.

[0063] In the above embodiments, although the invention has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory structures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed. The embodiments of the invention are intended to cover all such substitutions, modifications, and variations falling within the broad scope of the appended claims.

[0064] As will be understood by those skilled in the art, the computer-readable storage medium described in this embodiment allows for the implementation of all or part of the steps in the above method embodiments by computer program-related hardware. The aforementioned computer program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0065] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver, and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication between them. The memory is used to store computer programs, the communication interface is used to perform communication, and the processor and the transceiver are used to run the computer programs, so that the electronic terminal performs the steps of the above method.

[0066] In this embodiment, the memory may include random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device.

[0067] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0068] This invention can be used in a wide range of general-purpose or special-purpose computing system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices, etc.

[0069] This invention can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This invention can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0070] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for evaluating the spraying quality of a spraying robot based on laser point clouds, characterized in that: Includes the following steps: S1: Use a lidar sensor to collect baseline data before spraying and target data after spraying by the spraying robot; S2: Perform time-series averaging on multiple point cloud frames continuously acquired before and after spraying; S3: The Delaunay triangulation method is used to construct a triangular mesh model of the reference surface before spraying; S4: Calculate the minimum vertical distance from the point cloud to the triangular mesh to quantize the thickness; S5: Employs a multi-filtering process to remove noise and outliers; S6: Generate a thermal map of spray coating thickness based on thickness statistics and color mapping.

2. The method for evaluating the spraying quality of a spraying robot based on laser point clouds according to claim 1, characterized in that: Step S1 specifically includes the following steps: Before the spraying operation begins, the lidar sensor is exposed to the working environment and put into working condition; the lidar performs a rotational scan on the original surface of the tunnel to be sprayed, and obtains the three-dimensional geometric shape of the tunnel surface before spraying, as a reference data. After the spraying operation is completed, the lidar is used to perform a second scan of the sprayed tunnel surface at the same frame rate to obtain the three-dimensional shape of the tunnel surface covered by the sprayed coating. The resulting point cloud data is the target data used for quality assessment and thickness calculation.

3. The method for evaluating the spraying quality of a spraying robot based on laser point clouds according to claim 1, characterized in that: Step S2 specifically includes the following steps: Continuously acquire and cache the sequential data before and after spraying. Frame point cloud data, for the same pixel coordinates The spatial point sequence below is averaged in the time domain; let the first... k The coordinates of a point in a frame point cloud are The point cloud coordinates after mean processing Calculate using the following formula: in, This represents the pixel coordinates of the point cloud in the k-th frame. The three-dimensional coordinate vector at the location; The point cloud data before and after spraying are subjected to the time-series averaging process described above to generate the baseline point cloud before spraying. With the target point cloud after spraying .

4. The method for evaluating the spraying quality of a spraying robot based on laser point clouds according to claim 1, characterized in that: Step S3 specifically includes the following steps: S31: Point Cloud Preprocessing is performed, including removing outlier noise points and smoothing the point cloud using moving least squares as needed to suppress high-frequency noise. S32: To perform triangulation, the 3D point cloud is projected onto an optimally fitted 2D plane, which is determined by principal component analysis of all points and spanned by the directions of the largest and second-largest principal components; the projected 2D point set is denoted as... ; S33: Point set on a two-dimensional projection plane Perform Delaunay triangulation to generate a two-dimensional triangular mesh. This process forms triangles by connecting adjacent points and strictly adheres to the Delaunay criterion, which requires each triangle to satisfy the conditions of maximizing the minimum angle and the empty circle property. S34: Map each vertex in the 2D triangular mesh back to 3D space based on its original 3D coordinates and projection relationships, reconstructing a 3D triangular mesh model that covers the original point cloud. .

5. The method for evaluating the spraying quality of a spraying robot based on laser point clouds according to claim 1, characterized in that: Step S4 Specifically, the following steps are included: S41: For All triangular faces are used to construct a spatial index structure to locate the target point. Candidate triangular facets with the closest spatial location For candidate face patches From the vertex Define the normal vector as Calculation points Signed distance to the plane containing the patch : S42: Calculation point Projection point on the plane : Determine the projection point using the centroid coordinate method. Does it fall on a triangular facet? Internal: If the conditions are met, then the thickness If the projection point is located outside the triangular facet, then the calculation point... The shortest distance to the three sides of the triangular face ; S43: Thickness value is... Traverse all target points to generate initial thickness point cloud data. The intensity value at each point stores the corresponding coating thickness. .

6. The method for evaluating the spraying quality of a spraying robot based on laser point clouds according to claim 1, characterized in that: Step S5 specifically includes the following steps: S51: Perform connected component analysis and small region suppression: Treat the thickness point cloud as a two-dimensional intensity image, and identify connected regions based on eight-neighbor search; for regions with areas smaller than a threshold... Isolated small regions are identified as noise, and their thickness values ​​are replaced with the average thickness of their effective neighboring points. S52: Adaptive median filtering: Assuming the point cloud is ordered, for each point's thickness value, the median of all valid thickness values ​​within its local window is taken as the output to smooth impulse noise. The process is described as follows: in For point The neighborhood window; S53: Perform Gaussian smoothing filtering: Use a two-dimensional Gaussian kernel function to perform a weighted average of the thickness values ​​to eliminate high-frequency fluctuations. Defined as: Through the above multi-stage filtering, the denoised thickness point cloud is obtained.

7. The method for evaluating the spraying quality of a spraying robot based on laser point clouds according to claim 1, characterized in that: Step S6 specifically includes the following steps: S61: Based on the preset thickness acceptable threshold For each point, determine its thickness: Then, a specific color representing qualification is assigned; for thickness... For defective points, a color gradient mapping is applied based on the degree of thickness deviation to normalize the thickness value. Interval: in for The minimum thickness value in; S62: Normalization parameters are obtained through a predefined color mapping function. Converted to the corresponding RGB colors, the distribution and severity of insufficient thickness are clearly revealed; S63: Generate point clouds with color information The information is then displayed as a visual message and integrated into the human-machine interface of the painting robot to provide quality feedback and decision support for operators.

Images