A roadside laser radar calibration verification method based on intelligent barrel

Abstract

The application discloses a roadside laser radar calibration verification method based on an intelligent cone barrel, relates to the technical field of roadside device calibration, and comprises the following steps: acquiring a 3D model of the intelligent cone barrel; identifying 3D point clouds of the intelligent cone barrel; designing a point cloud registration algorithm and performing point cloud registration on the 3D model and the 3D point clouds of the intelligent cone barrel to obtain calibration parameters and calibration accuracy indexes of the roadside laser radar; setting distance threshold values and direction threshold values; calculating distance activation values and direction activation values; calculating corresponding matching scores according to the distance activation values and the direction activation values; acquiring perception data of the intelligent cone barrel according to the matching scores; and calculating accuracy indexes of roadside perception data. The application can output current sensor calibration errors in real time based on positioning information uploaded by the intelligent cone barrel and laser radar sensor data, can calculate calibration parameters of the roadside laser radar by using multiple intelligent cone barrels, and can also serve as a facility for monitoring roadside perception data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of roadside device calibration, in particular to a roadside laser radar calibration verification method based on an intelligent cone barrel. BACKGROUND

[0002] China's intelligent network connected vehicles have officially entered a new stage of city-level scale promotion and exploration of new business models. Currently, laser radar and cameras are usually used for perception at key intersections. Due to the lack of convenient calibration verification methods, the problem of poor perception data quality caused by inaccurate calibration often occurs during project acceptance.

[0003] In the prior art, the calibration verification of a roadside perception system is usually one of the following three methods: 1. selecting points on a high-precision map, which depends on a high-precision map, increases the calibration threshold, and is not universal; 2. using RTK or a total station to collect static target feature point coordinates, which consumes a lot of manpower and time; and 3. collecting trajectory data by a true value vehicle equipped with a combined inertial navigation system, which has high requirements for the calibration of the combined inertial navigation system of the true value vehicle, cannot be verified online, needs to be driven according to certain distinguishable trajectories, and is inconvenient for error verification of static targets due to traffic regulations.

[0004] As the most common traffic safety facility, a cone barrel has the characteristics of small volume and strong mobility and does not hinder traffic, so it can be placed in a test area for a long time for calibration verification. Based on this, the present application designs a roadside laser radar online calibration verification method based on an intelligent cone barrel, which can output the current sensor calibration error in real time based on the positioning information uploaded by the intelligent cone barrel and the laser radar sensor data, and can use multiple intelligent cone barrels to calculate the calibration parameters of the roadside laser radar and also serve as a facility for monitoring roadside perception data. SUMMARY

[0005] The present application provides a roadside laser radar calibration verification method based on an intelligent cone barrel to overcome at least one technical problem in the prior art.

[0006] The present application provides a roadside laser radar calibration verification method based on an intelligent cone barrel, which includes the following steps:

[0007] Collecting intelligent cone barrel positioning data and roadside perception system data; the intelligent cone barrel positioning data includes spatial positioning coordinates and attitude positioning information of the intelligent cone barrel, and the roadside perception system data includes point cloud data and roadside perception data, and the roadside perception data at least includes perception data of the intelligent cone barrel;

[0008] Obtaining a 3D model of the intelligent cone barrel according to the spatial positioning coordinates and the attitude positioning information of the intelligent cone barrel.

[0009] identifying a 3D point cloud of the smart cone barrel according to the point cloud data;

[0010] designing a point cloud registration algorithm;

[0011] performing point cloud registration on the 3D model and the 3D point cloud of the smart cone barrel by using the point cloud registration algorithm, to obtain a calibration parameter of the roadside laser radar;

[0012] calculating a calibration accuracy index according to the registration result; the calibration accuracy index at least includes a position accuracy and a pose accuracy;

[0013] setting a distance threshold and a direction threshold;

[0014] calculating a distance difference and a direction difference between the roadside perception data and the smart cone barrel positioning data;

[0015] calculating a distance activation value according to the distance threshold and the distance difference;

[0016] calculating a direction activation value according to the direction threshold and the direction difference;

[0017] calculating a corresponding matching score according to the distance activation value and the direction activation value;

[0018] obtaining perception data of the smart cone barrel according to the matching score;

[0019] calculating an accuracy index of the roadside perception data according to the perception data of the smart cone barrel, and spatial positioning coordinates and pose positioning information of the smart cone barrel.

