Dynamic compaction settlement measuring method and measuring system of laser radar

Abstract

The invention provides a dynamic compaction settlement measuring method and measuring system of a laser radar. The problems that in an existing method, when the dynamic compactor works, compaction settlement measurement is not accurate, and data collecting and processing real-time performance is insufficient are solved. The three-dimensional point cloud data are obtained through the laser radar and the gyroscope arranged on the tamping machine, and the three-dimensional point cloud data are filtered, so that the depth of the tamping pit is calculated; manual participation is not needed, the real-time performance of measurement is improved, potential safety hazards are reduced, and meanwhile the measurement precision is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of computer vision, and particularly relates to a method for measuring the tamping settlement of a dynamic consolidation machine using a lidar. Background Art

[0002] A dynamic consolidation machine usually refers to a compaction device that uses a lifting equipment to repeatedly lift a hammer weighing 80 - 400 kN (the heaviest reaching 2000 kN) to a height of 8 - 25 m (the highest reaching 40 m) and then let it fall freely when treating the foundation. It forms a powerful shock wave and high stress in the soil, thereby improving the strength of the foundation, reducing compressibility, improving the ability to resist vibration (seismic) liquefaction, eliminating collapsibility, improving the uniformity of the soil, reducing differential settlement, etc. According to the national standard of the People's Republic of China GB50202 - 2018 "Code for Acceptance of Construction Quality of Building Foundation Engineering", the main control items for the quality inspection standard of dynamic consolidation foundation are the strength, bearing capacity and deformation index of the foundation, and the general items are the drop distance of the rammer, the mass of the rammer, the number of tamping passes, the tamping sequence, the number of tamping blows, the tamping position, the tamping range, the intermittent time between the front and back passes, the average tamping settlement of the last two blows, the total tamping settlement and the site flatness, etc. Since the main control items can only be detected through static load, dynamic penetration, etc. after a period of time when the foundation treatment is completed, it is necessary to control the general items during the process for the final tamping settlement result.

[0003] Chinese invention patent 201610050755.1 discloses a laser automatic ranging and recording system for a dynamic tamping machine. This method uses laser measurement to measure the tamping amount. Due to the harsh working conditions on site and the splashing of soil, the laser is easily blocked, resulting in errors when the laser is irradiated to the bottom of the tamping pit, and the actual application effect is not good. Chinese invention patent 201820291125.8 controls the measurement by setting a magnetoelectric encoder on the winch and a PLC controller in the cab. The transmission shaft and other structures are required, resulting in a relatively complex overall structure and low measurement accuracy. Chinese invention patent 201921818680.2 sets a winch, and the winch arranges an ammonium magnet to the Hall sensor, which then transmits it to the encoder in the cab to provide the driver with real-time height information of the rammer. The tamping amount is not accurate enough and the installation and use are complicated, which cannot easily meet the new data requirements for dynamic tamping quality specifications. Chinese invention patent 202210035099.3 integrates multiple sensors and intelligent control systems, calculates the position of the steel cable, verifies and corrects according to the soil quality, and realizes the measurement and real-time monitoring of the tamping amount. However, the system is highly complex and relies on a pre-set soil database, and its adaptability to special environments is not strong enough. Chinese invention patent 202310672126.2 proposes a method for real-time monitoring of the tamping amount based on binocular vision and neural network models. By integrating binocular cameras and neural network models, real-time and high-precision monitoring of the tamping amount is realized. However, the camera may affect the image quality under complex lighting conditions such as strong light, shadows, and nighttime, and in the presence of dust and occlusion, thereby affecting the detection accuracy. Chinese invention patent CN202410617599.7 proposes a method for measuring the depth of irregular pits based on point cloud data. This method has high requirements on the quality of the point cloud itself, and the measurement accuracy is low when facing a large undulating ground or multiple different pits at the same time. Chinese invention patent CN202211328501.3 proposes a method for measuring tamping amount based on a TOF camera. The TOF camera is used to obtain the three-dimensional information of the tamping pit and calculate the tamping amount in real time. In this method, the reflectivity of the TOF camera itself is affected by the surface loose soil and flatness. The three-dimensional point cloud map obtained in the actual scene is prone to distortion, which in turn affects the accuracy of subsequent processing. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art and solve the problems of inaccurate measurement of tamping amount and insufficient real-time performance of data collection and processing during the operation of the dynamic tamping machine, the present invention provides a method and system for measuring the tamping amount of dynamic tamping using a laser radar. The purpose of the present invention is to improve the following aspects:

[0005] (1) The tamping amount measurement solutions actually implemented in the prior art require manual participation or the addition of a large number of linkage mechanical devices. The solution layout and operation are relatively complicated and there are safety hazards.

[0006] (2) The operating site of a dynamic compactor often has harsh working conditions, with mud splashing and the machine vibrating. The actual application effect of existing technical solutions is not good.

[0007] (3) The repeated precision of the settlement measurement solution applied at the present stage is not high, which affects the judgment of the dynamic compaction operation effect.

[0008] (4) When the solution involves human participation, the algorithm effect is easily affected by humans, and unreasonable operations will have a greater impact on the measurement results.

[0009] A method for measuring the dynamic compaction settlement amount of a lidar includes the following steps:

[0010] Step 1: Obtain the three-dimensional point cloud data of the lidar and the inclination angle of the gyroscope.

[0011] Obtain the three-dimensional point cloud data from the lidar; obtain the inclination angle of the lidar relative to the ground from the gyroscope; the lidar and the gyroscope are installed in the same plane; the lidar uses a right-handed coordinate system, where the x-axis is the front of the lidar, and the z-axis is the direction from the bottom to the top when the lidar moves vertically; the gyroscope uses a right-handed coordinate system, and the direction of the z-axis of the gyroscope is the same as the direction of gravity; the y-axis of the gyroscope is consistent with the y-axis direction of the lidar; the angle transformation between the lidar coordinate system and the gyroscope coordinate system is represented by Euler angles.

