Concrete strength automatic rebound detection method and device based on unmanned inspection vehicle

Abstract

The invention provides an automatic concrete strength rebound detection method and device based on an unmanned inspection vehicle, and relates to the technical field of automatic concrete strength rebound detection.The detection method comprises the steps that a navigation module carried by the unmanned inspection vehicle conducts automatic concrete strength rebound detection through a laser radar and vision fusion method in an autonomous navigation mode; moving to a detection operation area of a target concrete wall surface; a mechanical arm carried by the unmanned inspection vehicle adjusts the posture of the tail end of the mechanical arm according to the coordinates of the detection points and the normal vector of the fitting plane, so that a measuring head of a springback device carried at the tail end of the mechanical arm is in vertical contact with the wall surface; the springback device carries out a springback test according to preset parameters, obtains a springback value and transmits the springback value to the control processing module. According to the scheme, fusion analysis is carried out through the feature vector of the detection point wall image and the rebound curve feature to correct the influence of the surface state on the rebound value, abnormal data points are automatically screened out, and finally high-precision concrete strength calculation is achieved.

Description

Technical Field

[0001] This invention relates to the field of automatic rebound testing technology for concrete strength, and in particular to an automatic rebound testing method and device for concrete strength based on an unmanned inspection vehicle. Background Technology

[0002] Concrete strength testing is a crucial aspect of quality control in building construction. Traditional rebound hammer testing typically relies on manual operation, where personnel use a handheld rebound hammer to measure concrete strength point by point on the structural surface, indirectly calculating the concrete strength by reading the rebound values. However, this method suffers from low operational efficiency, poor data consistency, and significant human error. Especially in complex scenarios such as high-rise buildings, tunnels, and bridges, manual testing struggles to meet the requirements for long-term, comprehensive, and high-precision testing.

[0003] In recent years, the industry has attempted to combine rebound testing with automated equipment, using robotic arms or mobile robot platforms to achieve partially automated testing. However, existing systems generally suffer from insufficient detection and positioning accuracy, poor posture stability, and weak automated data processing capabilities. Furthermore, they cannot pre-identify flat areas on the wall surface and can only perform rebound testing in random areas. Uneven surfaces directly cause deviations in the contact angle of the rebound hammer, amplifying the error in the detection data.

[0004] For example, patent application CN119356406A, entitled "Collaborative Control System and Method for Building Concrete Strength Testing Robot", proposes a concrete strength testing device based on a mobile platform that can automatically complete rebound tests. However, this device is mainly designed for flat ground scenarios and is difficult to adapt to complex construction environments. At the same time, it cannot perform pre-analysis of the flatness of the test surface. The patent application with publication number CN223229405U, entitled "Automatic Rebound Detection Robot for Concrete Strength", proposes a system for automatic rebound detection using a robotic arm. Although it achieves partial automation, the accuracy of the robotic arm's end posture control is limited, resulting in poor detection performance on vertical or inclined surfaces. Furthermore, it is also unable to plan a reasonable detection area based on the actual flatness of the wall surface.

[0005] Furthermore, existing rebound detection systems still rely primarily on manual intervention in data fusion and automatic rebound curve fitting, making it impossible to achieve closed-loop control throughout the entire detection process. Summary of the Invention

[0006] The purpose of this invention is to provide an automatic rebound testing method and device for concrete strength based on an unmanned inspection vehicle, so as to solve at least one of the above-mentioned technical problems existing in the prior art.

[0007] Firstly, to solve the above-mentioned technical problems, the present invention provides an automatic rebound testing method for concrete strength based on an unmanned inspection vehicle, comprising the following steps: Step 1: The navigation module on the unmanned inspection vehicle uses a fusion method of lidar and vision to drive the unmanned inspection vehicle to move to the inspection area of ​​the target concrete wall in an autonomous navigation manner. Step 2: The laser scanning module on the unmanned inspection vehicle performs three-dimensional point cloud acquisition on the target concrete wall to obtain the overall point cloud data of the wall. Step 3: The control and processing module on the unmanned inspection vehicle preprocesses the overall point cloud data of the wall surface to obtain effective sample point cloud data, and then performs plane fitting to obtain the fitting plane; calculates the distance value from each effective sample point to the fitting plane; and selects areas with distance values ​​≤ preset flatness thresholds as effective detection areas and calculates the flatness. Step 4: Within the effective detection area, based on the principle of even grid division, (automatically) generate the coordinates of the detection points and the optimized detection path; Step 5: The (multi-degree-of-freedom) robotic arm mounted on the unmanned inspection vehicle adjusts the posture of the end of the robotic arm according to the coordinates of the detection point and the normal vector of the fitting plane, so that the probe of the rebound device mounted at the end of the robotic arm makes perpendicular contact with the wall. Step 6: The rebound device performs a rebound test according to preset parameters (automatic), obtains the rebound value and transmits it to the control processing module, and iteratively executes step 5 according to the detection optimization path until all detection points have completed the rebound test. Step 7: The control processing module performs strength conversion on the rebound value based on the coordinates of the detection point and the flatness to obtain the concrete strength; In this way, by using the above method, the target concrete wall surface can be automatically laser scanned and its flatness pre-analyzed. After screening out the effective detection area that meets the rebound detection conditions, the automatic rebound detection is carried out. This not only avoids the problems of low efficiency and large error of manual detection, but also avoids the deviation of the contact angle of the rebound device, and will not amplify the error of the detection data, thereby improving the detection efficiency and detection accuracy.

[0008] In one feasible implementation, the detection method further includes step 8, which involves curve fitting and visualization based on concrete strength for intuitive display.

[0009] In one feasible implementation, step 1, autonomous navigation includes: based on a lidar SLAM map, using the Adaptive Monte Carlo Localization (AMCL) method to achieve real-time positioning of the unmanned inspection vehicle; fusing IMU data to correct positioning drift, thereby ensuring a positioning error ≤ 3cm; using the Dynamic Window Method (DWA) to plan the optimal path from the starting point to a predetermined position in front of the target concrete wall, with path constraints including: avoiding obstacles in the map, minimizing path length, and minimizing the number of turns for the unmanned inspection vehicle; the lidar detects obstacles in the foreground in real time, and if an obstacle is detected, automatically plans a detour path (offset ≤ 0.5m, and not exceeding the detection area); after reaching the inspection work area, IMU data and lidar data are fused to correct the inspection vehicle's attitude, thereby ensuring a heading angle error ≤ 3cm. This is to ensure that the laser scanning direction is perpendicular to the wall.

