Laser radar noise point detection method and detection device, laser radar and medium

Abstract

The invention provides a laser radar noisy point detection method and device, a laser radar and a medium, the method is applied to the field of laser radar detection, and the method comprises the steps: obtaining a plurality of candidate decision values according to the feature parameters of a current data point and the feature parameters of M data points in a neighborhood window of the current data point; determining a target decision value of the current data point from the plurality of candidate decision values according to the feature parameters of the current data point and the feature parameters of the M data points; and determining whether the current data point is a noisy point or not according to the target judgment value. According to the method, the target decision value of the current data point can be determined from the plurality of candidate decision values through the feature parameters of the plurality of data points in the neighborhood window of the current data point, so that the accuracy of determining the target decision value is ensured. And whether the current data point is a noisy point is further judged in combination with the target judgment value, so that the noisy point in the point cloud can be accurately identified, and the detection accuracy of the laser radar is improved.

Description

Technical Field

[0001] The present application relates to the field of lidar detection, and more particularly, to a method and apparatus for detecting lidar noise points, a lidar, and a medium in the field of lidar detection. Background Art

[0002] Currently, with the continuous development of technology, in the field of detection, lidar has been widely used in the fields of intelligent robots, vehicle driverless, etc. due to its advantages of high resolution, good concealment, and strong anti-interference ability.

[0003] When the lidar detects real objects, if the rain and fog in the detection environment are relatively large or the dust is relatively dense, in addition to being reflected by real objects, the detection signals emitted by the lidar will also be reflected by rain and fog and dust, resulting in the presence of noise points such as rain and fog and dust in the point cloud data obtained by the lidar, which affects the detection effect and accuracy of the lidar.

[0004] In summary, in the process of lidar detection, how to accurately identify the noise points in the point cloud and improve the detection accuracy of the lidar has become an urgent problem to be solved. Summary of the Invention

[0005] The present application provides a method and apparatus for detecting lidar noise points, a lidar, and a medium. The method can determine the target determination value of the current data point from multiple candidate determination values through the characteristic parameters of multiple data points in the neighborhood window of the current data point, ensuring the accuracy of the determination of the target determination value. Further, by combining the target determination value to determine whether the current data point is a noise point, the noise points in the point cloud can be accurately identified, and the detection accuracy of the lidar can be improved.

[0006] In a first aspect, a method for detecting lidar noise points is provided. The method includes: obtaining multiple candidate determination values according to the characteristic parameters of the current data point and the characteristic parameters of M data points in the neighborhood window of the current data point; where the M data points are other data points in the neighborhood window of the current data point except the current data point, and M is an integer greater than 1; determining the target determination value of the current data point from the multiple candidate determination values according to the characteristic parameters of the current data point and the characteristic parameters of the M data points; and determining whether the current data point is a noise point according to the target determination value.

[0007] In combination with the first aspect, in some possible implementation manners, the feature parameter includes a distance value and a reflectivity feature value. Obtaining a plurality of candidate determination values according to the feature parameter of the current data point and the feature parameters of M data points within the neighborhood window of the current data point includes: obtaining a weighted distance difference set according to the distance value of the current data point, the distance values of the M data points, and the reflectivity feature values of the M data points; and obtaining the plurality of candidate determination values according to the weighted distance difference set.

[0008] In combination with the first aspect and the above implementation manners, in some possible implementation manners, obtaining a weighted distance difference set according to the distance value of the current data point, the distance values of the M data points, and the reflectivity feature values of the M data points includes: calculating the absolute value of the difference between the distance value of each of the M data points and the distance value of the current data point to obtain M absolute distance differences, forming a distance difference set; obtaining M weight coefficients according to the reflectivity feature values of the M data points, forming a weight coefficient set; and multiplying the absolute distance differences in the distance difference set by the corresponding weight coefficients in the weight coefficient set to obtain M weighted distance differences, forming the weighted distance difference set.

[0009] In combination with the first aspect and the above implementation manners, in some possible implementation manners, obtaining M weight coefficients according to the reflectivity feature values of the M data points, forming a weight coefficient set includes: when the reflectivity feature value Ref(i) of the i-th data point meets the high reflectivity requirement, calculating the weight coefficient as 0.5 / Ref(i) 1 / 2 ; when the reflectivity feature value Ref(i) of the i-th data point meets the low reflectivity requirement, calculating the weight coefficient as 1 / Ref(i) 1 / 2 .

[0010] In combination with the first aspect and the above implementation manners, in some possible implementation manners, obtaining the plurality of candidate determination values according to the weighted distance difference set includes: determining the average value of the X weighted distance differences with the smallest values in the weighted distance difference set as the first candidate determination value; determining the average value of the Y weighted distance differences with the smallest values in the weighted distance difference set as the second candidate determination value; determining the average value of the Z weighted distance differences with the smallest values in the weighted distance difference set as the third candidate determination value; where X < Y < Z ≤ M and X, Y, and Z are all positive integers.

[0011] Combined with the first aspect and the above implementation manners, in some possible implementation manners, determining the target determination value of the current data point from the multiple candidate determination values according to the characteristic parameters of the current data point and the characteristic parameters of the M data points includes: when the reflectivity characteristic values of at least two data points among the M data points meet the preset requirements, determining the target determination value as the first candidate determination value; when the reflectivity characteristic values of at least two data points among the M data points do not meet the preset requirements, determining the target determination value according to the characteristic parameters of the current data point.

[0012] Combined with the first aspect and the above implementation manners, in some possible implementation manners, the characteristic parameters further include a height value; determining the target determination value according to the characteristic parameters of the current data point includes: when the height value of the current data point is less than or equal to the ground height threshold, determining the target determination value as the first candidate determination value; when the height value of the current data point is greater than the ground height threshold, determining the target determination value according to the distance value of the current data point.

[0013] Combined with the first aspect and the above implementation manners, in some possible implementation manners, determining the target determination value according to the distance value of the current data point includes: when the distance value of the current data point is within the first distance interval, determining the target determination value as the second candidate determination value; when the distance value of the current data point is within the second distance interval, determining the target determination value as the third candidate determination value; wherein, the maximum value of the second distance interval is less than or equal to the minimum value of the first distance interval.