[0012] Step 2: Perform adaptive filtering processing on the three-dimensional point cloud data to obtain the point cloud after filtering out the outlier points.

[0013] Obtain the low-density point cloud from the three-dimensional point cloud data by the method of random downsampling; calculate the average distance from each point in the low-density point cloud to its neighboring points; calculate the global variance of the distance variance between each point and its neighboring points in the low-density point cloud; then perform adaptive voxelization filtering on the low-density point cloud according to the average distance from each point to its neighboring points; calculate the sampling range of the k nearest points and the threshold for judging outlier points for any point in the low-density point cloud according to the global variance of the distance variance between each point and its neighboring points in the low-density point cloud; use the sampling range of the k nearest points and the threshold for judging outlier points for any point in the low-density point cloud to perform statistical outlier filtering to obtain the point cloud after filtering out the outlier points.

[0014] Step 3: Calculate the depth of the dynamic compaction pit according to the point cloud after filtering out the outlier points.

[0015] The point cloud after filtering out the outliers is rotated along the y-axis; the rotation angle is the same as the inclination angle, so that the z-axis of the point cloud after filtering out the outliers is facing upward, and the rotated point cloud is obtained. Then, in the rotated point cloud, a histogram distributed along the z-axis is counted; when there is at least one peak in the histogram, the largest peak is taken as the point cloud of the ground part in the rotated point cloud, and the random sampling consensus algorithm RANSAC is used to extract the plane in the point cloud of the ground part, and the extracted plane is regarded as the ground position; when there is no peak in the histogram, the random sampling consensus algorithm RANSAC is applied to obtain the plane in the rotated point cloud, and the extracted plane is regarded as the ground position;

[0016] In the rotated point cloud, all point clouds with z-axis height above the ground are eliminated to obtain the point clouds of the pit wall and the pit bottom, and the point cloud corresponding to the ramming pit with the smallest x-coordinate on the laser radar's coordinate axis xOz plane is selected as the target ramming pit; in the rotated point cloud, the point cloud with z-axis height not less than n in the point cloud where the target ramming pit is located is taken as the pit bottom part, and the average pit bottom position is calculated; the ramming pit depth is calculated based on the pit bottom position and the ground position; the value of n is 5%.

[0017] Furthermore, the specific steps of performing adaptive filtering on the three-dimensional point cloud data to obtain the point cloud after filtering out the outliers are:

[0018] The low-density point cloud N is obtained by randomly downsampling the three-dimensional point cloud data. sampled :

[0019] N sampled =rendam_sample(N points ), where N points It is three-dimensional point cloud data;

[0020] Calculate the low-density point cloud N sampled The average distance d between each point and its neighboring points global_mean :

[0021]

[0022] where d ij is the distance from the i-th point to its j-th neighbor, k is the average point cloud density per cubic centimeter, is the average distance of k neighbors;

[0023] In the low-density point cloud, an adaptive voxelization filter voxelization (·) is performed according to the average distance from each point to its neighboring points to obtain the voxelized point cloud N voxelization , to reduce and unify the point cloud density:

[0024] N voxelization =voxelization(Nsampled , d global_mean );

[0025] Calculate the global variance d of the distance variance between each point in the low-density point cloud N sampled and its neighboring points global_dev :

[0026]

[0027] where σ i is the variance of the distance from each point to its neighboring points;

[0028] According to the value range of d global_dev , calculate the sampling range k of the k nearest points to any point in the low-density point cloud mean and the threshold t for judging outliers dev_threshold :

[0029] Use the sampling range k of the k nearest points to any point in the low-density point cloud mean and the threshold t for judging outliers dev_threshold to perform statistical outlier filtering and obtain the point cloud N after filtering out outliers inlier :

[0030] N inlier = statistical_outlier_removal(k mean , t dev_threshold )

[0031] where statical_outlier_removal(·) is statistical outlier filtering.

[0032] Furthermore, according to the value range of d global_dev , calculate the sampling range k of the k nearest points to any point in the low-density point cloud mean and the threshold t for judging outliers dev_threshold The calculation steps are as follows:

[0033] If d global_dev is greater than 0.8, then the value of k mean is 200; the value of t dev_threshold is 3.0;

[0034] If d global_dev is greater than 0.5 and less than or equal to 0.8, then the value of k mean is 150, and the value of t dev_threshold is 2.5;

[0035] If d global_dev is greater than 0.1 and less than or equal to 0.5, then the value of k mean is 100, and the value of t dev_thresholdThe value is 2.0;

[0036] If d global_dev is greater than 0.05 and less than or equal to 0.1, then k mean has a value of 75, and t dev_threshold has a value of 1.5:

[0037] If d global_dev is greater than 0 and less than or equal to 0.05, then k mean has a value of 25, and t dev_threshold has a value of 1.0.

[0038] Furthermore, the specific steps to calculate the depth of the tamping pit based on the point cloud after filtering out outliers are as follows:

[0039] Rotate the point cloud after filtering out outliers along the y-axis; the rotation angle is the same as the inclination angle, so that the z-axis of the point cloud after filtering out outliers is upward, obtaining the rotated point cloud. Subsequently, in the rotated point cloud, count the histogram of the distribution according to the z-axis;

[0040] When there is at least one peak in the histogram, in the rotated point cloud, take the largest peak as the point cloud of the ground part, and use the Random Sample Consensus algorithm RANSAC to extract the plane in the point cloud of the ground part, and regard the extracted plane as the ground G i :

[0041] G i ={(x, y, z)|A i x + B i y + C i z + D i = 0}, where i is the number of tamping times; A i , B i , C i and D i are the coefficients of the ground G i ;

[0042] When there is no peak in the histogram, in the rotated point cloud, apply the Random Sample Consensus algorithm RANSAC to obtain the plane where the ground is located:

[0043] G i ={(x, y, z)|A i x + B i y + C i z + D i = 0} as the ground;

[0044] In the rotated point cloud, remove all points with a z-axis height on the ground G iFor the above point cloud, the point cloud of the rammed pit wall and the bottom part of the pit is obtained. At this time, multiple rammed pits may be obtained. Select the point cloud corresponding to the rammed pit with the smallest x coordinate on the xOz plane of the lidar coordinate axis as the target rammed pit;

[0045] In the rotated point cloud, take the point cloud N where the z-axis height in the point cloud where the rammed pit is located is not lower than n bottom As the bottom part of the pit, calculate the average position of the bottom of the pit, Bottom i = mean(N bottom ), N bottom = {(x, y, z)|z < n×L z}, L z is the length of the z-axis from the rammed pit to the ground part, and mean(·) is to calculate the average value of all point coordinates in the point cloud N bottom ;

[0046] According to the bottom position Bottom i and the ground position G i , calculate the depth D of the rammed pit i :

[0047]

[0048] D i is the depth of the rammed pit calculated this time.