[0010] In one feasible implementation, the specific method for acquiring three-dimensional point clouds in step 2 includes: an unmanned inspection vehicle moves horizontally back and forth along a direction parallel to the target concrete wall at a preset speed and a preset horizontal moving distance. After each reciprocating cycle, the pitch angle of the lidar is adjusted according to a preset angle step size until the entire target concrete wall is completely scanned.

[0011] In one feasible implementation, the specific method for preprocessing the overall point cloud data of the wall surface in step 3 includes: Step a1: Calculate the average and standard deviation of the distance between each point and all points in its preset neighborhood using a statistical filtering algorithm, and remove points whose distance is greater than 3 times the standard deviation. Step a2: Filter using a preset voxel grid to reduce the number of point clouds, retain key features, and improve the efficiency of subsequent calculations; Step a3: Obtain the extrinsic parameter matrix using the eye-to-hand calibration method and convert the local coordinate system (laser scanning module) into the world coordinate system (unmanned inspection vehicle).

[0012] In one feasible implementation, the hand-eye calibration method specifically includes: Step b1: Fix the checkerboard calibration plate to the end of the (multi-degree-of-freedom) robotic arm mounted on the unmanned inspection vehicle, and keep it parallel to the target concrete wall. Step b2: Control the robotic arm to move within the laser scanning range, and collect several sets of calibration board images and calibration point cloud data in different postures through a binocular camera and laser scanning module; Step b3: Extract the corner coordinates of the calibration board image (using the findChessboardCorners function of the OpenCV tool); extract the corner coordinates of the calibration point cloud data (using the projectPointOnPlane function of the PCL tool). Step b4: Construct calibration equations and solve for the rotation matrix R' and translation vector T' from the local coordinate system to the world coordinate system (calibration error ≤ 0.02 mm). Step b5: Based on R' and T', perform coordinate transformation (using the transformPointCloud function of the PCL tool). The specific expression includes: P_world = R' × P_local + T'; where P_world represents a point in the world coordinate system and P_local represents a point in the local coordinate system.

[0013] In one feasible implementation, the specific method for plane fitting in step 3 includes: Step c1: Randomly select 3 non-collinear points from the valid sample point data. , , Calculate the equation of the plane to obtain a plane. The specific formulas include: ; in, , , and All are plane coefficients, and we have: ; ; ; ; Step c2: Calculate the distance from all points to the plane. The specific formula is as follows: ;in, Indicates the index of a point; Based on a preset distance threshold ,Will The points are taken as interior points, and the number of interior points is counted; Step c3: Iterate through step c1, selecting the plane with the most interior points as the candidate plane, until the iteration termination condition is met; the iteration termination condition is reaching a preset number of iterations or the proportion of interior points in the valid sample points is ≥80%. Step c4: Construct the objective function for the set of interior points of the candidate plane. The specific calculation formula includes: ; in, All are planar parameters, and have ; Through the Find the partial derivative of the calculation formula and minimize it. The optimized plane parameters are obtained, and the final fitted plane is determined.

[0014] In one feasible implementation, step 3, the method for screening the effective detection area specifically includes: Step d1: According to the color mapping rules, perform color rendering on the effective sample point cloud data to distinguish the effective detection area (flat area) by color. Step d2: Convert the spatial regions corresponding to all points representing the colors of the valid detection regions into binary images; Step d3: In the binary image, fill the holes in the spatial region using rectangular structuring elements; Step d4: Extract connected regions (using the findContours function of OpenCV tool), calculate the actual area of ​​each connected region, and consider connected regions with actual area ≥ preset area threshold as valid connected regions, thereby avoiding invalid detection due to local small flat areas; Step d5: Merge multiple effective connected regions within a preset spacing range to form the final effective detection area, so as to reduce the moving distance of the robotic arm carried by the unmanned inspection vehicle.

[0015] In one feasible implementation, step 4 specifically includes: Step e1: Within the effective detection area, generate detection points using the grid method to meet the point spacing requirements for rebound testing in GB / T50204-2015 standard; Step e2: Starting from the first detection point on the left side of the first row within the effective detection area, number the points sequentially to the right until the last detection point of the first row is reached. Then, starting from the first detection point on the right side of the second row, number the points sequentially to the left until the last detection point of the second row is reached. This process is repeated until the last detection point of the last row is completed. In this way, an optimized detection path is automatically generated in a serpentine manner, reducing the moving distance of the robotic arm and improving detection efficiency.

[0016] In one feasible implementation, step e1 specifically includes: Step e11: Establish a local coordinate system (e.g., the horizontal axis is the X-axis and the vertical axis is the Y-axis) with the lower left corner of the effective detection area as the origin, and obtain the X-axis and Y-axis; Step e12: Within the effective detection area, select detection points along the X-axis and Y-axis according to the preset grid spacing. The coordinates of the detection points are represented as follows: , This represents the two-dimensional plane coordinate index of the detection point. Represents the Z-axis coordinate index of the detection point; calculate using the plane equation. The value is recorded in the detection point list, and the specific expression is: ; Step e13: If there are detection points that are less than the preset grid spacing from the edge of the effective detection area, then the detection points are shifted away from the edge of the effective area by a preset offset distance to avoid the detection points being too close to the edge of the effective detection area and affecting the detection effect of the rebound device.

[0017] In one feasible implementation, step 5, the specific method for adjusting the end effector posture of the robotic arm includes: Step f1: Based on the coordinates of the detection point and the normal vector of the fitting plane To determine the target posture of the robotic arm's end effector: the Z-axis of the robotic arm's end effector coordinate system is in the same direction as the normal vector of the fitting plane, so as to ensure that the probe of the rebound device is perpendicular to the wall surface; the X-axis of the robotic arm's end effector coordinate system is consistent with the forward direction of the unmanned inspection vehicle. Step f2: Based on the preset safe distance and the first movement speed, (using the MoveIt! motion planning library of the ROS system) plan the robotic arm end effector to reach the safe position of the detection point coordinates from the current position.

[0018] In one feasible implementation, the specific method of the rebound test in step 6 includes: the end of the robotic arm moves at a second speed to make the probe of the rebound device mounted on the end of the robotic arm reach the coordinates of the detection point, and monitors the pressure sensor signal of the rebound device in real time until there is a pressure signal, indicating that the probe of the rebound device has contacted the wall surface, and then applies a preset pressure until the tilt sensor of the rebound device detects that the contact angle error is ≤ the first preset angle threshold; the rebound device triggers the rebound test according to the preset number of rebounds, and takes the average value as the rebound value.