[0014] Combined with the first aspect and the above implementation manners, in some possible implementation manners, determining the target determination value according to the characteristic parameters of the current data point further includes: performing a histogram statistics on the height values of the M data points according to N height intervals to obtain N statistical values; determining the upper limit value of the height interval corresponding to the maximum value of the N statistical values as the ground height threshold; wherein, N is an integer greater than 1.

[0015] Combined with the first aspect and the above implementation manners, in some possible implementation manners, determining whether the current data point is a noise point according to the target determination value includes: when the target determination value is greater than or equal to the determination threshold, determining that the current data point is a noise point; when the target determination value is less than the determination threshold, determining that the current data point is not a noise point.

[0016] In a second aspect, a detection device for lidar noise points is provided. The device includes: an acquisition module configured to obtain a plurality of candidate determination values according to the characteristic parameters of a current data point and the characteristic parameters of M data points within a neighborhood window of the current data point; wherein the M data points are other data points within the neighborhood window of the current data point except the current data point, and M is an integer greater than 1; a target determination value determination module configured to determine a target determination value of the current data point from the plurality of candidate determination values according to the characteristic parameters of the current data point and the characteristic parameters of the M data points; and a noise point determination module configured to determine whether the current data point is a noise point according to the target determination value.

[0017] In a third aspect, a lidar is provided, including a memory and a processor. The memory is configured to store executable program code, and the processor is configured to call and run the executable program code from the memory, so that the lidar executes the method in the first aspect or any possible implementation manner of the first aspect.

[0018] In a fourth aspect, a computer program product is provided. The computer program product includes: computer program code, which when running on a computer, causes the computer to execute the method in the first aspect or any possible implementation manner of the first aspect.

[0019] In a fifth aspect, a computer-readable storage medium is provided. The computer-readable storage medium stores computer program code, which when running on a computer, causes the computer to execute the method in the first aspect or any possible implementation manner of the first aspect.

[0020] In the embodiments of the present application, first, a plurality of candidate determination values are obtained through the characteristic parameters of the current data point and the characteristic parameters of M data points within the neighborhood window of the current data point. Then, in combination with the characteristic parameters of the current data point and the characteristic parameters of the M data points, a target determination value of the current data point is determined from the plurality of candidate determination values. Selecting the target determination value corresponding to the current data point from the plurality of candidate determination values can ensure a high matching degree between the target determination value and the current data point, guaranteeing the accuracy and rationality of the determination of the target determination value. Further, it is determined whether the current data point is a noise point according to the target determination value, and screening and determination are performed according to the characteristics of the noise point distribution of rain, fog, or dust, effectively identifying the noise points in the point cloud and improving the detection accuracy of the lidar. Description of the Drawings

[0021] Figure 1 is a schematic diagram of a scene for lidar noise point detection provided by an embodiment of the present application;

[0022] Figure 2It is a schematic flowchart of a method for detecting noise points of a lidar provided by an embodiment of the present application;

[0023] Figure 3 It is a schematic diagram of a scenario for selecting data points within a neighborhood window provided by an embodiment of the present application;

[0024] Figure 4 It is a schematic flowchart of another method for detecting noise points of a lidar provided by an embodiment of the present application;

[0025] Figure 5 It is a schematic structural diagram of a device for detecting noise points of a lidar provided by an embodiment of the present application;

[0026] Figure 6 It is a schematic structural diagram of a lidar provided by an embodiment of the present application. Detailed implementation manners

[0027] Next, the technical solutions in the present application will be clearly and elaborately described with reference to the accompanying drawings. Among them, in the description of the embodiments of the present application, unless otherwise specified, " / " means "or". For example, A / B may represent A or B. The "and / or" in the text is only a description of the association relationship between associated objects, indicating that there can be three relationships. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of the present application, "a plurality of" means two or more than two.

[0028] Hereinafter, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as implying or suggesting relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features.

[0029] Before introducing the method of the embodiments of the present application, the professional terms that may be involved in the embodiments of the present application will be explained first.

[0030] Lidar: A radar system that emits laser beams to detect the position, speed and other characteristic quantities of a target. Its working principle is to emit a detection signal (laser beam or detection light) to the target object, and then, based on the received echo signal reflected from the target object, obtain the pose information of the target object, such as parameters like distance, azimuth, height, speed, attitude, and even shape.

[0031] Currently, when identifying noise points in the point cloud collected by a lidar in the related art, it is specifically implemented by the following Figure 1 method.

[0032] Figure 1It is a schematic diagram of the scenario for lidar noise detection provided by an embodiment of the present application.

[0033] Exemplarily, as Figure 1 shown, during the operation of the lidar, first, a detection signal is emitted. After encountering an object, the detection signal is reflected by the object, and then the lidar obtains an echo signal.

[0034] After obtaining the echo signal, the echo signal is subjected to detection processing. According to the sampling results of the detection, the lidar can obtain relevant information about the object.

[0035] Affected by weather factors, in addition to real objects in the air, there may also be impurity objects such as dust, rain, and fog. Specifically, in the embodiment of the present application, a real object refers to an object from which the lidar can collect valid data during the detection process, such as vehicles, pedestrians, curbs, the ground, road signs, buildings, etc. on the road surface. During the actual detection process of the lidar, the detection signal can be reflected by real objects on the one hand and by impurity objects on the other hand. In other words, the echo signal received by the lidar includes both the echo signal corresponding to the real object and the echo signal corresponding to the impurity object. As a result, when the lidar processes the echo signal, the obtained point cloud includes data points of real objects (hereinafter referred to as "real object points") and data points of impurity objects (hereinafter referred to as noise points). The noise points in the point cloud will interfere with subsequent object recognition and judgment based on the point cloud, resulting in an impact on the detection accuracy of the lidar.

[0036] As Figure 1 shown, to remove the noise points in the point cloud, the lidar can perform filtering during the detection process after receiving the echo signal to remove the echo signal corresponding to the impurity object in the echo signal.

[0037] In a possible implementation, it is achieved by setting a suitable detection threshold. For example, the detection threshold is set to threshold A, and the value of threshold A is relatively small. When the received echo signal is present, the echo signal intensity of the real object is relatively large and is usually greater than threshold A, and can be sampled; the echo signal intensity of the impurity object is relatively small, and the echo signal with an intensity less than threshold A will not be sampled. By setting a suitable detection threshold, the echo signal with a signal intensity less than threshold A is filtered out, so as to play a role in filtering out the echo signal of the impurity object.