[0049] Furthermore, filter the coefficients of the plane where the ground is located to reduce errors; the steps for filtering the coefficients of the plane where the ground is located are:

[0050] Compare G i with the previously obtained ground G i-1 to judge the coincidence degree between G i and G i-1 :

[0051] Calculate condition 1 and condition 2;

[0052] The said condition 1 is to judge whether the cosine value cosθ of the included angle between G i and G i-1 is greater than 0.996;

[0053]

[0054] The said condition 2 is to judge whether the distance d between G i and G i-1 is less than 20;

[0055]

[0056] If either condition 1 or condition 2 is not satisfied, it is considered that the ground errors obtained from the two measurements are large; if both condition 1 and condition 2 are satisfied, it is considered that the ground errors obtained from the two measurements are small;

[0057] If the ground errors obtained from the two measurements are large, the specific method for performing mean filtering on the parameters of the current ground is as follows:

[0058]

[0059] where A i 、B i 、C i and D i are the ground coefficients after filtering;

[0060] Update the ground G i :

[0061] G i ={(x,y,z)|A' i x + B' i y + C' i z + D' i =0}

[0062] Update the ground G i using the ground coefficients after filtering as the ground for this calculation.

[0063] A measurement system for the dynamic compaction settlement measurement method using a lidar includes a lidar, a gyroscope, a support plate, a dynamic compactor, and a control computer; the support plate is provided on the dynamic compactor; the lidar and the gyroscope are provided on the support plate; the support plate is rigidly connected to the lidar and the gyroscope respectively; the observation range of the lidar covers the position of the bottom of the compaction pit; the lidar and the gyroscope are connected to the control computer through a network cable.

[0064] The steps of the system for the dynamic compaction settlement measurement method using a lidar are as follows:

[0065] Step S1: Initial measurement, obtain the inclination angle of the lidar relative to the ground;

[0066] The system obtains the inclination angle of the lidar relative to the ground from the gyroscope and saves the inclination angle information;

[0067] Start the dynamic compactor, the rammer completes the first compaction operation, and wait for the control cable of the dynamic compactor to lift the rammer;

[0068] Step S2: During the process of the rammer rising, the system collects three-dimensional point cloud data obtained by a single lidar scan, and calculates the depth of the compaction pit according to the three-dimensional point cloud data and the inclination angle; at the same time, calculate the settlement H i ,Hi = D i -D i-1 ; At this time, the dynamic compactor continues to perform the action of raising the rammer until the rammer reaches the predetermined height;

[0069] Step S3: Calculate the total tamping settlement H and the difference between the current tamping settlement and the previous tamping settlement, and judge whether the difference between the last two tamping settlements and the total tamping settlement meet the requirements of the tamping operation conditions:

[0070] The requirements for the tamping operation conditions are:

[0071] Condition 3: Calculate the current tamping settlement H i and the previous tamping settlement H i-1 difference ΔH:

[0072] ΔH = |H i -H i-1 |;

[0073] Judge whether △H is less than the value required by the actual project quality;

[0074] Condition 4: Calculate the total tamping settlement H as:

[0075]

[0076] Judge whether the total tamping settlement H reaches the value required by the actual project quality; where num is the current total number of tamping times;

[0077] If both Condition 3 and Condition 4 are satisfied, the tamping operation of the tamping pit reaches the operation requirements, and the tamping operation stops; if any one of Condition 3 and Condition 4 is not satisfied, continue to execute Step S2 until the project quality requirements are met.

[0078] Furthermore, in Step S2, to reduce errors, for the depths of the tamping pits calculated multiple times in one tamping operation, the z-score method of standard scores is used to eliminate bad points, and the average value of the depths of the tamping pits after eliminating bad points is calculated. The difference between the average value of the depths of the tamping pits multiple times and the previous depth of the tamping pit is used to obtain the updated tamping settlement, and the updated tamping settlement is used as the measurement result of this time.

[0079] The beneficial effects of the present invention compared with the existing tamping settlement measurement methods are:

[0080] (1), The equipment is light and the process is simple. The present invention is a measurement method based on lidar, and the three-dimensional point cloud of the tamping pit is obtained through lidar. Compared with the method of obtaining the rope length by a winch and an electric encoder counting or manually recording data and then calculating the tamping settlement, the present invention only needs to mount a lidar detection device to obtain the tamping pit information, and then perform analysis and calculation to obtain more accurate tamping settlement information.

[0081] (2) Strong robustness in dealing with harsh construction site conditions. The working site environment of the dynamic compactor is complex, and factors such as mud splashing, floating dust, strong wind, and vibration will seriously affect the quality of point cloud data. Existing laser ranging or encoder-based solutions are prone to measurement errors in such an environment. The adaptive filtering algorithm of the present invention can intelligently adjust the filtering parameters automatically according to the real-time working conditions, ensuring that high-precision point cloud data can be obtained even in the harshest environment. Through adaptive filtering, the system can not only effectively remove noise points but also retain the characteristics of the real tamping surface, thus greatly improving the accuracy and reliability of the measurement results. In addition, adaptive filtering does not require manual intervention, avoiding measurement errors caused by unreasonable manual parameter settings, and truly realizing automation and intelligence. At the same time, because a large-scale ground is used as the measurement reference object, it can ensure the repeated accuracy of multiple measurements and avoid measurement errors caused by vibration and equipment skew.