[0019] In one feasible implementation, step 6 of the rebound test method further includes an anomaly handling method: if the pressure sensor signal exceeds the pressure threshold or the fitting angle error exceeds the second preset angle threshold, the rebound test is determined to be a failure, the detection point is marked and the next detection point is skipped so that subsequent manual retesting can be performed.

[0020] In one feasible implementation, step 6, the specific method for the springback test, further includes an adaptive vector compensation method, specifically comprising: Step g1: The attitude of the robotic arm end is sensed in real time by using the attitude sensor installed at the end of the robotic arm. Step g2: Calculate the spatial angle difference between the pose and the normal vector of the fitted plane; Step g3: Compensate the spatial angle difference to the servo motor of the robotic arm to adjust the end effector angle of the robotic arm; iteratively execute step a2 until the spatial angle difference is less than or equal to the difference threshold. In this way, the direction of probe movement can always be consistent with the normal of the wall being tested, avoiding the contact angle error between the probe and the wall caused by uneven wall surfaces, reducing the rebound value deviation problem caused by posture error, and thus greatly improving the physical consistency and repeatability of the test results.

[0021] In one feasible implementation, step 6, the specific method for the rebound test, further includes an abnormal rebound value screening method, specifically including: Step h1: Use the camera mounted on the end of the robotic arm to collect real-time images of the wall at the detection point; Step h2: Extract feature vectors based on the wall image. The specific expressions include: ; in, The texture roughness feature is represented by the following expressions: ;in, This represents the total number of pixels in the vertical and horizontal directions of the wall image; Indicates the pixel position of the wall image The grayscale value at that location; This represents a two-dimensional discrete Laplacian operator used to quantify local second-order variations in image intensity, reflecting surface micro-irregularities. The crack index characteristic is represented by the following specific expressions: ;in, This represents the region of interest defined in the wall image; express Total number of pixels within; Indicated in pixels The binary output of the Canny edge detection operator (i.e., if the pixel is an edge, the value is 1; if the pixel is not an edge, the value is 0). This is represented (by conventional connected component analysis). The pixel length of the longest continuous crack within the crack; This represents the average pixel width of all (detected) crack lines; It represents a very small positive number, used to prevent the denominator from being 0; The expression representing the degree of color change includes: ;in, This represents the color histogram vector of the current wall image; The base color histogram vector representing the wall image; The L2 norm of a vector is represented, which is the Euclidean distance. The characteristics representing color uniformity are specifically expressed as follows: ;in, They represent The variance of pixel values ​​in the three color channels: red, green, and blue. Step h3: Define the comprehensive wall surface influencing factor This is used to quantify the systematic shift in rebound value caused by wall defects. Specific expressions include: ; in, Indicates the reference offset; These are all weighting coefficients, which can be calibrated through conventional paired experiments; To represent local flatness, the specific expressions include: ;in, Represents a valid sample point cloud dataset The Middle A local point cloud dataset The number of points in the middle; express The Middle One point; express The set of neighborhood points; This indicates the calculation of the covariance matrix; This represents the smallest eigenvalue of the computed matrix. If this eigenvalue is close to 0, it indicates that the local region is close to a plane. Step h4, based on The rebound value is corrected using the following formulas: ; in, Indicates the first The original rebound value of each detection point; express Correction value; This represents the correction strength coefficient, used to control the maximum correction range of the impact of wall defects on the rebound value; The threshold representing the baseline wall state can be taken from the training dataset. The median; This represents the normalization parameter, used for adjustment. The sensitivity of the function can be taken from the training dataset. Standard deviation; The function, namely the hyperbolic tangent function, is used to smoothly limit the correction amount to a certain value. Within the range; Step h5: Define the anomaly value segmentation The calculation formula, specifically the expression, includes: ; in, Each represents a weight coefficient for each outlier, and satisfies the following conditions: ; Indicates the first The statistical dispersion of the rebound values ​​at each detection point is expressed as follows: ;in, These represent the sample mean and sample standard deviation of all raw rebound values, respectively. Indicates the first The Mahalanobis dispersion of the feature vector of each detection point is expressed as follows: ; in, This represents the sample mean vector of all feature vectors; The sample covariance matrix representing all eigenvectors; Indicates the first The point cloud feature dispersion of each detection point is specifically expressed as follows: ; in, These represent the sample mean and sample standard deviation of all local smoothness, respectively; Step h6, if If the value is greater than the anomaly detection threshold, then the first one is determined. If the rebound value of a test point is an abnormal rebound value, it will be screened out.

[0022] In one feasible implementation, step 6, the specific method for the springback test, further includes a stiffness compensation method, specifically comprising: Step i1: Construct an empirical mapping between rebound value and transferred energy to quantify the rebound value error caused by arm-body compliance and joint flutter. The specific expression includes: ; in, Indicates the rebound value; Represents a mapping function; This represents the energy actually transferred to the concrete wall surface. The specific energy conservation model formula is as follows: ,in, This represents the impact energy of the probe in an ideal rebound device. This indicates that the robotic arm absorbs energy in a compliant manner; the specific calculation formula is as follows: ;in, Represents the impact force function. Represents the unit vector along the direction of the rebound impact. express The transpose of ; Represents the joint angle vector function of the robotic arm; The compliance matrix of the robotic arm's end effector is represented by the following formula: ,in, Let represent the Cartesian stiffness matrix of the robotic arm's end effector, and have . ,in, This represents the joint stiffness matrix (torque / angular deformation). Represents the Jacobian matrix of the robotic arm; Step i2: Construct the joint stiffness amplification factor This is used to quantify the degree of stiffness improvement of different (models) of robotic arms; Introducing matrices, specific expressions include: ; ; in, This represents the improved joint stiffness matrix; This represents the improved compliance matrix; In this way, the position / angle of the robotic arm end is precisely locked through high rigidity control, avoiding energy loss and position deviation caused by the deformation of the arm body when the rebound device strikes, thus achieving rigid locking and impedance gain enhancement. Step i3: Based on the displacement caused by the impact force, construct an approximate formula for improving the robot arm's compliance with energy absorption: In order to optimize energy distribution; among which, This indicates that the improved robotic arm is more compliant with energy absorption; therefore, the overall stiffness of the robotic arm is increased. This can reduce the impact energy loss during rebound testing by several times. This significantly reduces rebound test errors; Step i4: Construct the joint angle perturbation covariance matrix The impact propagates linearly to the end of the robotic arm, and the standard deviation along the rebound impact direction is calculated. In order to quantify the joint perturbation error of the robotic arm, the specific expressions include: ; Step i5: Perform tests on the same detection point. After taking the average of several independent measurements, we have: This significantly reduces disturbance error; Step i6: Measure the force and displacement at the end of the robotic arm in real time using a (high-bandwidth) force / displacement sensor mounted on the robotic arm, and based on... Real-time estimation and deduction of the robotic arm's The rebound value of a single measurement is corrected, and the specific expression includes: ; in, This represents the corrected variance of the rebound value; Indicates the variance of the rebound value; express The derivative; express The estimated variance.