[0020] Optionally, the point cloud registration algorithm includes a point cloud registration formula and a constraint condition; wherein,

[0021] the point cloud registration formula is represented as p i represents an i-th point in the 3D point cloud, q i represents a corresponding point in the 3D model, R represents a rotation matrix, R includes a roll angle a, a pitch angle b and a yaw angle g, t represents a translation vector, and N represents a number of points; w i represents a weight coefficient of the i-th point cloud;

[0022] the constraint condition includes a pose change angle constraint and a convergence condition constraint;

[0023] the pose change angle constraint is represented as a min ≤ a ≤ a max , b min ≤ b ≤ b max , and g min ≤ g ≤ gmax ; wherein, a max , a min respectively represent the upper and lower limits of the roll angle change, b max , b min respectively represent the upper and lower limits of the pitch angle change, g max , g min respectively represent the upper and lower limits of the yaw angle change.

[0024] The convergence condition constraint includes a dynamic constraint and a static constraint, the dynamic constraint is represented as e dynamic = e0+ a v, wherein e0 represents a basic error threshold in a static scene, a represents a weight factor adjusted according to dynamic changes of the scene, v represents a motion speed of the cone barrel, e dynamic represents a convergence error threshold; and the static constraint is represented as p error < e position , r error < e rotation ; wherein p error represents a position error threshold, e position represents a position error threshold, t i,current represents a translation vector of the ith point cloud in this iteration, t i,previous represents a translation vector of the ith point cloud in the last iteration, r error represents an attitude error threshold, r error = |R current - R previous |, e rotation represents an attitude error threshold, R current represents a rotation matrix in this iteration, R previous represents a rotation matrix in the last iteration.

[0025] Optionally, the weight coefficient w i of the point cloud includes a multi-scale error weighting weight or a sensor fusion weight;

[0026] The multi-scale error weighting weight refers to that the weight coefficient w i of the point cloud decreases with the decrease of the number of point clouds;

[0027] The sensor fusion weight refers to that the weight w vision of the point cloud data collected by the roadside perception system and the weight w imu of the intelligent cone barrel positioning data are summed, and is represented as w i = w vision + w imu .

[0028] Optionally, the point cloud registration algorithm is used for point cloud registration of the 3D model and the 3D point cloud of the intelligent cone barrel, and specifically,

[0029] judging whether the intelligent cone barrel is in a static state, if yes, based on the attitude change angle constraint and the static constraint, performing point cloud registration on the 3D model and the 3D point cloud of the intelligent cone barrel by using the point cloud registration formula, and removing abnormal points in the registration process; if no, based on the attitude change angle constraint and the dynamic constraint, performing point cloud registration on the 3D model and the 3D point cloud of the intelligent cone barrel by using the point cloud registration formula, and removing abnormal points in the registration process.

[0030] Optionally, abnormal points are removed in the registration process, specifically:

[0031] an abnormal threshold d is set threshold ;

[0032] the point cloud registration error is calculated according to the point cloud registration formula;

[0033] judging whether the point cloud registration error is greater than the abnormal threshold, if yes, the corresponding point cloud is an abnormal point, and the point cloud is removed; if no, the corresponding point cloud is a normal point, and the point cloud is retained.

[0034] Optionally, the distance activation value is represented as wherein, D t represents the distance activation value, δ d represents the distance threshold, d i represents the distance difference between the roadside perception data and the intelligent cone barrel positioning data obtained by the i-th detection, n represents the detection number, and λ d represents the distance conversion factor.

[0035] Optionally, the direction activation value is represented as wherein, H t represents the direction activation value, δ h represents the direction threshold, h i represents the direction difference between the roadside perception data and the intelligent cone barrel positioning data obtained by the i-th detection, n represents the detection number, and λ h represents the direction conversion factor.

[0036] Optionally, the matching score is represented as wherein, M represents the matching score, and S(·) represents the Sigmoid activation function.

[0037] Optionally, the perception data of the intelligent cone barrel is obtained according to the matching score, specifically:

[0038] determining whether the matching score is greater than a matching threshold, if greater, indicating that the matching is successful, and the corresponding roadside perception data is the perception data of the intelligent cone barrel; if less than or equal to, indicating that the matching is unsuccessful, and the corresponding roadside perception data is not the perception data of the intelligent cone barrel.