[0082] (3) Fast data acquisition and processing result output. The existing manual measurement scheme not only takes a long time to obtain data but also takes a longer time for subsequent data feedback and correction, reducing the tamping operation speed. The present invention can automatically record and store the tamping settlement information, and the work of data analysis and recording can be completed within a millisecond response time, greatly improving work efficiency.

[0083] (4) Quantifying and automatically analyzing the tamping operation. The existing solution for the winch does not propose the function of data recording. The system of the present invention outputs information such as point cloud-related parameters and tamping settlement obtained from the tamping operation, facilitating the supervision and verification of the tamping operation. Description of the Drawings

[0084] Figure 1 is the algorithm flow chart;

[0085] Figure 2 is the operation flow chart. Detailed Embodiment

[0086] A method for measuring the tamping settlement of dynamic compaction using lidar includes the following steps:

[0087] Step 1: Obtain the three-dimensional point cloud data of the lidar and the inclination angle of the gyroscope;

[0088] Obtain the three-dimensional point cloud data from the lidar; obtain the inclination angle of the lidar relative to the ground from the gyroscope; the lidar and the gyroscope are installed on the same plane;

[0089] The lidar uses a right-handed coordinate system, where the x-axis is the front of the lidar, and the z-axis is the direction from the bottom to the top when the lidar moves vertically; the gyroscope uses a right-handed coordinate system, and the direction of the z-axis of the gyroscope is the same as the direction of gravity; the y-axis of the gyroscope is consistent with the y-axis direction of the lidar; the angular transformation between the coordinate system of the lidar and the coordinate system of the gyroscope is represented by Euler angles;

[0090] Step 2: Perform adaptive filtering on the three-dimensional point cloud data;

[0091] Obtain the low-density point cloud N from the three-dimensional point cloud data by random downsampling sampled to reduce the point cloud density:

[0092] N sampled = rendam_sample(N points ), where N points is the three-dimensional point cloud data;

[0093] Calculate the average distance d from each point in the low-density point cloud N sampled to its neighboring points: global_mean :

[0094]

[0095] where d ij is the distance from the i-th point to its j-th neighbor, k is the average point cloud density per cubic centimeter, is the average distance of k neighbors;

[0096] In the low-density point cloud, perform adaptive voxelization filtering voxelization(·) according to the average distance from each point to its neighboring points to obtain the voxelized point cloud N voxelization to reduce and unify the point cloud density:

[0097] N voxelization = voxelization(N sampled , d global_mean );

[0098] Calculate the global variance d of the distance variance between each point and its neighboring points in the low-density point cloud N sampled : global_dev :

[0099]

[0100] where σ i is the variance of the distance from each point to its neighboring points;

[0101] Calculate the filtering parameters k global_dev and t according to the range of the d mean valuedev_threshold :

[0102] If d global_dev is greater than 0.8, then k mean has a value of 200; t dev_threshold has a value of 3.0;

[0103] If d global_dev is greater than 0.5 and less than or equal to 0.8, then k mean has a value of 150, t dev_threshold has a value of 2.5;

[0104] If d global_dev is greater than 0.1 and less than or equal to 0.5, then k mean has a value of 100, t dev_threshold has a value of 2.0;

[0105] If d global_dev is greater than 0.05 and less than or equal to 0.1, then k mean has a value of 75, t dev_threshold has a value of 1.5:

[0106] If d global_dev is greater than 0 and less than or equal to 0.05, then k mean has a value of 25, t dev_threshold has a value of 1.0:

[0107] The filtering parameter k mean is for any point in the N sampled point cloud, selecting the k nearest points as the sampling range for filtering; the filtering parameter t dev_threshold is the threshold for judging outliers;

[0108] Using the filtering parameters k mean and t dev_threshold to perform statistical outlier filtering to filter out outliers and obtain the point cloud N inlier after filtering out outliers:

[0109] N inlier = statistical_outlier_removal(k mean , t dev_threshold )

[0110] where statical_outlier_removal(·) is statistical outlier filtering;

[0111] Step 3: Calculate the depth of the ramming pit based on the point cloud after filtering out outliers;

[0112] Rotate the point cloud after filtering out the outliers along the y-axis; the rotation angle is the same as the dip angle, so that the z-axis of the point cloud after filtering out the outliers faces upward, obtaining the rotated point cloud. Subsequently, in the rotated point cloud, count the histogram of the distribution according to the z-axis;

[0113] When there is at least one peak in the histogram, in the rotated point cloud, take the largest peak as the point cloud of the ground part, and use the Random Sample Consensus algorithm RANSAC to extract the plane in the point cloud of the ground part, and regard the extracted plane as the ground G i :

[0114] G i ={(x, y, z)|A i x + B i y + C i z + D i = 0}, where i is the number of tamping times at this time; A i 、B i 、C i and D i are the coefficients of the ground G i ;

[0115] When there is no peak in the histogram, in the rotated point cloud, apply the Random Sample Consensus algorithm RANSAC to obtain the plane where the ground is located:

[0116] G i ={(x, y, z)|A i x + B i y + C i z + D i = 0} as the ground;

[0117] If the current ground spatial position is quite different from the previous one, the ground parameters will be mean-filtered according to historical data to reduce the impact of contingency;

[0118] Compare G i with the ground G i-1 obtained last time to judge the coincidence degree between G i and G i-1 :

[0119] The steps to judge the coincidence degree between G i and G i-1 are as follows:

[0120] Condition 1: Calculate the cosine value of the angle between G i and G i-1 . The cosine value of the angle between two planes is:

[0121]

[0122] Determine whether cosθ is greater than 0.996;

[0123] Condition 2: Calculate G i and G i-1 The distance between the two planes is:

[0124]

[0125] Determine whether d is less than 20;