[0023] In one feasible implementation, the specific calculation formula for strength conversion in step 7 includes: ; ; in, Indicates the first Concrete strength at each testing point; Indicates the first The rebound value at each detection point; This indicates the correction value for concrete strength. This represents the flatness correction factor, and the specific calculation formula is as follows: ;in, Indicates the first The flatness of each test point (i.e.) In this way, the standard GB / T50204-2015 can be met, and the concrete strength can be effectively corrected according to the flatness, thereby improving the accuracy of the concrete strength.

[0024] In one feasible implementation, step 8, the visualization process includes: generating a flatness-concrete strength overlay color map; generating an overall concrete strength heat map based on the concrete strength values ​​of the detection points using the Kriging interpolation method; and storing the overall point cloud data of the wall surface, the coordinates of the detection points, the rebound value, the concrete strength correction value, and the flatness as a CSV file and uploading it to the BIM platform.

[0025] Secondly, based on the same inventive concept, this application also provides an automatic rebound testing device for concrete strength based on an unmanned inspection vehicle, including an unmanned inspection vehicle equipped with a navigation module, a laser scanning module, a control processing module, and a robotic arm; a rebound device is provided at the end of the robotic arm; the control processing module is electrically connected to the navigation module, the laser scanning module, the robotic arm, and the rebound device.

[0026] By adopting the above technical solution, the present invention has the following beneficial effects: This invention provides an automatic rebound detection method and device for concrete strength based on an unmanned inspection vehicle. By introducing a high-precision navigation module that integrates lidar and vision, the unmanned inspection vehicle achieves centimeter-level autonomous positioning and dynamic path planning in complex environments. Compared with existing solutions that rely solely on GPS or single laser ranging navigation, this solution can perform three-dimensional environmental modeling and detection area identification of the structural surface. It can still operate stably even in areas without satellite signals, such as construction sites, tunnels, and bridge webs, fundamentally improving the positioning reliability and environmental adaptability of the unmanned inspection vehicle.

[0027] This application utilizes attitude sensor feedback and motor fine-tuning control to ensure that the impact direction of the rebound device always aligns with the normal direction of the concrete wall being tested. Compared to previous methods involving fixed angles or manual adjustments, this application can automatically adapt to vertical, inclined, and curved surfaces, significantly reducing rebound value deviations caused by attitude errors and improving detection accuracy and repeatability.

[0028] This application establishes a joint angle perturbation covariance matrix, a robotic arm Jacobian matrix, and a robotic arm end-effector compliance matrix to estimate the propagation of joint flutter error to end-effector displacement error during impact. Force / displacement sensors are used to measure the force and displacement at the robotic arm end-effector in real time. An energy conservation model is used to estimate and compensate for the impact energy absorbed by the robotic arm structure online, making the corrected rebound value approach the measurement results from the rigid base. A short-term rigidity locking and impedance gain enhancement mechanism is triggered at the moment of impact to improve the transient stiffness of the robotic arm joints and reduce energy loss and flutter propagation. Random errors caused by joint flutter are compensated through multiple measurements, statistical analysis, and modeling, making the random fluctuations more consistent with the number of measurements. By attenuating and avoiding the accumulation of systematic bias, this solution can achieve rebound measurement accuracy and repeatability close to that of a fixed rigid support in a mobile platform environment, thus solving the problem of rebound deviation introduced by the robotic arm.

[0029] This application fuses and analyzes feature vectors (such as texture, cracks, and degree of discoloration due to pollution) from wall images at detection points with rebound curve features to correct the influence of surface condition on rebound values, automatically filter out abnormal data points, dynamically correct coefficients, and ultimately achieve high-precision concrete strength calculation. Compared to existing schemes that rely mainly on manual readings or static curve fitting, this application significantly improves the intelligence, adaptability, and anti-interference capability of data fitting. Attached Figure Description

[0030] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0031] Figure 1 A flowchart of an automatic rebound testing method for concrete strength based on an unmanned inspection vehicle is provided in this embodiment of the invention. Figure 2 Flowchart of another automatic rebound testing method for concrete strength based on an unmanned inspection vehicle provided in this embodiment of the invention; Figure 3 A three-dimensional structural diagram of an automatic rebound testing device for concrete strength based on an unmanned inspection vehicle, provided in an embodiment of the present invention; Figure 4 for Figure 3 Another viewpoint; Figure label: 1-Unmanned inspection vehicle; 2-robotic arm; 21-robotic arm end effector; 3-rebound device. Detailed Implementation

[0032] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0034] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0035] The present invention will be further explained below with reference to specific embodiments.

[0036] Example 1: like Figure 1 As shown in the figure, this embodiment provides an automatic rebound testing method for concrete strength based on an unmanned inspection vehicle, which includes the following steps: Step 1: The navigation module on the unmanned inspection vehicle moves to the inspection area of ​​the target concrete wall through autonomous navigation by using a fusion method of lidar and vision. Step 2: The laser scanning module on the unmanned inspection vehicle performs three-dimensional point cloud acquisition on the target concrete wall to obtain the overall point cloud data of the wall. Step 3: The control and processing module on the unmanned inspection vehicle preprocesses the overall point cloud data of the wall surface to obtain effective sample point cloud data, and then performs plane fitting to obtain the fitting plane; calculates the distance value from each effective sample point to the fitting plane; and selects areas with distance values ​​≤ preset flatness thresholds as effective detection areas and calculates the flatness. Step 4: Within the effective detection area, based on the principle of even grid division, (automatically) generate the coordinates of the detection points and the optimized detection path; Step 5: The (multi-degree-of-freedom) robotic arm mounted on the unmanned inspection vehicle adjusts the posture of the end of the robotic arm according to the coordinates of the detection point and the normal vector of the fitting plane, so that the probe of the rebound device mounted at the end of the robotic arm makes perpendicular contact with the wall. Step 6: The rebound device performs a rebound test according to preset parameters (automatic), obtains the rebound value and transmits it to the control processing module, and iteratively executes step 5 according to the detection optimization path until all detection points have completed the rebound test. Step 7: The control processing module performs strength conversion on the rebound value based on the coordinates of the detection point and the flatness to obtain the concrete strength; In this way, by using the above method, the target concrete wall surface can be automatically laser scanned and its flatness pre-analyzed. After screening out the effective detection area that meets the rebound detection conditions, the automatic rebound detection is carried out. This not only avoids the problems of low efficiency and large error of manual detection, but also avoids the deviation of the contact angle of the rebound device, and will not amplify the error of the detection data, thereby improving the detection efficiency and detection accuracy.