[0039] Optionally, the perception data of the intelligent cone barrel comprises perception coordinates and a perception pose of the intelligent cone barrel; and the precision indicator comprises at least a position error and a pose error.

[0040] According to the perception data of the intelligent cone barrel and the spatial positioning coordinates and the pose positioning information of the intelligent cone barrel, a precision indicator of the roadside perception data is calculated, specifically:

[0041] According to the perception coordinates and the spatial positioning coordinates of the intelligent cone barrel, the position error is calculated.

[0042] According to the perception pose and the pose positioning information of the intelligent cone barrel, the pose error is calculated.

[0043] The innovation points of the embodiments of the present application include:

[0044] 1. In the embodiments, based on the positioning information uploaded by the intelligent cone barrel and the laser radar sensor data, the calculation of the roadside laser radar calibration parameters is performed, the high cost of the high-precision map is avoided, and the map range constraint is also avoided. Since the intelligent cone barrel itself is a feature in the laser radar point cloud, it is more convenient to find the feature point in the high-precision map in the laser radar point cloud, and the calibration precision is not lost, which is one of the innovation points of the embodiments of the present application.

[0045] 2. In the embodiments, based on the positioning information uploaded by the intelligent cone barrel and the laser radar sensor data, the current sensor calibration error is output in real time, the calibration verification is completed, and since the intelligent cone barrel can send its own coordinates in real time, the time of the heterogeneous data is synchronized, and the problem of inaccurate calibration result caused by the movement of the feature point in the existing calibration verification method is avoided, which is one of the innovation points of the embodiments of the present application.

[0046] 3. In the embodiments, based on the positioning information uploaded by the intelligent cone barrel and the laser radar sensor data, the roadside perception data is monitored, and since the intelligent cone barrel can be placed in the test site for a long time, it is not easy to affect the traffic flow, it is convenient to troubleshoot the error caused by the exposure time of the mechanical laser radar, and in the application scene, the true value car scheme is more suitable for long-time monitoring and early warning of the roadside perception data, which is one of the innovation points of the embodiments of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0048] Figure 1 A flow chart of the calibration verification method provided by the embodiment of the present application;

[0049] Figure 2 A flow chart of the outlier rejection method provided by the embodiment of the present application. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments only constitute some embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0051] It should be noted that the terms "comprising" and "having" and any variations thereof in the embodiments of the present application and the drawings are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed, or can optionally include other steps or units inherent to the process, method, product or device.

[0052] The embodiment of the present application discloses a roadside laser radar calibration verification method based on intelligent cone barrel. The following will be described in detail respectively.

[0053] Figure 1 A flow chart of the calibration verification method provided by the embodiment of the present application, please refer to Figure 1 The roadside laser radar calibration verification method based on intelligent cone barrel provided by the embodiment of the present application comprises:

[0054] Step 1: Collect intelligent cone barrel positioning data and roadside perception system data; the intelligent cone barrel positioning data includes spatial positioning coordinates and attitude positioning information of the intelligent cone barrel, and the roadside perception system data includes point cloud data and roadside perception data, and the roadside perception data at least includes perception data of the intelligent cone barrel;

[0055] Step 2: Obtain the 3D model of the intelligent cone barrel according to the spatial positioning coordinates and the attitude positioning information of the intelligent cone barrel;

[0056] Step 3: Identify the 3D point cloud of the intelligent cone barrel according to the point cloud data;

[0057] Step 4: Design a point cloud registration algorithm;

[0058] Step 5: Use the point cloud registration algorithm to perform point cloud registration on the 3D model and 3D point cloud of the intelligent cone barrel, and obtain the calibration parameters of the roadside laser radar;

[0059] Step 6: Calculate the calibration accuracy index according to the registration result; the calibration accuracy index at least includes position accuracy and attitude accuracy;

[0060] Step 7: Set a distance threshold and a direction threshold;

[0061] Step 8: Calculate the distance difference and direction difference between the roadside perception data and the intelligent cone barrel positioning data;

[0062] Step 9: Calculate the distance activation value according to the distance threshold and the distance difference;

[0063] Step 10: Calculate the direction activation value according to the direction threshold and the direction difference;

[0064] Step 11: Calculate the corresponding matching score according to the distance activation value and the direction activation value;

[0065] Step 12: Obtain the perception data of the intelligent cone barrel according to the matching score;

[0066] Step 13: Calculate the accuracy index of the roadside perception data according to the perception data of the intelligent cone barrel, and the spatial positioning coordinates and attitude positioning information of the intelligent cone barrel.