[0038] However, when filtering and denoising using the above method, if the detection threshold is set too high, it is possible to cause missed detection of real objects, especially real objects at a long distance or real objects with a low reflectivity, whose echo signal intensity is small, resulting in the real object being erroneously deleted; if the detection threshold is set too low, the noise points cannot be effectively filtered out.

[0039] Based on the above problems, an embodiment of the present application proposes a method for detecting noise points of a lidar. This method can determine the target determination value of the current data point from multiple candidate determination values through the characteristic parameters of multiple data points within the neighborhood window of the current data point, ensuring the accuracy and rationality of the determination of the target determination value. Further, by combining the target determination value to determine whether the current data point is a noise point, the noise points in the point cloud can be accurately identified, improving the detection accuracy of the lidar.

[0040] Before introducing the method of the embodiment of the present application, the application scenario of the method of the embodiment of the present application will be introduced first.

[0041] It should be understood that for the noise points in the point cloud, they are generally densely distributed above the ground within a range where the distance value is less than a preset distance. Based on the distribution characteristics of the above noise points in the point cloud, for any data point in the point cloud, when determining whether the current data point is a noise point, the distance value of the current data point can be compared first. If the distance value of the current data point is greater than or equal to the preset distance, the lidar determines that the current data point is not a noise point; if the distance value of the current data point is less than the preset distance, the lidar determines that the current data point may be a noise point and further judgment of the current data point is required. Among them, the preset distance can be obtained through data statistics. For example, the preset distance can be 50m.

[0042] In the case where the current data point may be a noise point, the lidar can further determine whether the current data point is a noise point through the method of the embodiment of the present application. In other words, the method of the embodiment of the present application is specifically applicable to the data points in the point cloud whose distance values are less than the preset distance.

[0043] The following introduces a method for detecting noise points of a lidar provided by an embodiment of the present application.

[0044] Figure 2 is a schematic flowchart of a method for detecting noise points of a lidar provided by an embodiment of the present application.

[0045] Exemplarily, as Figure 2 shown, the method 200 includes the following steps 201 - step 203:

[0046] Step 201, obtain multiple candidate determination values according to the characteristic parameters of the current data point and the characteristic parameters of M data points within the neighborhood window of the current data point; where the M data points are other data points within the neighborhood window of the current data point except the current data point, and M is an integer greater than 1.

[0047] It should be understood that from the working principle of the lidar, during the detection process of the lidar, after the lidar completes a scanning cycle and scans the entire field of view, a frame of point cloud is obtained, and a frame of point cloud includes all the data points detected within the entire field of view.

[0048] The characteristic parameters of each data point may include three-dimensional coordinates, reflectivity eigenvalues, speed, grayscale information, etc. In the embodiments of the present application, the characteristic parameters of the data points mainly include three-dimensional coordinates and reflectivity eigenvalues. Among them, the three-dimensional coordinates of the data point refer to the three-dimensional coordinates of the data point in the lidar coordinate system. There are specifically various ways to establish the lidar coordinate system. For example, the lidar coordinate system may have the center of the scanning device that deflects and detects the light beam of the lidar as the coordinate origin O, the X-axis horizontally to the left, the Y-axis vertically upward, and the Z-axis along the forward direction of the lidar. In the lidar coordinate system, the coordinates of any data point i in the point cloud are denoted as (X i , Y i , Z i ). Among them, X i represents the distance between the data point i and the YOZ plane, that is, the horizontal value of the data point i; Y i represents the distance between the data point i and the XOZ plane, that is, the height value of the data point i; Z i represents the distance between the data point i and the XOY plane, that is, the distance value calculated based on the flight time between the emission moment of the detection signal and the reception moment of the echo signal of the data point i. The reflectivity eigenvalue refers to the reflectivity of the object detected by the lidar, denoted as Ref(i). In the embodiments of the present application, the reflectivity eigenvalues of the noise points meet the low reflectivity requirements, and the reflectivity eigenvalues of the real object points meet the high reflectivity requirements.

[0049] When determining whether the current data point is a noise point, the lidar may use the current data point as the center point and select M data points within the neighborhood window. The M data points are the other data points within the neighborhood window except the current data point, and M is an integer greater than 1.

[0050] Optionally, in the embodiments of the present application, the size of the neighborhood window is 5*5, and it can also be adjusted according to the actual situation. The embodiments of the present application do not limit the size of the neighborhood window, and the size of the neighborhood window is taken as 5*5 in the embodiments of the present application for illustration.

[0051] Figure 3 is a schematic diagram of a scenario for selecting data points within the neighborhood window provided by the embodiments of the present application.

[0052] Exemplarily, as Figure 3 shows, the current frame of the point cloud includes multiple data points, and each data point is numbered in the order from left to right and from top to bottom. In the current frame of the point cloud, the row number of any data point i is m, and the column number is n.

[0053] The current data point R is the 38th data point among the multiple data points, the row number m = 3, and the column number n = 8.

[0054] When selecting the neighborhood window of the current data point, the lidar can take the current data point R as the center and select multiple data points within a 5*5 neighborhood window, that is, 25 data points included from the 1st row to the 5th row and from the 6th column to the 10th column, where M = 24.

[0055] As Figure 3 shown, the 24 data points are sequentially denoted as "A1, A2,..., A24" from left to right and from top to bottom.

[0056] Furthermore, the lidar can determine multiple candidate decision values based on the characteristic parameters of the current data point and the characteristic parameters of the 24 data points.

[0057] When the lidar calculates multiple candidate decision values, the characteristic parameters mainly include the distance value and the reflectivity characteristic value.

[0058] In a possible implementation manner, multiple candidate decision values are obtained according to the characteristic parameters of the current data point and the characteristic parameters of the M data points within the neighborhood window of the current data point, including step 2011 - step 2012 below.

[0059] Step 2011, obtain a weighted distance difference set according to the distance value of the current data point, the distance values of the M data points, and the reflectivity characteristic values of the M data points.

[0060] Step 2012, obtain multiple candidate decision values according to the weighted distance difference set.

[0061] When calculating the weighted distance difference, if the current data point is a real object point, the higher the similarity between the current data point and the data points in the neighborhood, the smaller the obtained weighted distance difference; on the contrary, if the current data point is a noise point, the lower the similarity between the current data point and the data points in the neighborhood, the larger the obtained weighted distance difference. In the embodiments of the present application, by calculating the weighted distance difference, the difference between each data point in the neighborhood and the current data point can be reflected.