[0037] Furthermore, such as Figure 2 As shown, the detection method also includes step 8, which involves curve fitting and visualization based on concrete strength for intuitive display.

[0038] Further, in step 1, autonomous navigation includes: based on a lidar SLAM map, using the Adaptive Monte Carlo Localization (AMCL) method to achieve real-time positioning of the unmanned inspection vehicle; fusing IMU data to correct positioning drift, thereby ensuring a positioning error ≤ 3cm; using the Dynamic Window Method (DWA), planning the optimal path from the starting point to a predetermined position (e.g., 5m) in front of the target concrete wall, with path constraints including: avoiding obstacles in the map, minimizing path length, and minimizing the number of turns for the unmanned inspection vehicle; the lidar detects obstacles (e.g., objects with three-dimensional dimensions ≥ 0.3m × 0.3m × 0.3m) within a real-time range (e.g., 1m) in front, and if an obstacle is detected, automatically planning a detour path (offset ≤ 0.5m, not exceeding the detection area); after reaching the inspection work area, fusing IMU and lidar data to correct the inspection vehicle's attitude, thereby ensuring a heading angle error ≤ 0.5m. This is to ensure that the laser scanning direction is perpendicular to the wall.

[0039] Furthermore, in step 2, the specific method for acquiring three-dimensional point clouds includes: the unmanned inspection vehicle moves horizontally back and forth along a direction parallel to the target concrete wall at a preset speed (e.g., 0.05 m / s) and a preset horizontal moving distance. After each reciprocating cycle, the pitch angle of the lidar is adjusted according to a preset angle step size until the entire target concrete wall is completely scanned.

[0040] Furthermore, in step 3, the specific method for preprocessing the overall point cloud data of the wall surface includes: Step a1: Calculate the average and standard deviation of the distance between each point and all points in its preset neighborhood using a (conventional) statistical filtering algorithm, and remove points whose distance is greater than 3 times the standard deviation. Step a2: Filter the data using a preset voxel grid (e.g., 0.005m × 0.005m × 0.005m) to reduce the number of point clouds, retain key features, and improve the efficiency of subsequent calculations. Step a3: Obtain the extrinsic parameter matrix using the eye-to-hand calibration method and convert the local coordinate system (laser scanning module) into the world coordinate system (unmanned inspection vehicle).

[0041] Furthermore, the hand-eye calibration method specifically includes: Step b1: Fix the checkerboard calibration plate (e.g., a calibration plate with a size of 0.5m×0.5m, a corner spacing of 5cm, and a number of corner points of 9×7) to the end of the (multi-degree-of-freedom) robotic arm mounted on the unmanned inspection vehicle, and keep it parallel to the target concrete wall. Step b2: Control the robotic arm to move within the laser scanning range, and acquire several sets (e.g., 10 sets) of calibration board images and calibration point cloud data in different postures through a binocular camera and laser scanning module; Step b3: Extract the corner coordinates of the calibration board image (using the findChessboardCorners function of the OpenCV tool); extract the corner coordinates of the calibration point cloud data (using the projectPointOnPlane function of the PCL tool). Step b4: Construct calibration equations and solve for the rotation matrix R' and translation vector T' from the local coordinate system to the world coordinate system (calibration error ≤ 0.02 mm). Step b5: Based on R' and T', perform coordinate transformation (using the transformPointCloud function of the PCL tool). The specific expression includes: P_world = R' × P_local + T'; where P_world represents a point in the world coordinate system and P_local represents a point in the local coordinate system.

[0042] Furthermore, the specific method for plane fitting in step 3 includes: Step c1: Randomly select 3 non-collinear points from the valid sample point data. , , Calculate the equation of the plane to obtain a plane. The specific formulas include: ; in, , , and All are plane coefficients, and we have: ; ; ; ; Step c2: Calculate the distance from all points to the plane. The specific formula is as follows: ;in, Indicates the index of a point; Based on a preset distance threshold (e.g., 0.5mm), will The points are taken as interior points, and the number of interior points is counted; Step c3: Iterate through step c1, selecting the plane with the most interior points as the candidate plane, until the iteration termination condition is met; the iteration termination condition is reaching a preset number of iterations (e.g., 50 times) or the proportion of interior points in the valid sample points is ≥80%. Step c4: Construct the objective function for the set of interior points of the candidate plane. The specific calculation formula includes: ; in, All are planar parameters, and have ; Through the Find the partial derivative of the calculation formula and minimize it. The optimized plane parameters are obtained, and the final fitted plane is determined.

[0043] Furthermore, in step 3, the method for screening the effective detection area specifically includes: Step d1: According to the color mapping rules, the effective sample point cloud data is color-rendered to distinguish the effective detection area (flat area) by color; the color mapping rules are shown in Table 1. Table 1

[0044] Step d2: Convert the spatial regions corresponding to all points representing the colors of the effective detection areas into binary images (e.g., a resolution of 0.001m / pixel, corresponding to a 2.5m × 5m wall area). Step d3: In the binary image, fill the holes in the spatial region using rectangular structuring elements (e.g., 100*100 pixels); Step d4: Using the `findContours` function of OpenCV, extract connected components and calculate the actual area of ​​each connected component (e.g., pixel area × 0.001 × 0.001). Ensure the actual area is greater than or equal to a preset area threshold (e.g., ...). A connected region is considered a valid connected region, thus avoiding invalid detection due to local small flat areas. Step d5: Merge multiple effective connected regions within a preset spacing range (e.g., ≤5cm) to form the final effective detection area, so as to reduce the moving distance of the robotic arm carried by the unmanned inspection vehicle.