[0067] Specifically, please refer to Figure 1 The roadside laser radar calibration verification method based on the intelligent cone barrel provided by the embodiment of the present application verifies the calibration and calibration error based on the positioning information uploaded by the intelligent cone barrel and the laser radar sensor data. Therefore, first, the intelligent cone barrel positioning data and the roadside perception system data need to be collected through step 1. The intelligent cone barrel positioning data includes the spatial positioning coordinates and the attitude positioning information of the intelligent cone barrel, and the roadside perception system data includes the point cloud data and the roadside perception data. The roadside perception data at least includes the perception data of the intelligent cone barrel.

[0068] The core function of the intelligent cone barrel system is to realize accurate positioning through an RTK (Real-time kinematic) module, which includes a GNSS (Global Navigation Satellite System) and an IMU (Inertial Measurement Unit). The satellite signal is received by the GNSS antenna, and the position is corrected through the IMU, so as to realize the positioning accuracy of the intelligent cone barrel at the centimeter level, and send the real-time position data of the intelligent cone barrel to the embedded computing module, and through the combination of other sensors (such as an accelerometer, a gyroscope, etc.), data fusion and processing are carried out, and the position and attitude of the cone barrel are calculated in real time. At the same time, the embedded computing module is also responsible for uploading the processed data through the communication module, and receiving remote instructions.

[0069] In this embodiment, the intelligent cone barrel system uses a 12V DC battery for power supply, so as to ensure the continuous work in an outdoor environment. The battery is charged from the alternating current mains through a charging module, so as to ensure that the battery can supply power for a long time.

[0070] After obtaining the positioning data of the intelligent cone barrel, in step 2, the 3D model of the intelligent cone barrel can be built according to the spatial positioning coordinates and the attitude positioning information of the intelligent cone barrel. The 3D point cloud obtained by the roadside perception system positioning includes the 3D point cloud of the intelligent cone barrel and the 3D point cloud of other traffic participants such as people and vehicles, so in step 3, the 3D point cloud of the intelligent cone barrel needs to be identified from all the point cloud data.

[0071] After obtaining the 3D point cloud of the intelligent cone barrel, the 3D model of the intelligent cone barrel can be used to register the 3D point cloud of the intelligent cone barrel, so as to realize the calibration of the roadside laser radar. When performing point cloud registration, the point cloud registration algorithm needs to be registered, therefore, the present application designs a point cloud registration algorithm through step 4.

[0072] The present application is based on the ICP (iterative closest point) point cloud matching algorithm, and combines the unique geometric features of the intelligent cone barrel and the multi-source sensing information, and designs a point cloud registration formula and constraint condition suitable for the intelligent cone barrel.

[0073] When designing the point cloud registration formula, the weight coefficient of the point cloud is added on the basis of the ICP, therefore, the point cloud registration formula is represented as wherein, p i represents the i th point in the 3D point cloud, q i represents the corresponding point in the 3D model, R represents a rotation matrix, R includes a roll angle a, a pitch angle b and a yaw angle g, t represents a translation vector, N represents the number of points, and w iThe weight coefficient of the i-th point cloud is represented. It is solved by SVD (singular value decomposition), and when E is the minimum, the corresponding R and t are the optimal R and t.

[0074] In the point cloud registration process, changing the number of point clouds and the weight coefficient of the point cloud will affect the registration result. Therefore, in the embodiment, the weight coefficient w of the point cloud is set i The multi-scale error weighting weight can be set, and when the multi-scale error weighting weight is set, the corresponding relationship of the point cloud at different scales is set as different weights, so that the weight coefficient w of the point cloud i is reduced as the number of point clouds decreases.

[0075] In addition, the traditional ICP algorithm only considers the point coordinates in the point cloud, and does not consider the attitude data measured by the IMU, which may affect the registration accuracy. Therefore, the weight coefficient w of the point cloud is set i The sensor fusion weight can also be set, that is, the weight coefficient w of the point cloud is set i The weight w of the point cloud data collected by the roadside perception system vision The weight w of the intelligent cone barrel positioning data imu The sum of the weights of the point cloud data collected by the roadside perception system and the intelligent cone barrel positioning data is w i The weight coefficient of the point cloud is w vision +w imu .