[0062] In order to take into account the numerical characteristics of the distance value and the reflectivity characteristic value, during the process of obtaining the weighted distance difference set, first determine the absolute values of M distance differences, and then use the reflectivity characteristic value as the weight coefficient to obtain the weighted distance difference set according to the absolute values of the distance differences and the corresponding weight coefficients. In a possible implementation manner, obtaining the weighted distance difference set according to the distance value of the current data point, the distance values of the M data points, and the reflectivity characteristic values of the M data points includes step 20111 - step 20113 below.

[0063] Step 20111, calculate the absolute values of the differences between the distance values of the M data points and the distance value of the current data point respectively to obtain the absolute values of M distance differences, forming a distance difference set.

[0064] Exemplarily, taking Figure 3 the current data point R shown in the figure and the other 24 data points within the neighborhood window opened with the current data point R as the center as an example for illustration. In this example, M = 24. Denote the distance value of the current data point as Z R , and denote the 24 distance values corresponding to the 24 data points as Z A1 , Z A2 , Z A3 ,.....Z A24 . Calculate the absolute values of the differences between the 24 distance values and Z R , obtaining 24 absolute values of distance differences, and denote the 24 absolute values of distance differences as Z (A1)_dif , Z (A2)_dif , Z (A3)_dif ......, Z (A24)_dif . The 24 absolute values of distance differences form a set of distance differences, and denote the set of distance differences as Z dif .

[0065] Specifically, the set of distance differences Z dif can be represented by the following expression (1):

[0066] Z dif = {Z (A1)_dif , Z (A2)_dif , Z (A3)_dif ,......, Z (A24)_dif} (1)

[0067] Through the above process, the lidar can obtain the set of distance differences Z dif .

[0068] Step 20112: Obtain M weight coefficients based on the reflectivity eigenvalue of M data points, and form a set of weight coefficients.

[0069] After obtaining the set of distance differences, considering the influence of the reflectivity eigenvalue on data point detection, based on the reflectivity eigenvalue of each of the 24 data points, obtain 24 sets of weight coefficients.

[0070] In a possible implementation manner, obtaining M weight coefficients based on the reflectivity eigenvalue of M data points and forming a set of weight coefficients can be obtained through the following steps.

[0071] When the reflectivity eigenvalue Ref(i) of the i-th data point meets the high reflectivity requirement, calculate the weight coefficient as 0.5 / Ref(i) 1 / 2 . When the reflectivity eigenvalue Ref(i) of the i-th data point meets the low reflectivity requirement, calculate the weight coefficient as 1 / Ref(i) 1 / 2 .

[0072] The lidar can sample the echo signal using multiple detection thresholds, such as including a first threshold and a second threshold, where the first threshold is greater than the second threshold, that is, the first threshold is a high threshold and the second threshold is a low threshold. When the intensity of the echo signal is relatively large, it can be sampled by both the first threshold and the second threshold at the same time; at this time, the probability that the echo signal is returned after being reflected by a real object is greater, and the output reflectivity eigenvalue is an even number, meeting the high reflectivity requirement. When the intensity of the echo signal is relatively small, it can be sampled by the second threshold but cannot be sampled by the first threshold; at this time, the probability that the echo signal is returned after being reflected by an impurity object is greater, and the output reflectivity eigenvalue is an odd number, meeting the low reflectivity requirement.

[0073] The reflectivity eigenvalue can be obtained from the reflectivity value output after the lidar samples the echo signal using two detection thresholds. Since the lidar originally outputs a binary reflectivity code, converting the binary reflectivity code to a decimal reflectivity code is the reflectivity eigenvalue.

[0074] Therefore, based on whether the reflectivity eigenvalue of each data point among 24 data points is specifically an odd value or an even value, the lidar can determine the weight coefficient of each data point. Specifically, the weight coefficient of a data point can be expressed by the following expression (2).

[0075]

[0076] Among them, in expression (2):

[0077] W (i) : Among 24 data points, the weight coefficient of any data point i, as Figure 3 shown, i takes values from A1 - A24;

[0078] Ref(i): The reflectivity eigenvalue of data point i.

[0079] Through the above process, the lidar can obtain the weight coefficient corresponding to each of the 24 data points, and the 24 weight coefficients form a weight coefficient set, denoted as W.

[0080] Specifically, the weight coefficient set W can be expressed by the following expression (3).

[0081] W = {W (A1) , W (A2) , W (A3) ,......, W (A24)} (3)

[0082] Through the above process, the lidar can obtain the weight coefficient set W.

[0083] Step 20113: Multiply the absolute value of the distance difference in the distance difference set by the corresponding weight coefficient in the weight coefficient set to obtain M weighted distance differences, forming a weighted distance difference set.

[0084] Specifically, when determining the weighted distance difference set according to the distance difference set and the weight coefficient set, for any one of the 24 data points, first determine the absolute value of the distance difference corresponding to this data point from the distance difference set, and determine the weight coefficient corresponding to this data point from the weight coefficient set. Then multiply the absolute value of the distance difference corresponding to this data point and the weight coefficient corresponding to this data point to obtain the weighted distance difference corresponding to this data point. Thus, 24 weighted distance differences corresponding to 24 data points are obtained, forming the weighted distance difference set.

[0085] Exemplarily, as Figure 3 shown, for data point A1 among the 24 data points, in the distance difference set Z dif , the lidar can determine that the absolute value of the distance difference corresponding to data point A1 is Z (A1)_dif . In the weight coefficient set W, the lidar can determine that the weight coefficient corresponding to data point A1 is W (A1) . Multiply the above absolute value of the distance difference Z (A1)_dif and the weight coefficient W (A1) to obtain the weighted distance difference corresponding to data point A1, denoted as Q (A1) .

[0086] By analogy, the determination process of the weighted distance differences corresponding to other data points is exactly the same as that of the weighted distance difference of data point A1, and will not be elaborated here.

[0087] Through the above process, the lidar can obtain the weighted distance difference set, denoted as Q.

[0088] Specifically, the weighted distance difference set Q can be expressed by the following expression (4).