[0045] Furthermore, step 4 specifically includes: Step e1: Within the effective detection area, generate detection points using a grid method (e.g., a 500mm*500mm grid method) to meet the point spacing requirements for rebound testing in the GB / T50204-2015 standard. Step e2: Starting from the first detection point on the left side of the first row within the effective detection area, number the points sequentially to the right until the last detection point of the first row is reached. Then, starting from the first detection point on the right side of the second row, number the points sequentially to the left until the last detection point of the second row is reached. This process is repeated until the last detection point of the last row is completed. In this way, an optimized detection path is automatically generated in a serpentine manner, reducing the moving distance of the robotic arm and improving detection efficiency.

[0046] Further, step e1 specifically includes: Step e11: Establish a local coordinate system (e.g., the horizontal axis is the X-axis and the vertical axis is the Y-axis) with the lower left corner of the effective detection area as the origin, and obtain the X-axis and Y-axis; Step e12: Within the effective detection area, select detection points along the X-axis and Y-axis according to the preset grid spacing (e.g., 500mm). The coordinates of the detection points are represented as follows: , This represents the two-dimensional plane coordinate index of the detection point. Represents the Z-axis coordinate index of the detection point; calculate using the plane equation. The value is recorded in the detection point list, and the specific expression is: ; Step e13: If there are detection points that are less than the preset grid spacing from the edge of the effective detection area, the detection points are shifted away from the edge of the effective area by a preset offset distance (e.g., 25mm) to avoid the detection points being too close to the edge of the effective detection area and affecting the detection effect of the rebound device.

[0047] Furthermore, in step 5, the specific method for adjusting the posture of the robotic arm's end effector includes: Step f1: Based on the coordinates of the detection point and the normal vector of the fitting plane To determine the target posture of the robotic arm's end effector: the Z-axis of the robotic arm's end effector coordinate system is in the same direction as the normal vector of the fitting plane, so as to ensure that the probe of the rebound device is perpendicular to the wall surface; the X-axis of the robotic arm's end effector coordinate system is consistent with the forward direction of the unmanned inspection vehicle. Step f2: Based on the preset safe distance (e.g., the distance between the robotic arm and the wall is ≥5cm) and the first movement speed (e.g., 0.1m / s), (using the MoveIt! motion planning library of the ROS system) plan the robotic arm end to reach the safe position of the detection point coordinates from the current position.

[0048] Further, the specific method for the rebound test in step 6 includes: the end effector of the robotic arm moves at a second speed to bring the probe of the rebound device mounted on the end effector to the detection point coordinates, and the pressure sensor signal of the rebound device is monitored in real time until a pressure signal is detected, indicating that the probe of the rebound device has contacted the wall surface. Then, a preset pressure (e.g., 2.2 ± 0.2 N) is applied until the tilt sensor of the rebound device detects that the contact angle error is ≤ the first preset angle threshold (e.g., ...). The rebound device triggers a rebound test according to a preset number of rebounds (e.g., 3 times), and takes the average value as the rebound value.

[0049] Furthermore, in step 6, the specific method for the rebound test also includes an anomaly handling method: if the pressure sensor signal exceeds the pressure threshold (e.g., 3N) or the contact angle error exceeds the second preset angle threshold (e.g., ... If the test fails, the rebound test is considered a failure. The test point is marked and the test is skipped to the next test point so that a manual retest can be performed later.

[0050] Furthermore, in step 6, the specific method for the springback test also includes an adaptive vector compensation method, specifically comprising: Step g1: The attitude of the robotic arm end is sensed in real time by using the attitude sensor installed at the end of the robotic arm. Step g2: Calculate the spatial angle difference between the pose and the normal vector of the fitted plane; Step g3: Compensate the spatial angle difference to the servo motor of the robotic arm to adjust the end effector angle of the robotic arm; iteratively execute step a2 until the spatial angle difference is less than or equal to the difference threshold. In this way, the direction of probe movement can always be consistent with the normal of the wall being tested, avoiding the contact angle error between the probe and the wall caused by uneven wall surfaces, reducing the rebound value deviation problem caused by posture error, and thus greatly improving the physical consistency and repeatability of the test results.

[0051] Furthermore, in step 6, the specific method for the rebound test also includes an abnormal rebound value screening method, specifically including: Step h1: Use the camera mounted on the end of the robotic arm to collect real-time images of the wall at the detection point; Step h2: Extract feature vectors based on the wall image. The specific expressions include: ; in, The texture roughness feature is represented by the following expressions: ;in, This represents the total number of pixels in the vertical and horizontal directions of the wall image; Indicates the pixel position of the wall image The grayscale value at that location; This represents a two-dimensional discrete Laplacian operator used to quantify local second-order variations in image intensity, reflecting surface micro-irregularities. The crack index characteristic is represented by the following specific expressions: ;in, This represents the region of interest defined in the wall image; express Total number of pixels within; Indicated in pixels The binary output of the Canny edge detection operator (i.e., if the pixel is an edge, the value is 1; if the pixel is not an edge, the value is 0). This is represented (by conventional connected component analysis). The pixel length of the longest continuous crack within the crack; This represents the average pixel width of all (detected) crack lines; It represents a very small positive number, used to prevent the denominator from being 0; The expression representing the degree of color change includes: ;in, This represents the color histogram vector of the current wall image; The base color histogram vector representing the wall image; The L2 norm of a vector is represented, which is the Euclidean distance. The characteristics representing color uniformity are specifically expressed as follows: ;in, They represent The variance of pixel values ​​in the three color channels: red, green, and blue. Step h3: Define the comprehensive wall surface influencing factor This is used to quantify the systematic shift in rebound value caused by wall defects. Specific expressions include: ; in, Indicates the reference offset; These are all weighting coefficients, which can be calibrated through conventional paired experiments; To represent local flatness, the specific expressions include: ;in, Represents a valid sample point cloud dataset The Middle A local point cloud dataset The number of points in the middle; express The Middle One point; express The set of neighborhood points; This indicates the calculation of the covariance matrix; This represents the smallest eigenvalue of the computed matrix. If this eigenvalue is close to 0, it indicates that the local region is close to a plane. Step h4, based on The rebound value is corrected using the following formulas: ; in, Indicates the first The original rebound value of each detection point; express Correction value; This represents the correction strength coefficient, used to control the maximum correction range of the impact of wall defects on the rebound value; The threshold representing the baseline wall state can be taken from the training dataset. The median; This represents the normalization parameter, used for adjustment. The sensitivity of the function can be taken from the training dataset. Standard deviation; The function, namely the hyperbolic tangent function, is used to smoothly limit the correction amount to a certain value. Within the range; Step h5: Define the anomaly value segmentation The calculation formula, specifically the expression, includes: ; in, Each represents a weight coefficient for each outlier, and satisfies the following conditions: ; Indicates the first The statistical dispersion of the rebound values ​​at each detection point is expressed as follows: ;in, These represent the sample mean and sample standard deviation of all raw rebound values, respectively. Indicates the first The Mahalanobis dispersion of the feature vector of each detection point is expressed as follows: ; in, This represents the sample mean vector of all feature vectors; The sample covariance matrix representing all eigenvectors; Indicates the first The point cloud feature dispersion of each detection point is specifically expressed as follows: ; in, These represent the sample mean and sample standard deviation of all local smoothness, respectively; Step h6, if If the value is greater than the anomaly detection threshold, then the first one is determined. If the rebound value of a test point is an abnormal rebound value, it will be screened out.