[0076] In addition to setting the weight coefficient, a constraint condition such as an attitude change angle constraint and a convergence condition constraint is added to improve the accuracy and stability of the iteration.

[0077] The attitude change angle constraint is a limitation on the rotation matrix R. Due to the physical characteristics of the intelligent cone barrel, its attitude will only change within a small range, so in the point cloud registration process, the attitude range is narrowed by setting the attitude change angle constraint, so that the change range of the rotation angle is within a reasonable range, thereby improving the running efficiency of the algorithm.

[0078] Assuming that the rotation angle in the current ICP iteration is represented by Euler angles, the Euler angles (α, β, γ) correspond to the roll angle, pitch angle, and yaw angle, respectively. In order to ensure the rationality of the attitude, the constraint condition of the attitude angle is set as α min ≤α≤α max , β min ≤β≤β max , γ min ≤γ≤γ max , wherein α max , α min represent the upper and lower limits of the roll angle change, which can take values of plus or minus 30 degrees; β max , β min represent the upper and lower limits of the pitch angle change, which can take values of plus or minus 30 degrees; γ max, γ min respectively represent the upper and lower limits of the yaw angle change, and can take values of plus or minus 120.

[0079] Generally, the intelligent cone barrel is in a static state, so the convergence condition can be set as a static constraint, but when a strong wind or other extreme conditions cause the position of the intelligent cone barrel to change, the static constraint is not applicable. Therefore, in order to adapt to different scenarios, the present application sets the convergence condition to include static constraints and dynamic constraints. During the registration process, according to the state of the intelligent cone barrel, dynamic constraints or static constraints are selected to ensure that the point cloud registration is continuously effective.

[0080] wherein the dynamic constraint is represented as ∈ dynamic = ∈0+α·v, wherein ∈0represents a basic error threshold in a static scenario, α represents a weight factor adjusted according to dynamic changes in the scenario, and v represents the motion speed of the cone barrel, and ∈ dynamic represents the convergence error threshold.

[0081] When the intelligent cone barrel is in a static state, in order to comprehensively consider the changes in position and attitude, the static constraint is defined as a joint convergence condition, and the static constraint is represented as p error < ∈ position , r error < ∈ rotation ; wherein p error represents a position error threshold, ∈ position represents a position error threshold, t i,current represents the translation vector of the i-th point cloud in this iteration, t i,previous represents the translation vector of the i-th point cloud in the last iteration, r error represents an attitude error threshold, r error = |R current -R previous |, ∈ rotation represents an attitude error threshold, R current represents the rotation matrix in this iteration, R previous represents the rotation matrix in the last iteration. When the position error and the attitude error both meet the requirements, it is considered that the registration is completed, and the iteration is terminated.

[0082] After designing the point cloud registration algorithm, the point cloud registration of the 3D model and the 3D point cloud of the intelligent cone barrel can be performed in step 5. After the registration is completed, the calibration parameters of the roadside laser radar can be obtained, that is, the calculation of the calibration parameters of the roadside laser radar is completed.

[0083] When the point cloud registration is performed, the state of the intelligent cone barrel needs to be determined first. When the intelligent cone barrel is in a static state, the point cloud registration formula is used for registration, and the attitude change angle constraint and the static constraint are adopted in the registration process to ensure the registration accuracy and effectiveness. When the intelligent cone barrel is in a dynamic state, the attitude change angle constraint and the dynamic constraint are adopted in the registration process.

[0084] In addition, in order to improve the robustness of registration, noise and abnormal points need to be removed in each iteration registration. In the embodiment, the random sample consensus (RANSAC) method is used to filter abnormal points.

[0085] After obtaining the registration result, in step 6, the calibration accuracy index can be calculated according to the registration result to obtain the calibration error of the current sensor. Here, the calibration accuracy index at least includes the position accuracy and the attitude accuracy. The position accuracy and the calibration accuracy refer to the Euclidean distance and the Euler angle error before and after the registration of the point cloud to be measured (3D point cloud of the intelligent cone barrel).