[0089] Q = {Q (A1) , Q (A2) , Q (A3) ,......, Q (A24)} (4)

[0090] According to the elements included in the weighted distance difference set, and according to the different numbers of candidate determination values, multiple candidate determination values may include a first candidate determination value, a second candidate determination value, and a third candidate determination value.

[0091] In a possible implementation, according to the weighted distance difference set, multiple candidate determination values are obtained, including steps 20121 to 20123 below.

[0092] Step 20121: Determine the average value of the X weighted distance differences with the smallest values in the weighted distance difference set as the first candidate determination value.

[0093] Step 20122: Determine the average value of the Y weighted distance differences with the smallest values in the weighted distance difference set as the second candidate determination value.

[0094] Step 20123: Determine the average value of the Z weighted distance differences with the smallest values in the weighted distance difference set as the third candidate determination value, where X < Y < Z ≤ M and X, Y, and Z are all positive integers.

[0095] Optionally, in the embodiments of the present application, X = 3, Y = 6, and Z = 15.

[0096] Exemplarily, in the process of determining the first candidate determination value, the second candidate determination value, and the third candidate determination value, for the 24 weighted distance differences in the weighted distance difference set, the lidar can sort the 24 weighted distance differences in ascending order of value to obtain the sorted weighted distance difference set Q'.

[0097] Among them, the first candidate determination value refers to the average value of the 3 weighted distance differences with the smallest values in the weighted distance difference set Q'. When the weighted distance differences in the weighted distance difference set Q' are sorted in ascending order, take the average value of the first 3 weighted distance differences in Q' as the first candidate determination value, denoted as "E1"; the second candidate determination value refers to the average value of the first 6 weighted distance differences in Q'; the third candidate determination value refers to the average value of the first 15 weighted distance differences in Q'.

[0098] E1, E2, and E3 can be respectively represented by the following expression (5).

[0099]

[0100] Another exemplarily, the lidar can also not sort the 24 weighted distance differences, directly compare the values in size, respectively select 3 weighted distance differences with smaller values, 6 weighted distance differences with smaller values, and 15 weighted distance differences with smaller values from the 24 weighted distance differences, and respectively calculate the average values to obtain the first candidate determination value, the second candidate determination value, and the third candidate determination value.

[0101] So far, the process of step 201 for determining multiple candidate determination values ends.

[0102] As can be seen from the calculation processes of the foregoing multiple candidate determination values, the first candidate determination value is the smallest, and the third candidate determination value is the largest.

[0103] The smaller the distance difference between the M data points in the neighborhood and the current data point, the smaller the corresponding weighted distance difference; at the same time, when the echo signal intensity is large, the reflectivity eigenvalue is an even number, and the weight coefficient is also smaller. If the current data point corresponds to a real object, it is very likely that the M data points in its neighborhood also correspond to the same real object, and the M weighted distance differences calculated therefrom are generally small. On the contrary, if the current data point corresponds to an impurity object, the correlation between the M data points in its neighborhood and the current data point is poor, and the M weighted distance differences calculated therefrom are generally large.

[0104] Further, the lidar can determine the target determination value of the current data point from multiple candidate determination values through the following step 202.

[0105] Step 202: Determine the target determination value of the current data point from multiple candidate determination values according to the characteristic parameters of the current data point and the characteristic parameters of the M data points.

[0106] Specifically, when determining the target determination value according to the characteristic parameters of the current data point and the characteristic parameters of the M data points, based on the characteristics of the characteristic parameters corresponding to the real object points and noise points in the point cloud, and combining the point cloud characteristics of rain, fog or dust noise points, the lidar can determine the matching target determination value through the characteristic parameters of the current data point and the M data points in the neighborhood.

[0107] In a possible implementation manner, determining the target determination value of the current data point from multiple candidate determination values according to the characteristic parameters of the current data point and the characteristic parameters of the M data points includes the following steps 2021 to 2022.

[0108] Step 2021: When the reflectivity eigenvalues of at least two data points among the M data points meet the preset requirements, determine the target determination value as the first candidate determination value.

[0109] Among them, the reflectivity eigenvalue meeting the preset requirements means that the reflectivity eigenvalue is an even value greater than zero.

[0110] As can be seen from the foregoing, when the reflectivity eigenvalue is an even number, the probability that the echo signal is returned after being reflected by a real object is relatively high. If the current data point corresponds to a real object, it is very likely that the M data points in its neighborhood also correspond to the same real object, that is, the correlation between the current data point and the M data points in the neighborhood is strong. When the reflectivity eigenvalues of at least 2 of the 24 data points are all even numbers, it indicates that the probability that the current data point is a real object point is relatively high, and the possibility that the current data point is a noise point is very low. At this time, select the first candidate determination value E1 with the weakest constraint as the target determination value E.

[0111] Step 2022: When the reflectivity eigenvalue of at least two data points among the M data points does not meet the preset requirements, determine the target determination value according to the characteristic parameters of the current data point.

[0112] For the case where the probability that the current data point is not a real object point is high, more judgments need to be continued. The lidar needs to combine the characteristic parameters of the current data point to determine the target determination value of the current data point.

[0113] Optionally, in addition to the distance value and the reflectivity eigenvalue, when determining the target determination value, the characteristic parameters further include the height value, that is, the Y of the current data point in the lidar coordinate system (i) 。

[0114] In a possible implementation, determining the target determination value according to the characteristic parameters of the current data point includes the following steps 20221 - step 20222.

[0115] Step 20221: When the height value of the current data point is less than or equal to the ground height threshold, determine the target determination value as the first candidate determination value.

[0116] It should be understood that the ground height threshold can be the pre - calibrated reference ground height. However, the pre - calibrated reference ground height may not match the actual scene, such as the scenes where the ground height changes like uphill and downhill. The ground height threshold can also be calculated according to the eigenvalue of the data point. The lidar can perform statistics on the height values of the M data points according to different height intervals, and determine the ground height threshold according to the height interval where the data points are most concentrated.

[0117] Specifically, the determination of the ground height threshold can include the following steps.

[0118] First, perform histogram statistics on the height values of the M data points in N height intervals to obtain N statistical values, where N is an integer greater than 1.

[0119] Continuing with the previous example for illustration, for the 24 height values corresponding to 24 data points, N height intervals can be set. For example, the N height intervals can be [a, b), [b, c), [c, d). Based on the preset 3 height intervals, perform histogram statistics on the height values of the 24 data points, and thus obtain the 3 distribution quantities corresponding to the 24 height values in the 3 height intervals, that is, obtain 3 statistical values.