[0052] Furthermore, in step 6, the specific method for the springback test also includes a stiffness compensation method, specifically including: Step i1: Construct an empirical mapping between rebound value and transferred energy to quantify the rebound value error caused by arm-body compliance and joint flutter. The specific expression includes: ; in, Indicates the rebound value; Represents a mapping function; This represents the energy actually transferred to the concrete wall surface. The specific energy conservation model formula is as follows: ,in, This represents the impact energy of the probe in an ideal rebound device. This indicates that the robotic arm absorbs energy in a compliant manner; the specific calculation formula is as follows: ;in, Represents the impact force function. Represents the unit vector along the direction of the rebound impact. express The transpose of ; Represents the joint angle vector function of the robotic arm; The compliance matrix of the robotic arm's end effector is represented by the following formula: ,in, Let represent the Cartesian stiffness matrix of the robotic arm's end effector, and have . ,in, This represents the joint stiffness matrix (torque / angular deformation). Represents the Jacobian matrix of the robotic arm; Step i2: Construct the joint stiffness amplification factor This is used to quantify the degree of stiffness improvement of different (models) of robotic arms; Introducing matrices, specific expressions include: ; ; in, This represents the improved joint stiffness matrix; This represents the improved compliance matrix; In this way, the position / angle of the robotic arm end is precisely locked through high rigidity control, avoiding energy loss and position deviation caused by the deformation of the arm body when the rebound device strikes, thus achieving rigid locking and impedance gain enhancement. Step i3: Based on the displacement caused by the impact force, construct an approximate formula for improving the robot arm's compliance with energy absorption: In order to optimize energy distribution; among which, This indicates that the improved robotic arm is more compliant with energy absorption; therefore, the overall stiffness of the robotic arm is increased. This can reduce the impact energy loss during rebound testing by several times. This significantly reduces rebound test errors; Step i4: Construct the joint angle perturbation covariance matrix The impact propagates linearly to the end of the robotic arm, and the standard deviation along the rebound impact direction is calculated. In order to quantify the joint perturbation error of the robotic arm, the specific expressions include: ; Step i5: Perform tests on the same detection point. After taking the average of several independent measurements, we have: This significantly reduces disturbance error; Step i6: Measure the force and displacement at the end of the robotic arm in real time using a (high-bandwidth) force / displacement sensor mounted on the robotic arm, and based on... Real-time estimation and deduction of the robotic arm's The rebound value of a single measurement is corrected, and the specific expression includes: ; in, This represents the corrected variance of the rebound value; Indicates the variance of the rebound value; express The derivative; express The estimated variance.

[0053] Furthermore, in step 7, the specific calculation formula for the strength conversion includes: ; ; in, Indicates the first Concrete strength at each testing point; Indicates the first The rebound value at each detection point; This indicates the correction value for concrete strength. This represents the flatness correction factor, and the specific calculation formula is as follows: ;in, Indicates the first The flatness of each test point (i.e.) In this way, the standard GB / T50204-2015 can be met, and the concrete strength can be effectively corrected according to the flatness, thereby improving the accuracy of the concrete strength.

[0054] Furthermore, in step 8, the visualization process includes: generating a flatness-concrete strength overlay color map (e.g., gray areas indicate the test points and their corresponding concrete strength values, and red and blue areas indicate the flatness levels); generating an overall concrete strength heat map based on the concrete strength values ​​of the test points using Kriging interpolation; and storing the overall wall point cloud data, test point coordinates, rebound values, concrete strength correction values, and flatness (data) as a CSV file and uploading it to the BIM platform (supporting export in IFC format).

[0055] Example 2: like Figure 3-4 As shown, this embodiment provides an automatic rebound testing device for concrete strength based on an unmanned inspection vehicle, including an unmanned inspection vehicle 1. The unmanned inspection vehicle 1 is equipped with a navigation module, a laser scanning module, a control processing module, and a robotic arm 2. A rebound device 3 is provided at the end 21 of the robotic arm. The control processing module is electrically connected to the navigation module, the laser scanning module, the robotic arm 2, and the rebound device 3.

[0056] Furthermore, the specific parameters of the laser scanning module include: Scan frequency: 100Hz; Point cloud resolution: 0.05mm; Working distance: 0.5-50m; Measurement accuracy: ±0.1mm; The laser scanning module is fixed to the top front end of the unmanned inspection vehicle 1, and the scanning direction is perpendicular to the forward direction of the unmanned inspection vehicle 1 (towards the concrete wall). The installation height can be 1.2m. The laser scanning module is connected to the control and processing module via Ethernet, so that the transmission rate reaches 1Gbps and the delay time is ≤10ms.

[0057] Furthermore, the robotic arm 2 has six degrees of freedom, specifically the UR5e model, with a load capacity ≥ 5 kg and a repeatability accuracy of [missing information]. The robotic arm 2 is installed at the top middle and rear end of the unmanned inspection vehicle 1, with a distance of more than 300mm between it and the laser scanning module to avoid scanning obstruction.

[0058] Furthermore, the control processing module includes a processor (NVIDIA Jetson AGXXavier, 8-core ARM Cortex-A78, 32GB memory), an SSD storage (1TB), an Ethernet interface (4 channels, connecting to the laser scanning module and the navigation module), an RS485 interface (6 channels, connecting to the robotic arm 2 and the rebound device 3), and an HDMI interface (2 channels, connecting to the display screen).