[0086] In addition to calculating the calibration parameters and the calibration error, the positioning information uploaded by the intelligent cone barrel and the laser radar sensor data can also be used to monitor the accuracy of the roadside perception data. In addition to the perception data of the intelligent cone barrel, the roadside perception data also includes the perception data of other road users such as vehicles and pedestrians. The positioning data of the intelligent cone barrel is used to monitor the roadside perception data to determine the accuracy of the sensor. Therefore, the perception data of the intelligent cone barrel needs to be identified from all the roadside perception data, that is, target matching is performed. In the embodiment, the point-to-point matching relationship is determined by calculating the joint probability to complete the perception data accuracy monitoring, and the synthesis rule of the probability is based on the DS evidence theory.

[0087] When the perception data accuracy monitoring is performed, first, in step 7, the distance threshold δ d and the direction threshold δ h are set. In step 8, the distance difference and the direction difference between the roadside perception data and the positioning data of the intelligent cone barrel are calculated. In order to improve the calculation accuracy, the average value of the distance difference and the direction difference in a period of time is taken as the final distance difference and direction difference when the distance difference and the direction difference are calculated. Here, the distance difference is represented as The direction difference is represented as Where d i represents the distance difference between the roadside perception data and the positioning data of the intelligent cone barrel obtained by the i-th detection, h i represents the direction difference between the roadside perception data and the positioning data of the intelligent cone barrel obtained by the i-th detection, and n represents the number of detections.

[0088] In step 9, the distance activation value is calculated as where Dt denotes distance activation value, λ d denotes distance conversion factor, by setting distance conversion factor, data can be more dispersed, avoid data too concentrated when affecting the judgment result or cause error.

[0089] In step 10, according to the direction threshold and the direction difference, the direction activation value can be calculated, and the direction activation value is represented as Wherein, H t denotes distance activation value, λ h denotes direction conversion factor. Similarly, by setting the direction conversion factor, the data can be more dispersed, and the situation that the judgment result is affected or error is caused when the data is too concentrated is avoided.

[0090] In step 11, according to the distance activation value and the direction activation value, the corresponding matching score is calculated as Wherein, M represents the matching score, S(·) represents the Sigmoid activation function, and all data is mapped to 0 to 1 by using the activation function.

[0091] After obtaining the matching score of each point, in step 12, the perception data of the intelligent cone barrel is identified by using the matching score. When identifying the perception data of the intelligent cone barrel, the size relationship of the matching score and the matching threshold can be compared. When the matching score is greater than the matching threshold, the matching is successful, indicating that the current roadside perception data is the perception data of the intelligent cone barrel; otherwise, the matching is unsuccessful, indicating that the current roadside perception data is not the perception data of the intelligent cone barrel. It should be noted that the matching threshold can be set according to experience, and the present application does not make specific limitation.

[0092] After identifying the perception data of the intelligent cone barrel by the matching algorithm, in step 13, the precision index of the roadside perception data can be calculated. The precision index includes the accuracy of position and attitude, that is, the position error and the attitude error. The position error can be calculated by using the perception coordinates of the intelligent cone barrel and the spatial positioning coordinates, and the attitude error can be calculated according to the perception attitude of the intelligent cone barrel and the attitude positioning information.

[0093] The roadside laser radar calibration verification method based on the intelligent cone barrel provided by the present application is based on the positioning information uploaded by the intelligent cone barrel and the laser radar sensor data, and the calculation of the roadside laser radar calibration parameters is carried out, which eliminates the high cost of high-precision maps and is not restricted by the range of the map. Since the intelligent cone barrel itself is a feature in the laser radar point cloud, it is more convenient to find the feature points in the high-precision map in the laser radar point cloud, and the calibration accuracy is not lost.

[0094] The present application is based on the positioning information uploaded by the intelligent cone barrel and the laser radar sensor data, and real-time output of the current sensor calibration error is realized, calibration verification is completed, and since the intelligent cone barrel can send its own coordinates in real time, the time of the heterogeneous data is synchronized, and the problem of inaccurate calibration results caused by the movement of feature points in the existing calibration verification method is avoided.

[0095] In addition, the roadside laser radar calibration verification method based on the intelligent cone barrel provided by the present application can also monitor the roadside perception data based on the positioning information uploaded by the intelligent cone barrel and the laser radar sensor data, and since the intelligent cone barrel can be placed in the test site for a long time, it is not easy to affect the traffic flow, it is convenient to troubleshoot the error caused by the exposure time of the mechanical laser radar, and in the application scenario, it is more suitable for long-time monitoring and early warning of the roadside perception data than the true value car scheme.