[0126] If any one of Condition 1 and Condition 2 is not satisfied, it is considered that the ground error obtained from the two measurements is large; if both Condition 1 and Condition 2 are satisfied, the ground error obtained from the two measurements is small;

[0127] If the ground error obtained from the two measurements is large, perform mean filtering on the parameters of the current ground to reduce the accidental influence. The specific method is:

[0128] Update the ground G i to G i ={(x,y,z)|A' i x + B' i y + C' i z + D' i = 0};

[0129] In the rotated point cloud, remove all the point clouds with the z-axis height above the ground G i to obtain the point clouds of the rammed pit wall and the pit bottom. At this time, multiple rammed pits may be obtained. Select the point cloud corresponding to the rammed pit with the smallest x-coordinate on the xOz plane of the lidar coordinate axis as the target rammed pit;

[0130] In the rotated point cloud, take the point cloud N bottom with the z-axis height not lower than n in the point cloud where the rammed pit is located as the pit bottom part, and calculate the average position Bottom i = mean(N bottom ), N bottom ={(x,y,z)|z < n×L z}, L z is the length of the z-axis from the rammed pit to the ground part, and mean(·) is to calculate the average value of all point coordinates in the point cloud N bottom ;

[0131] According to the pit bottom position Bottom i and the ground position G i , calculate the rammed pit depth D i :

[0132]

[0133] The value of n is 5%.

[0134] A system for measuring the tamping settlement amount using a lidar, comprising a lidar, a gyroscope, a support plate, a tamping machine and a control computer; the support plate is arranged on the tamping machine; the lidar and the gyroscope are arranged on the support plate; the support plate is rigidly connected to the lidar and the gyroscope respectively; the observation range of the lidar covers the position of the bottom of the tamping pit; the lidar and the gyroscope are connected to the control computer through a network cable;

[0135] The steps of the system for measuring the tamping settlement amount using a lidar are as follows:

[0136] Step S1: Initial measurement, obtaining the inclination angle of the lidar relative to the ground;

[0137] The system obtains the inclination angle of the lidar relative to the ground from the gyroscope and saves the inclination angle information;

[0138] Start the tamping machine, the rammer completes the first tamping operation, and wait for the control cable of the tamping machine to lift the rammer;

[0139] Step S2: During the process of the rammer rising, the system collects the three-dimensional point cloud data obtained by a lidar scan once, and calculates the depth of the tamping pit using the method for measuring the tamping settlement amount of the lidar according to the three-dimensional point cloud data and the inclination angle; at the same time, calculate the settlement amount H i = D i - D i-1 ; At this time, the tamping machine continues to perform the action of lifting the rammer until the rammer reaches the predetermined height;

[0140] Step S3: Calculate the total settlement amount H and H i and the difference from the previous settlement amount H i-1 to analyze whether the difference between the last two settlement amounts and the total settlement amount meet the requirements of the tamping operation:

[0141] The requirements for the tamping operation are as follows:

[0142] Condition 3: Calculate H i and H i-1 The difference between the settlement amounts of the two tamping blows is:

[0143] ΔH = |H i - H i-1 |;

[0144] Judge whether △H is less than the value required by the actual project quality;

[0145] Condition 4: Calculate the total settlement amount H, and the total settlement amount is:

[0146]

[0147] Determine whether H reaches the value required for the actual project quality; where num is the total number of tamping times currently.

[0148] If both Condition 3 and Condition 4 are satisfied, the tamping pit tamping operation meets the operation requirements, and the tamping operation is stopped; if either Condition 3 or Condition 4 is not satisfied, continue to execute Step S2 until the project quality requirements are met.

[0149] In Step S2, to reduce errors, for the depths of the tamping pits calculated multiple times in one tamping operation, the standard score method z-score is used to eliminate bad points, and the average value of the depths of the tamping pits after eliminating bad points is calculated. The difference between the average value of the depths of the tamping pits and the depth of the previous tamping pit is used to obtain the updated tamping amount, and the updated tamping amount is used as the measurement result of this time.

[0150] The present invention will be further described below in conjunction with the drawings and embodiments.

[0151] The present invention proposes a method for measuring the tamping amount based on lidar. When the dynamic compactor performs the tamping operation, it measures and analyzes the tamping amount, tamping times, etc. of the dynamic compaction foundation quality acceptance code, so as to determine whether the tamping operation meets the project quality requirements. The technical solution is as follows:

[0152] (1) 3D point cloud scanning of the tamping pit. The system first needs to be installed on the tamper to keep the tamping pit to be measured within the detection range of the lidar. During the operation of the system, the lidar continuously scans the tamping pit to construct a 3D point cloud map of the tamping pit for subsequent processing; the gyroscope simultaneously records information such as the scanning angle of the lidar for deviation correction during subsequent processing.

[0153] (2) Adaptive filtering. First, analyze and process the point cloud data measured in Step (1), and filter the noise data through the adaptive filtering algorithm. This algorithm first calculates features such as the distribution density of the input point cloud in space, and then adaptively adjusts the filtering parameters according to these features to ensure high-quality point cloud data can be obtained under different working conditions. Adaptive filtering can not only effectively remove low-density noise points caused by geological problems, vibration, and dust diffuse reflection, but also retain the characteristics of the real tamping surface, thus providing a reliable basis for subsequent calculation of the tamping amount.

[0154] (3) Rotate the point cloud and draw the point cloud histogram. According to the gyroscope data corresponding to the point cloud in Step (2), perform a spatial transformation on the point cloud data to make the point cloud data collected at different times basically maintain the same spatial position and angle, which is convenient for comparing the depth data at different times. Subsequently, draw a point cloud histogram along the z-axis direction to obtain the spatial distribution information of the point cloud.

[0155] (4) Distinguish the ground and the position of the tamping pits. According to the point cloud histogram in step (3), first determine the height range where the ground is located, intercept this part of the data, fit the position of the ground, and correct it according to the historical data of the previous ground. The part below the ground is judged as the positions of each tamping pit.