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for automatic rebound testing of concrete strength based on an unmanned inspection vehicle, characterized in that, include: Step 1: The navigation module on the unmanned inspection vehicle moves to the inspection area of ​​the target concrete wall through autonomous navigation by using a fusion method of lidar and vision. Step 2: The laser scanning module on the unmanned inspection vehicle performs three-dimensional point cloud acquisition on the target concrete wall to obtain the overall point cloud data of the wall. Step 3: The control and processing module on the unmanned inspection vehicle preprocesses the overall point cloud data of the wall surface to obtain effective sample point cloud data, and then performs plane fitting to obtain the fitting plane; calculates the distance value from each effective sample point to the fitting plane; and selects areas with distance values ​​≤ preset flatness thresholds as effective detection areas and calculates the flatness. Step 4: Within the effective detection area, generate the coordinates of the detection points and the optimized detection path based on the principle of even grid division; Step 5: The robotic arm mounted on the unmanned inspection vehicle adjusts the posture of the end of the robotic arm according to the coordinates of the detection point and the normal vector of the fitting plane, so that the probe of the rebound device mounted on the end of the robotic arm makes perpendicular contact with the wall. Step 6: The rebound device performs a rebound test according to preset parameters, obtains the rebound value and transmits it to the control processing module, and iteratively executes step 5 according to the detection optimization path until all detection points have completed the rebound test. Step 7: The control processing module calculates the strength of the rebound value based on the coordinates of the detection point and the flatness to obtain the concrete strength.

2. The detection method according to claim 1, characterized in that, It also includes step 8, which involves curve fitting and visualization based on concrete strength.

3. The detection method according to claim 1, characterized in that, In step 1, autonomous navigation includes: Based on LiDAR SLAM maps, an adaptive Monte Carlo positioning method is used to achieve real-time positioning of unmanned inspection vehicles, and IMU data is fused to correct positioning drift. Using the dynamic window method, the optimal path is planned from the starting point to a predetermined position in front of the target concrete wall. The path constraints include: avoiding obstacles in the map, minimizing the path length, and minimizing the number of turns for the unmanned inspection vehicle. The lidar detects obstacles in the area in front in real time. If an obstacle is detected, it automatically plans a detour path. After reaching the inspection work area, the IMU data and lidar data are fused to correct the attitude of the inspection vehicle.

4. The detection method according to claim 1, characterized in that, In step 3, the specific method for preprocessing the overall point cloud data of the wall surface includes: Step a1: Calculate the average and standard deviation of the distance between each point and all points in its preset neighborhood using a statistical filtering algorithm, and remove points whose distance is greater than 3 times the standard deviation. Step a2: Filter using a preset voxel grid; Step a3: Obtain the extrinsic parameter matrix using the hand-eye calibration method, and convert the local coordinate system into the world coordinate system.

5. The detection method according to claim 4, characterized in that, The hand-eye calibration method specifically includes: Step b1: Fix the chessboard calibration plate to the end of the robotic arm mounted on the unmanned inspection vehicle, and keep it parallel to the target concrete wall. Step b2: Control the robotic arm to move within the laser scanning range, and collect several sets of calibration board images and calibration point cloud data in different postures through a binocular camera and laser scanning module; Step b3: Extract the corner coordinates of the calibration board image; extract the corner coordinates of the calibration point cloud data; Step b4: Construct calibration equations and solve for the rotation matrix R' and translation vector T' from the local coordinate system to the world coordinate system; Step b5: Based on R' and T', perform coordinate transformation. The specific expression includes: P_world = R' × P_local + T'; where P_world represents a point in the world coordinate system and P_local represents a point in the local coordinate system.

6. The detection method according to claim 1, characterized in that, The specific methods for plane fitting in step 3 include: Step c1: Randomly select 3 non-collinear points from the valid sample point data. , , Calculate the equation of the plane to obtain a plane. The specific formulas include: ； in, , , and All are plane coefficients, and we have: ； ； ； ； Step c2: Calculate the distance from all points to the plane. The specific formula is as follows: ;in, Indicates the index of a point; Based on a preset distance threshold ,Will The points are taken as interior points, and the number of interior points is counted; Step c3: Iterate through step c1 and select the plane with the most interior points as the candidate plane until the iteration ends. Step c4: Construct the objective function for the set of interior points of the candidate plane. The specific calculation formula includes: ； in, All are planar parameters, and have ; Through the Find the partial derivative of the calculation formula and minimize it. The optimized plane parameters are obtained, and the final fitted plane is determined.

7. The detection method according to claim 1, characterized in that, In step 3, the method for screening the effective detection area specifically includes: Step d1: According to the color mapping rules, perform color rendering on the effective sample point cloud data to distinguish the effective detection area by color; Step d2: Convert the spatial regions corresponding to all points representing the colors of the valid detection regions into binary images; Step d3: In the binary image, fill the holes in the spatial region using rectangular structuring elements; Step d4: Extract connected components, calculate the actual area of ​​each connected component, and consider connected components with an actual area ≥ a preset area threshold as valid connected components; Step d5: Merge multiple valid connected regions within the preset spacing range to form the final valid detection region.

8. The detection method according to claim 1, characterized in that, In step 5, the specific methods for adjusting the end effector posture of the robotic arm include: Step f1: Based on the coordinates of the detection point and the normal vector of the fitting plane To determine the target posture of the robotic arm's end effector: the Z-axis of the robotic arm's end effector coordinate system is in the same direction as the normal vector of the fitted plane; the X-axis of the robotic arm's end effector coordinate system is in the same direction as the forward direction of the unmanned inspection vehicle. Step f2: Based on the preset safe distance and the first movement speed, plan the safe position for the end effector of the robotic arm to reach the detection point coordinates from the current position.

9. The detection method according to claim 1, characterized in that, The specific method for the rebound test in step 6 includes: the end of the robotic arm moves at a second speed to make the probe of the rebound device mounted on the end of the robotic arm reach the coordinates of the detection point, and monitors the pressure sensor signal of the rebound device in real time. When there is a pressure signal, a preset pressure is applied until the tilt sensor of the rebound device detects that the contact angle error is ≤ the first preset angle threshold. The rebound device triggers the rebound test according to the preset number of rebounds and takes the average value as the rebound value.

10. An automatic rebound testing device for concrete strength based on an unmanned inspection vehicle, employing the testing method described in any one of claims 1-9, characterized in that, The system includes an unmanned inspection vehicle equipped with a navigation module, a laser scanning module, a control and processing module, and a robotic arm; the robotic arm has a rebound device at its end; the control and processing module is electrically connected to the navigation module, the laser scanning module, the robotic arm, and the rebound device.

Images