[0096] Optionally, Figure 2 A flowchart for removing abnormal points provided by the embodiment of the present application is shown in Figure 1 and Figure 2 In step 5, the abnormal points are removed in the registration process, specifically: step 51, setting an abnormal threshold d threshold ; step 52, calculating the point cloud registration error according to the point cloud registration formula; step 53, judging whether the point cloud registration error is greater than the abnormal threshold, if yes, the corresponding point cloud is an abnormal point, and the point cloud is removed; if not, the corresponding point cloud is a normal point, and the point cloud is retained.

[0097] Specifically, please refer to Figure 2 When removing abnormal points, first, a consistency threshold is defined in step 51, that is, an abnormal threshold d threshold is set, and the point cloud whose registration error exceeds the threshold is regarded as an abnormal point and removed. Therefore, in step 52, the point cloud registration error is calculated according to the point cloud registration formula, and it should be noted that the point cloud registration error here is the registration result obtained by each iteration in step 5.

[0098] After obtaining the threshold and the registration error, step 53 can be used to judge whether the point cloud registration error is greater than the abnormal threshold, when it is greater, the corresponding point cloud is an abnormal point, and the point cloud is removed; otherwise, the corresponding point cloud is a normal point, and the point cloud is retained. The abnormal threshold d threshold in the embodiment of the present application can be set with noise level and prior knowledge, and the present application does not make specific limitation on this.

[0099] Those skilled in the art can understand that the drawings are only schematic diagrams of an embodiment, and the modules or processes in the drawings are not necessarily necessary for implementing the present application.

[0100] Those skilled in the art can understand that the modules in the device in the embodiments can be distributed in the device in the embodiments as described in the embodiments, or can be changed to be located in one or more devices different from the embodiments. The modules in the above embodiments can be combined into one module, or can be further split into multiple sub-modules.

[0101] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for calibrating and verifying roadside lidar based on intelligent cones, characterized in that, include: Collect smart cone positioning data and roadside perception system data; the smart cone positioning data includes the spatial positioning coordinates and attitude positioning information of the smart cone, and the roadside perception system data includes point cloud data and roadside perception data, and the roadside perception data includes at least the perception data of the smart cone; The 3D model of the intelligent cone is obtained based on its spatial positioning coordinates and attitude positioning information. The 3D point cloud of the smart cone is identified based on the point cloud data; Design a point cloud registration algorithm; The point cloud registration algorithm is used to register the 3D model and 3D point cloud of the smart cone to obtain the calibration parameters of the roadside lidar. The calibration accuracy index is calculated based on the registration results; the calibration accuracy index includes at least position accuracy and attitude accuracy. Set distance and direction thresholds; Calculate the distance difference and direction difference between the roadside sensing data and the smart cone positioning data; The distance activation value is calculated based on the distance threshold and the distance difference; The direction activation value is calculated based on the direction threshold and the direction difference; The corresponding matching score is calculated based on the distance activation value and the direction activation value; Based on the matching score, obtain the perception data of the smart cone; Based on the sensing data of the intelligent traffic cone, as well as the spatial positioning coordinates and attitude positioning information of the intelligent traffic cone, the accuracy index of the roadside sensing data is calculated.

2. The method for calibration and verification of roadside lidar based on intelligent cones according to claim 1, characterized in that, The point cloud registration algorithm includes a point cloud registration formula and constraints; wherein... The point cloud registration formula is expressed as follows: p i Let q represent the i-th point in the 3D point cloud. i Represents the corresponding point in the 3D model, R represents the rotation matrix, R includes roll angle α, pitch angle β and yaw angle γ, t represents the translation vector, and N represents the number of points; w i This represents the weight coefficient of the i-th point cloud; The constraints include attitude change angle constraints and convergence condition constraints; The attitude change angle constraint is represented by α. min ≤α≤α max β min ≤β≤β max γ min ≤γ≤γ max ; where α max α min These represent the upper and lower limits of the roll angle variation, β and β, respectively. max β min These represent the upper and lower limits of the pitch angle variation, γ and γ respectively. max γ min These represent the upper and lower limits of the yaw angle variation, respectively. The convergence constraints include dynamic constraints and static constraints, and the dynamic constraints are represented as ∈ dynamic =∈0+αv, where ∈0 represents the basic error threshold in a static scene, α represents the weighting factor adjusted according to the dynamic changes in the scene, and v represents the speed of the cone's movement. dynamic The static constraint is represented by p, which represents the convergence error threshold. error <∈ position r error <∈ rotation ; where p error Indicates the position error threshold. ∈ position t represents the position error threshold. i,current Let t represent the translation vector of the i-th point cloud in this iteration. i,previous Let r represent the translation vector of the i-th point cloud in the previous iteration. error R represents the attitude error threshold. error =|R current -R previous |,∈ rotation R represents the attitude error threshold. current Let R represent the rotation matrix for this iteration. previous This represents the rotation matrix from the previous iteration.