[0120] Then, determine the upper limit value of the height interval corresponding to the maximum value of the N statistical values as the ground height threshold.

[0121] After obtaining N statistical values, the lidar can determine the maximum value from them, that is, determine which height interval among the N height intervals corresponds to the largest number of data points. Since the ground height is continuous and occupies a large range in the field of view, there are many detected ground data points and the height values are concentrated. By statistically analyzing the height values of the data points in the neighborhood, the height interval with the largest number of distributed data points can be identified, and the upper limit value of the height interval is determined as the ground height threshold. Exemplarily, if among 3 height intervals, the statistical value of 24 data points in the height interval [b, c) is the largest, the lidar can determine the height value c as the ground height threshold.

[0122] After determining the ground height threshold, the lidar can compare the height value of the current data point with the ground height threshold.

[0123] It should be understood that since the noise points are the data points corresponding to impurities such as rain, fog, and dust. In the environment, rain, fog, and dust are mostly fine particles, with small volume and light weight, and have a small terminal velocity in the air. Affected by air convection, this type of impurity is more likely to float in the air relative to real objects. Therefore, the height value of the noise points is greater than the height value of the ground.

[0124] Therefore, when the height value of the current point is less than or equal to the ground height threshold, it indicates that the current data point most likely corresponds to the echo signal returned by the ground. The current data point is a ground point, not a noise point. When the current data point is likely to be a real object point, the first determination value E1 is selected as the target determination value E. Exemplarily, if the ground height threshold is c and the height value of the current data point is less than or equal to c, the lidar determines E1 as the target determination value.

[0125] Step 20222, when the height value of the current data point is greater than the ground height threshold, determine the target determination value according to the distance value of the current data point.

[0126] On the other hand, when the height value of the current point is greater than the ground height threshold, more judgments need to be continued.

[0127] Since the distance values of the noise points of rain, fog, or dust are usually concentrated above the ground with a distance value less than 20m, and there are only sporadic scattered points distributed within the distance range of 20 - 50m at most, the lidar can further combine the above distribution characteristics of the distance values of the noise points to determine the target determination value.

[0128] When the distance value of the current data point is within the first distance interval, determine the target determination value as the second candidate determination value. When the distance value of the current data point is within the second distance interval, determine the target determination value as the third candidate determination value, and the maximum value of the second distance interval is less than or equal to the minimum value of the first distance interval.

[0129] Optionally, the first distance interval is [20, 50), and the second distance interval is (0, 20). Exemplarily, the distance value of the current data point is 30m, which is within the first distance interval. Considering the characteristic that there are sporadic scattered points in the noise points at most within the range of the distance value of 20 - 50m in the above description, it shows that the possibility that the current data point is a noise point is relatively low. Considering the previous judgment result, the possibility that the current data point is a noise point is medium, and the lidar can determine the second candidate judgment value E2 as the target judgment value E.

[0130] Exemplarily, the distance value of the current data point is 10m, which is within the second distance interval. Considering the characteristic that the noise points are concentrated above the ground with a distance value less than 20m in the above description, it shows that the possibility that the current data point is a noise point is high. The lidar can determine the third candidate judgment value E3 as the target judgment value E.

[0131] So far, step 202 ends.

[0132] Through the above step 202, the lidar can determine the target judgment value of the current data point from multiple candidate judgment values to be used to judge whether the current data point is a noise point. Through the eigenvalue law of the data points in the neighborhood, for the case where the current data point is likely to be a real object and less likely to be a noise point, the first candidate judgment value with the weakest constraint is determined as the target judgment value; for the case where the eigenvalue of the data points in the neighborhood does not meet the real object judgment condition, the height value of the current data point is further compared. For the case where the current data point is likely to be a ground point, the first candidate judgment value with the weakest constraint is also determined as the target judgment value; if neither of the above two cases is satisfied, the possibility that the current data point is a noise point is relatively high. On this basis, the distance of the current data point is further compared. Considering the distance distribution characteristics of the point clouds of rain, fog, and dust, for the current data point that does not conform to the distribution characteristics, the second candidate judgment value with medium constraint is determined as the target judgment value, and for the current data point that conforms to the distribution characteristics, the third candidate judgment value with the strongest constraint is determined as the target judgment value.

[0133] Step 203, according to the target judgment value, determine whether the current data point is a noise point.

[0134] Specifically, after obtaining the target judgment value of the current data point, the lidar determines whether the current data point is a noise point by comparing the target judgment value with the judgment threshold. The judgment threshold is preset and can be determined according to the candidate judgment values calculated from the detected points of known real objects and the candidate judgment values calculated from known noise points, to determine the judgment threshold E'.

[0135] Determining whether the current data point is a noise point according to the target judgment value includes:

[0136] When the target determination value is greater than or equal to the determination threshold, it is determined that the current data point is a noise point. When the target determination value is less than the determination threshold, it is determined that the current data point is not a noise point.

[0137] Specifically, when the target determination value E is greater than or equal to E′, the lidar can determine that the current data point is a noise point. When the target determination value E is less than E′, the lidar can determine that the current data point is a valid point such as a real object point or a ground line.

[0138] Through the above steps 201 - step 203, the lidar can implement the process of noise point detection.

[0139] As can be seen from the foregoing, if the current detection point is a detection point of a real object, the correlation of the characteristic parameters of the detection points in its neighborhood is good, and the overall M weighted distance differences are relatively small, and thus the multiple candidate determination values calculated therefrom are also small. When determining the target determination value from multiple candidate determination values, it is also the first candidate determination value with the weakest constraint that is used as the target determination value. When the final target determination value is compared with the determination threshold, it is determined not to be a noise point. On the contrary, if the current detection point is a noise point, the correlation of the characteristic parameters of the detection points in its neighborhood is poor, and the overall M weighted distance differences are relatively large, and thus the multiple candidate determination values calculated therefrom are also relatively large. When determining the target determination value from multiple candidate determination values, since the characteristic parameter characteristics of the current point conform to the data characteristics of the point clouds of rain, fog, and dust, the second candidate determination value with medium constraint or the third candidate determination value with the strongest constraint is determined as the target determination value. When the final target determination value is compared with the determination threshold, it is determined to be a noise point.