[0156] (5) Extract the target tamping pit and calculate the depth. According to the position of the tamping pit obtained in step (4), select the tamping pit with the smallest x coordinate on the xOy plane of the lidar coordinate axis as the target tamping pit. After extracting this part of the point cloud, obtain the position of the bottom of the pit. Subsequently, calculate the distance from the bottom of the pit to the ground plane as the depth of the tamping pit for this detection.

[0158] (6) Obtain the tamping settlement. Analyze the data in step (5), compare the latest ten historical data, filter out the outliers, obtain the final depth data of the tamping pit, and then subtract the final depth data of the previous time to obtain the tamping settlement for this time. The measured tamping settlement result will be sent to the UI for display, and a signal will be sent to remind when the construction quality requirements are met.

[0159] (7) Save and export the tamping settlement data. Keep records of the data obtained in each of steps (1), (5), and (6), including the depth measured each time, the tamping settlement, the scanning angle, the pit number, etc., and the data table and work report can be exported, which is convenient for the supervision and verification of the tamping work.

[0160] To more clearly illustrate the technical solutions implemented in the present invention, the following will briefly introduce each module required in the description of the embodiments. Figure 1 Shows the algorithm operation process of this method. Starting from the system startup, it goes through automatic parameter setting, point cloud acquisition and preprocessing, feature extraction, depth analysis, until finally calculating the pit settlement; while Figure 2 depicts the operation process of this method in actual application, covering the operation process from the rammer operation to data acquisition, and after obtaining the tamping settlement through algorithm processing, it is displayed and exported. The whole process closely focuses on how to efficiently and accurately evaluate the tamping data and provides detailed analysis results for users. These two flowcharts not only help to understand the working principle and technical details of the present invention, but also provide guidance for actual operation.

[0161] Refer to Figure 1 and Figure 2 , the implementation steps of the present invention are as follows:

[0162] Step 1, initial measurement. Fix the lidar on the dynamic compactor or at a position on the ground where the bottom of the tamping pit can be expected to be observed. Read the inclination angle of the lidar relative to the ground from the gyroscope, and the system will save the inclination angle information. Subsequently, release the steel cable to let the rammer fall freely from a predetermined height in the air for the first ramming operation. After completing the ramming action, wait for the dynamic compactor to control the steel cable to raise the rammer.

[0163] Step 2: During the process of the rammer rising, the system continuously collects the three-dimensional point cloud data obtained by lidar scanning and calculates the depth of the tamping pit based on the data. After a certain delay (e.g., 10 seconds) when the rammer no longer blocks the lidar field of view, click the measurement button. The system calculates the tamping settlement of two tamping blows based on the obtained depth of the tamping pit. At this time, the dynamic compactor continues to perform the rammer rising action until the rammer reaches the predetermined height.

[0164] Step 3: After the system obtains the point cloud data, it first performs adaptive filtering processing. First, random downsampling is used to reduce the point cloud density to obtain a low-density point cloud N sampled = rendam_sample(N points ), where N points is the original point cloud. Then calculate the average distance d global_mean of each point to its neighboring points to perform adaptive voxelization filtering voxelization(·) to reduce and unify the point cloud density:

[0165]

[0166] N voxelization = voxelization(N sampled , d global_mean )

[0167] where d ij is the distance from the i-th point to its k-th neighbor, k is the average point cloud density per cubic centimeter, is the average distance of k neighbors, and N voxelization is the voxelized point cloud. Then calculate the global variance d global_dev of the distance variance between each point and its neighboring points in the point cloud, and obtain the filtering parameters according to the empirical function get_parameters(·) obtained through multiple experiments for statistical outlier filtering statical_outlier_removal(·) to filter out outliers:

[0168]

[0169] k mean , t dev_threshold = get_parameters(d global_dev )

[0170] N inlier = statistical_outlier_removal(k mean , t dev_threshold )

[0171] where N inlierTo filter out the point cloud after outliers, σ i is the variance of the distance from each point to its neighboring points, k mean and t dev_threshold is the filtering parameter obtained through the empirical function. After testing, most of the noise data has been filtered out.

[0172] Step 4: After obtaining low-noise data, first rotate the point cloud so that its z-axis is facing upward according to the predicted angle between the laser radar and the ground, and then calculate the distribution histogram of the point cloud along the z-axis. And obtain the distribution trend of the point cloud on the z-axis based on the histogram results.

[0173] When there is at least one peak, the largest peak is taken as the point cloud of the ground part, and a plane is extracted from it using the RANSAC algorithm. The plane is regarded as the ground G i ={(x,y,z)|A i x+B i y+C i z+D i =0}, where i is the number of tamping times; when there is no peak, take all point clouds and apply the RANSAC method to obtain the plane G where the ground is located i ={(x,y,z)|A i x+B i y+C i z+D i =0} as the ground. If the current ground spatial position is significantly different from the past, the ground parameters will be averaged and filtered based on historical data to reduce the impact of chance.

[0174] Then remove all z-axis heights above the ground G i The above point clouds can obtain the point clouds of the pit wall and the bottom. At this time, multiple pits may be obtained, so the pit with the smallest x coordinate on the coordinate axis xOz plane is selected as the target pit.

[0175] Take the point cloud whose z-axis height is not less than a certain ratio (e.g. 5%) in the point cloud where the ramming pit is located as the pit bottom part, and calculate the average position B of the pit bottom i =mean(N bottom ), N bottom ={(x,y,z)|z <n%×L z}, L z It is the z-axis length from the ramming pit to the ground.

[0176] According to the bottom position B i and ground position G i , calculate the depth of the ramming pit As a measurement result.

[0177] Step 5: After the measurement button is pressed, the system selects the results of several recent (e.g., 10 times) measurements, eliminates the bad points by the z-score method, and finally calculates the average value of the multiple measurement results as the result of this measurement. The result is saved, and the tamping settlement H is calculated at the same time. i = D i - D i-1 , and all the data are plotted on the software interface.