3. The method for calibration and verification of roadside lidar based on intelligent cones according to claim 2, characterized in that, The weight coefficient w of the point cloud i This includes multi-scale error weighting or sensor fusion weighting; The multi-scale error weighting refers to the weighting coefficients w of the point cloud. i It shrinks as the number of point clouds decreases; The sensor fusion weight refers to the weight w of the point cloud data collected by the roadside sensing system. vision Weight w of the smart cone positioning data imu The sum, denoted as w i =w vision +w imu .

4. The method for calibration and verification of roadside lidar based on intelligent cones according to claim 3, characterized in that, The point cloud registration algorithm is used to perform point cloud registration on the 3D model and 3D point cloud of the intelligent cone, specifically as follows: If the intelligent cone is stationary, determine whether it is. If so, based on the attitude change angle constraint and the static constraint, use the point cloud registration formula to perform point cloud registration on the 3D model and 3D point cloud of the intelligent cone, and remove outliers during the registration process. If not, based on the attitude change angle constraint and the dynamic constraint, use the point cloud registration formula to perform point cloud registration on the 3D model and 3D point cloud of the intelligent cone, and remove outliers during the registration process.

5. The method for calibration and verification of roadside lidar based on intelligent cones according to claim 4, characterized in that, Outliers are removed during the registration process, specifically: Set the exception threshold d threshold ; Calculate the point cloud registration error according to the point cloud registration formula; Determine whether the point cloud registration error is greater than the abnormal threshold. If it is, the corresponding point cloud is an abnormal point and is removed. If not, the corresponding point cloud is a normal point and is retained.

6. The method for calibration and verification of roadside lidar based on intelligent cones according to claim 1, characterized in that, The distance activation value is represented as Among them, D t δ represents the distance activation value. d d represents the distance threshold. i The distance difference between the roadside sensing data and the smart cone positioning data obtained in the i-th detection is represented by λ, where n represents the number of detections and λ represents the distance difference between the roadside sensing data and the smart cone positioning data obtained in the i-th detection. d This represents the distance conversion factor.

7. The method for calibration and verification of roadside lidar based on intelligent cones according to claim 6, characterized in that, The direction activation value is represented as Among them, H t Indicates the direction activation value, δ h h represents the direction threshold. i λ represents the directional difference between the roadside sensing data and the smart cone positioning data obtained in the i-th detection, n represents the number of detections, and λ represents the direction difference between them. h This represents the direction conversion factor.

8. The method for calibration and verification of roadside lidar based on intelligent cones according to claim 7, characterized in that, The matching score is represented as Where M represents the matching score and S(·) represents the Sigmoid activation function.

9. The method for calibration and verification of roadside lidar based on intelligent cones according to claim 1, characterized in that, Based on the matching score, the perception data of the smart cone is obtained, specifically: If the matching score is greater than the matching threshold, it indicates a successful match, and the corresponding roadside perception data is the perception data of the smart cone; if it is less than or equal to the threshold, it indicates a failed match, and the corresponding roadside perception data is not the perception data of the smart cone.

10. The method for calibration and verification of roadside lidar based on intelligent cones according to claim 1, characterized in that, The sensing data of the smart cone includes the sensing coordinates and sensing posture of the smart cone; the accuracy indicators include at least position error and posture error. Based on the sensing data of the intelligent traffic cone, and the spatial positioning coordinates and attitude positioning information of the intelligent traffic cone, the accuracy index of the roadside sensing data is calculated, specifically as follows: The position error is calculated based on the sensing coordinates and spatial positioning coordinates of the intelligent cone. The attitude error is calculated based on the perceived attitude and attitude positioning information of the intelligent cone.

Images