[0140] In summary, in the embodiments of the present application, first, multiple candidate determination values are obtained through the characteristic parameters of the current data point and the characteristic parameters of M data points in the neighborhood window of the current data point. Then, in combination with the characteristic parameters of the current data point and the characteristic parameters of the M data points, the target determination value of the current data point is determined from multiple candidate determination values. Selecting the target determination value corresponding to the current data point from multiple candidate determination values can ensure a high matching degree between the target determination value and the current data point, and ensure the accuracy and reasonableness of the determination of the target determination value. Further, it is determined whether the current data point is a noise point through the target determination value, and screening and judgment are performed according to the characteristics of the noise distribution of rain, fog, or dust, effectively identifying the noise points in the point cloud. For the detected points identified as noise points, the lidar can remove them and output the denoised point cloud, which is convenient for the backend to perform clustering and recognition of objects based on the output point cloud and perform the next processing, improving the detection accuracy of the lidar.

[0141] To facilitate understanding of the solution provided in the embodiments of the present application, the following Figure 4 is used to introduce in detail the method for detecting noise points of the lidar provided in the embodiments of the present application.

[0142] Figure 4 It is a schematic flowchart of another method for detecting noise points of a lidar provided by an embodiment of the present application.

[0143] Exemplarily, as Figure 4 shown, this method 400 includes the following steps 401-step 415:

[0144] Step 401, determine whether the distance value of the current data point is greater than a preset distance.

[0145] Based on the distribution characteristics of the distance values of the noise points, when identifying the current data point, the lidar can first determine whether the distance value of the current data point is greater than a preset distance.

[0146] When the distance value of the current data point is greater than the preset distance, directly execute step 415, that is, determine that the current data point is a real object point.

[0147] When the distance value of the current data point is less than or equal to the preset distance, it indicates that the current data point may be a noise point, and continue to execute step 402.

[0148] Step 402, determine the characteristic parameters of M data points within the neighborhood window centered on the current data point.

[0149] Step 403, according to the distance value of the current data point and the distance values of the M data points, determine the absolute values of M distance differences to form a distance difference set.

[0150] Step 404, according to the reflectivity characteristic values of the M data points, determine M weight coefficients to form a weight coefficient set.

[0151] Step 405, according to the distance difference set and the weight coefficient set, determine a weighted distance difference set.

[0152] Step 406, according to the weighted distance difference set, determine a plurality of candidate determination values, and the plurality of candidate determination values include a first candidate determination value, a second candidate determination value, and a third candidate determination value.

[0153] Steps 402-step 406 have the same inventive concept as step 201 in method 200. For specific reference, see the description in step 201, and details are not repeated here.

[0154] Step 407, determine the number of data points among the M data points that meet the preset requirements.

[0155] When the number of data points that meet the preset requirements is greater than or equal to 2, execute step 408.

[0156] When the number of data points that meet the preset requirements is less than 2, execute step 409.

[0157] Step 408, determine that the target determination value is the first candidate determination value.

[0158] Step 409, whether the height value of the current data point is less than or equal to the ground height threshold.

[0159] When the height value of the current data point is less than or equal to the ground height threshold, return to Step 408.

[0160] When the height value of the current data point is greater than the ground height threshold, execute Step 410.

[0161] Step 410, determine the distance interval where the distance value of the current data point is located.

[0162] When the distance interval where the distance value of the current data point is located is the first distance interval (0 - 20m), execute Step 412.

[0163] When the distance interval where the distance value of the current data point is located is the second distance interval (20 - 50), execute Step 411.

[0164] Step 411, determine that the target determination value is the second candidate determination value.

[0165] Step 412, determine that the target determination value is the third candidate determination value.

[0166] The above Steps 407 - 412 have the same inventive concept as Step 202 in Method 200. For specific descriptions, please refer to the description in Step 202 and will not be elaborated here.

[0167] Step 413, determine whether the target determination value is less than the determination threshold.

[0168] When the target determination value is less than the determination threshold, execute Step 415.

[0169] When the target determination value is greater than or equal to the determination threshold, execute Step 414.

[0170] Step 414, determine that the current data point is a noise point.

[0171] Step 415, determine that the current data point is a real object point.

[0172] Steps 413 - 415 have the same inventive concept as Step 203 in Method 200. For specific descriptions, please refer to the description in Step 203 and will not be elaborated here.

[0173] Figure 5 It is a schematic structural diagram of a lidar noise detection device provided by an embodiment of the present application.

[0174] Exemplarily, as Figure 5 shown, the device 500 includes:

[0175] An acquisition module 501, configured to obtain a plurality of candidate determination values according to the feature parameters of the current data point and the feature parameters of M data points within the neighborhood window of the current data point; wherein, the M data points are other data points within the neighborhood window of the current data point except the current data point, and M is an integer greater than 1;

[0176] A target determination value determination module 502, configured to determine the target determination value of the current data point from the plurality of candidate determination values according to the feature parameters of the current data point and the feature parameters of the M data points;

[0177] A noise point determination module 503, configured to determine whether the current data point is a noise point according to the target determination value.

[0178] In a possible implementation manner, the feature parameters include a distance value and a reflectivity feature value, and the acquisition module 501 is specifically configured to: obtain a weighted distance difference set according to the distance value of the current data point, the distance values of the M data points, and the reflectivity feature values of the M data points; and obtain the plurality of candidate determination values according to the weighted distance difference set.

[0179] In a possible implementation manner, the acquisition module 501 is further configured to: calculate the absolute values of the differences between the distance values of the M data points and the distance value of the current data point respectively to obtain M absolute distance differences, which form a distance difference set; obtain M weight coefficients according to the reflectivity feature values of the M data points, which form a weight coefficient set; multiply the absolute distance differences in the distance difference set by the corresponding weight coefficients in the weight coefficient set to obtain M weighted distance differences, which form the weighted distance difference set.

[0180] In a possible implementation manner, the acquisition module 501 is further configured to: when the reflectivity feature value Ref(i) of the i-th data point meets the high reflectivity requirement, calculate the weight coefficient as 0.5 / Ref(i) 1 / 2 ; when the reflectivity feature value Ref(i) of the i-th data point meets the low reflectivity requirement, calculate the weight coefficient as 1 / Ref(i) 1 / 2 .