[0178] Step 6: Compare the calculated result H in Step 5 i with the previous line settlement H i-1 and the total required tamping settlement H, analyze the change in the tamping settlement in the last two times and whether the final tamping settlement data meets the tamping requirements. If it meets the requirements, the tamping operation of this tamping pit reaches the operation requirements; otherwise, continue to execute Steps 1-5 until the engineering quality requirements are met.

[0179] Generally speaking, the present invention is a method for measuring tamping settlement based on lidar, which can use relatively simple devices to reduce the burden of tamping settlement measurement, control the measurement error within the centimeter level under the harsh working conditions such as mud splashing, floating dust, and strong wind during tamping operations, and at the same time display the tamping settlement measurement results in real time, greatly increasing the tamping work efficiency and effectively ensuring the quality of dynamic compaction operations.

Claims

1. A method for measuring the amount of strong compaction by laser radar, characterized in that: The steps include: Step 1: Obtain the 3D point cloud data of the laser radar and the inclination angle of the gyroscope; Obtain three-dimensional point cloud data from the laser radar; obtain the inclination angle of the laser radar relative to the ground from the gyroscope; the laser radar and the gyroscope are installed on the same plane; Step 2: Perform adaptive filtering on the 3D point cloud data to obtain the point cloud after filtering out the outliers; The low-density point cloud is obtained by randomly downsampling the three-dimensional point cloud data; the average distance from each point in the low-density point cloud to its adjacent points is calculated; Calculate the global variance of the distance variance between each point and its neighboring points in the low-density point cloud; Then, according to the average distance from each point to its neighboring points, adaptive voxel filtering is performed on the low-density point cloud; According to the global variance of the distance variance between each point in the low-density point cloud and its neighboring points, the sampling range of the k points closest to any point in the low-density point cloud and the threshold for judging outliers are calculated; Use the k-point sampling range closest to any point in the low-density point cloud and the threshold for judging outliers to perform statistical outlier filtering to obtain the point cloud after filtering out the outliers; Step 3: Calculate the pit depth based on the point cloud after filtering out the outliers; The point cloud after filtering out the outliers is rotated along the y-axis; the rotation angle is the same as the inclination angle, so that the z-axis of the point cloud after filtering out the outliers is facing upward, and a rotated point cloud is obtained. Then, a histogram of the distribution along the z-axis is counted in the rotated point cloud; When there is at least one peak in the histogram, in the rotated point cloud, the largest peak is taken as the point cloud of the ground part, and the random sampling consensus algorithm RANSAC is used to extract the plane in the point cloud of the ground part, and the extracted plane is regarded as the ground position; when there is no peak in the histogram, in the rotated point cloud, the random sampling consensus algorithm RANSAC is applied to obtain the plane, and the extracted plane is regarded as the ground position; in the rotated point cloud, all point clouds with z-axis height above the ground are eliminated to obtain the point clouds of the pit wall and the pit bottom, and the point cloud corresponding to the ramming pit with the smallest x coordinate on the laser radar coordinate axis xOz plane is selected as the target ramming pit; in the rotated point cloud, the point cloud with z-axis height not less than n in the point cloud where the target ramming pit is located is taken as the pit bottom part, and the average pit bottom position is calculated; the ramming pit depth is calculated according to the pit bottom position and the ground position; the value of n is 5%.

2. The method for measuring the amount of strong compaction by laser radar according to claim 1, characterized in that: In step 1, the laser radar adopts a right-handed coordinate system, wherein the x-axis is directly in front of the laser radar, and the z-axis is the direction from the bottom to the top when the laser radar moves vertically; the gyroscope adopts a right-handed coordinate system, and the direction of the z-axis of the gyroscope is the same as the direction of gravity; the y-axis of the gyroscope is consistent with the y-axis direction of the laser radar; the angle transformation between the laser radar coordinate system and the gyroscope coordinate system is expressed using Euler angles.

3. The method for measuring the amount of compaction by laser radar according to claim 1, characterized in that: The specific steps of performing adaptive filtering on the three-dimensional point cloud data to obtain the point cloud after filtering out the outliers are: The low-density point cloud N is obtained by randomly downsampling the three-dimensional point cloud data. sampled : N sampled =rendam_sample(N points ), where N points It is three-dimensional point cloud data; Calculate the low-density point cloud N sampled The average distance d between each point and its neighboring points global_mean : where d ij is the distance from the i-th point to its j-th neighbor, k is the average point cloud density per cubic centimeter, is the average distance of k neighbors; In the low-density point cloud, an adaptive voxelization filter voxelization (·) is performed according to the average distance from each point to its neighboring points to obtain the voxelized point cloud N voxelization , to reduce and unify the point cloud density: N voxelization =voxelization(N sampled ,d global_mean ); Calculate the low-density point cloud N sampled The global variance d of the distance variance between each point and its neighboring points global_dev : where σ i is the variance of the distance from each point to its neighboring points; According to d global_dev The range of values ​​is calculated by sampling the k points closest to any point in the low-density point cloud. mean And the threshold t for judging outliers dev_threshold : Use the k points closest to any point in the low-density point cloud to sample the range k mean And the threshold t for judging outliers dev_threshold Perform statistical outlier filtering to obtain the point cloud N after filtering out the outliers inlier : N inlier =statistical_outlier_removal(k mean ,t dev_threshold ) where statical_outlier_removal(·) is the statistical outlier filter.

4. The method for measuring the amount of strong compaction by laser radar according to claim 3 is characterized in that: According to d global_dev The range of values ​​is calculated by sampling the k points closest to any point in the low-density point cloud. mean And the threshold t for judging outliers dev_threshold The calculation steps are: If d global_dev If k is greater than 0.8, mean The value of is 200; t dev_threshold The value of is 3.0; If d global_dev If k is greater than 0.5 and less than or equal to 0.8, then mean The value of t is 150. dev_threshold The value of is 2.5; If d global_dev If k is greater than 0.1 and less than or equal to 0.5, then mean The value of t is 100. dev_threshold The value of d is 2.0; if d global_dev If k is greater than 0.05 and less than or equal to 0.1, then mean The value of t is 75. dev_threshold The value of is 1.5: If d global_dev If k is greater than 0 and less than or equal to 0.05, then mean The value of t is 25. dev_threshold The value of is 1.

0.

5. The method for measuring the amount of strong compaction by laser radar according to claim 1, characterized in that: The specific steps of computing the pit depth based on the point computing after filtering out the outliers are as follows: The point cloud after filtering out the outliers is rotated along the y-axis; the rotation angle is the same as the inclination angle, so that the z-axis of the point cloud after filtering out the outliers is facing upward, and a rotated point cloud is obtained. Then, a histogram of the distribution along the z-axis is counted in the rotated point cloud; When the histogram has at least one peak, the largest peak is taken as the ground point cloud in the rotated point cloud, and the plane is extracted from the ground point cloud using the random sampling consensus algorithm RANSAC, and the extracted plane is regarded as the ground G i : G i ={(x,y,z)|A i x+B i y+C i z+D i =0}, i is the number of tamping times; A i , B i , C i and D i G i The coefficient of When there is no peak in the histogram, the random sampling consensus algorithm RANSAC is applied to obtain the plane where the ground is located in the rotated point cloud: G i ={(x,y,z)|A i x+B i y+C i z+D i =0} as ground; In the rotated point cloud, remove all points whose z-axis height is above the ground G. i The above point clouds are used to obtain the point clouds of the pit wall and the pit bottom. At this time, multiple pits may be obtained. The point cloud corresponding to the pit with the smallest x coordinate on the laser radar coordinate axis xOz plane is selected as the target pit. In the rotated point cloud, select point N whose z-axis height is not less than n in the point cloud where the ramming pit is located bottom As the bottom part of the pit, calculate the average position of the bottom of the pit i =mean(N bottom ), N bottom ={(x,y,z)|z <n×L z }, L z is the z-axis length of the ramming pit to the ground, mean(·) is the calculated point cloud N bottom The mean of all point coordinates in ; According to the bottom of the pit i and ground position G i , calculate the pit depth D i : D i This is the ramming pit depth calculated this time.

6. The method for measuring the amount of strong compaction by laser radar according to claim 5, characterized in that: The coefficients of the plane where the ground is located are filtered to reduce the error. The steps of filtering the coefficients of the plane where the ground is located are: G i Compared with the ground G obtained last time i-1 Compare and judge G i With G i-1 The overlap between: Calculate conditions 1 and 2; The condition 1 is to judge G i With G i-1 Is the cosine value cosθ of the angle between them greater than 0.996? The condition 2 is to judge G i With G i-1 Whether the distance d between them is less than 20; If any one of conditions 1 and 2 is not met, it is considered that the ground error obtained by the two measurements is large; if conditions 1 and 2 are met at the same time, it is considered that the ground error obtained by the two measurements is small; If the ground error obtained from two measurements is large, the specific method for performing mean filtering on the current ground parameters is: Where A′ i , B′ i , C′ i and D′ i is the ground coefficient after filtering; Update the ground G using the filtered ground coefficients i : G i ={(x,y,z)|A' i x+B' i y+C' i z+D' i =0} Update the ground G with the filtered ground coefficient i As the ground for this calculation.

7. A measurement system for measuring the amount of compaction using any one of the laser radars described in claims 1 to 6, characterized in that: It includes a laser radar, a gyroscope, a support plate, a dynamic compaction machine and a control computer; the support plate is arranged on the dynamic compaction machine; the laser radar and the gyroscope are arranged on the support plate; the support plate is rigidly connected to the laser radar and the gyroscope respectively; the observation range of the laser radar covers the position of the bottom of the tamping pit; the laser radar and the gyroscope are connected to the control computer through a network cable.

8. The measuring system for measuring the amount of compaction using a laser radar according to claim 7, characterized in that: The measurement steps are: Step S1: initial measurement, obtaining the inclination angle of the laser radar relative to the ground; The system obtains the inclination of the laser radar relative to the ground from the gyroscope and saves the inclination information; Start the tamping machine, and the tamping hammer completes the first tamping operation, and wait for the tamping machine control cable to raise the tamping hammer; Step S2: During the process of the rammer rising, the system collects a laser radar scan to obtain three-dimensional point cloud data, and calculates the depth of the ramming pit using the laser radar's strong ramming sinking measurement method based on the three-dimensional point cloud data and the inclination angle; at the same time, the ramming sinking amount H is calculated. i , H i =D i -D i-1 ; At this time, the tamping machine continues to perform the tamping hammer raising action until the tamping hammer reaches the predetermined height; Step S3: Calculate the total tamping amount H and the difference between the current tamping amount and the previous tamping amount, and determine whether the difference between the last two tamping amounts and the total tamping amount meet the tamping operation requirements: The requirements for tamping operations are: Condition 3: Calculate the tamping amount H i Compared with the last tamping amount H i-1 The difference ΔH: ΔH=|H i -H i-1 |; Determine whether △H is less than the value required by the actual engineering quality; Condition 4: Calculate the total tamping amount H as: Determine whether the total tamping amount H reaches the value required for the actual project quality; where num is the current total number of tamping times; if conditions 3 and 4 are met at the same time, the tamping operation of the tamping pit meets the operation requirements and the tamping operation is stopped; if any one of conditions 3 and 4 is not met, continue to execute step S2 until the project quality requirements are met.

9. The steps of the measuring system of the method for measuring the amount of compaction using a laser radar according to claim 8 are characterized in that: In step S2, in order to reduce errors, the standard score method z-score is used to eliminate bad points from the tamping pit depths calculated multiple times in a tamping operation, and the average value of the multiple tamping pit depths after eliminating the bad points is calculated. The average value of the multiple tamping pit depths is subtracted from the previous tamping pit depth to obtain an updated tamping amount, and the updated tamping amount is used as the measurement result of this time.

Images