[0181] In a possible implementation, the obtaining module 501 is further configured to: determine the average value of the X weighted distance differences with the smallest values in the weighted distance difference set as the first candidate determination value; determine the average value of the Y weighted distance differences with the smallest values in the weighted distance difference set as the second candidate determination value; determine the average value of the Z weighted distance differences with the smallest values in the weighted distance difference set as the third candidate determination value; where X < Y < Z ≤ M and X, Y, and Z are all positive integers.

[0182] In a possible implementation, the target determination value determining module 502 is specifically configured to: when the reflectivity eigenvalue of at least two data points among the M data points meets the preset requirement, determine the target determination value as the first candidate determination value; when the reflectivity eigenvalue of at least two data points among the M data points does not meet the preset requirement, determine the target determination value according to the feature parameters of the current data point.

[0183] In a possible implementation, the feature parameter further includes a height value; the target determination value determining module 502 is further configured to: when the height value of the current data point is less than or equal to the ground height threshold, determine the target determination value as the first candidate determination value; when the height value of the current data point is greater than the ground height threshold, determine the target determination value according to the distance value of the current data point.

[0184] In a possible implementation, the target determination value determining module 502 is further configured to: when the distance value of the current data point is within the first distance interval, determine the target determination value as the second candidate determination value; when the distance value of the current data point is within the second distance interval, determine the target determination value as the third candidate determination value; where the maximum value of the second distance interval is less than or equal to the minimum value of the first distance interval.

[0185] In a possible implementation, the target determination value determining module 502 is further configured to: perform a histogram statistics on the height values of the M data points in N height intervals to obtain N statistical values; determine the upper limit value of the height interval corresponding to the maximum value of the N statistical values as the ground height threshold; where N is an integer greater than 1.

[0186] In a possible implementation, the noise point determining module 503 is specifically configured to: when the target determination value is greater than or equal to the determination threshold, determine that the current data point is a noise point; when the target determination value is less than the determination threshold, determine that the current data point is not a noise point.

[0187] Figure 6 It is a schematic structural diagram of a lidar provided by an embodiment of the present application.

[0188] Exemplarily, such as Figure 6As shown in the figure, the lidar 600 includes: a memory 601 and a processor 602. Among them, an executable program code 6011 is stored in the memory 601, and the processor 602 is used to call and execute the executable program code 6011 to execute a method for detecting lidar noise points.

[0189] In this embodiment, the lidar can be divided into functional modules according to the above method examples. For example, each functional module can be corresponding, or two or more functions can be integrated into one processing module. The above integrated module can be implemented in the form of hardware. It should be noted that the division of modules in this embodiment is illustrative, only a logical function division, and there can be other division methods in actual implementation.

[0190] In the case of dividing each functional module according to each function, the lidar can include: an acquisition module, a determination module, etc. It should be noted that all relevant contents of each step involved in the above method embodiment can be cited in the function description of the corresponding functional module, and will not be repeated here.

[0191] The lidar provided in this embodiment is used to execute the above method for detecting lidar noise points, so it can achieve the same effect as the above implementation method.

[0192] In the case of adopting an integrated unit, the lidar can include a processing module and a storage module. Among them, the processing module can be used to control and manage the actions of the lidar. The storage module can be used to support the lidar to execute mutual program codes and data, etc.

[0193] Among them, the processing module can be a processor or a controller, which can implement or execute various exemplary logical blocks, modules, and circuits described in combination with the disclosure of the present application. The processor can also be a combination that realizes computing functions, such as a combination of one or more microprocessors, a combination of digital signal processing (DSP) and a microprocessor, etc. The storage module can be a memory.

[0194] This embodiment also provides a computer-readable storage medium. Computer program code is stored in the computer-readable storage medium. When the computer program code runs on a computer, it causes the computer to execute the above-related method steps to implement a method for detecting lidar noise points in the above embodiment.

[0195] This embodiment also provides a computer program product. When the computer program product runs on a computer, it causes the computer to execute the above-related steps to implement a method for detecting lidar noise points in the above embodiment.

[0196] In addition, the lidar provided by the embodiments of the present application may specifically be a chip, component or module, and the lidar may include a connected processor and a memory; wherein, the memory is used to store instructions, and when the lidar runs, the processor can call and execute the instructions to enable the chip to execute a method for detecting lidar noise points in the above embodiments.

[0197] Among them, the lidar, computer-readable storage medium, computer program product or chip provided by this embodiment are all used to execute the corresponding methods provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding methods provided above, and will not be elaborated here.

[0198] Through the description of the above embodiments, those skilled in the art can understand that for the convenience and brevity of description, only the above-mentioned division of each functional module is used as an example. In actual applications, the above functions can be allocated to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above.

[0199] In the embodiments provided in the present application, it should be understood that the disclosed device and method can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the devices or units can be in electrical, mechanical or other forms.

[0200] The above content is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for detecting noise points of a lidar, characterized in that, The method includes: Obtaining a plurality of candidate determination values according to the characteristic parameters of the current data point and the characteristic parameters of M data points within the neighborhood window of the current data point; wherein, the M data points are other data points within the neighborhood window of the current data point except the current data point, and M is an integer greater than 1; Determining the target determination value of the current data point from the plurality of candidate determination values according to the characteristic parameters of the current data point and the characteristic parameters of the M data points; Determining whether the current data point is a noise point according to the target determination value.

2. The detection method according to claim 1, wherein The characteristic parameters include a distance value and a reflectivity characteristic value, and the obtaining a plurality of candidate determination values according to the characteristic parameters of the current data point and the characteristic parameters of M data points within the neighborhood window of the current data point includes: Obtaining a weighted distance difference set according to the distance value of the current data point, the distance values of the M data points, and the reflectivity characteristic values of the M data points; Obtaining the plurality of candidate determination values according to the weighted distance difference set.

3. The detection method according to claim 2, wherein The obtaining a weighted distance difference set according to the distance value of the current data point, the distance values of the M data points, and the reflectivity characteristic values of the M data points includes: Calculating the absolute value of the difference between the distance value of each of the M data points and the distance value of the current data point to obtain M absolute distance differences, forming a distance difference set; Obtaining M weight coefficients according to the reflectivity characteristic values of the M data points, forming a weight coefficient set; Multiplying the absolute distance differences in the distance difference set by the corresponding weight coefficients in the weight coefficient set to obtain M weighted distance differences, forming the weighted distance difference set.

4. The detection method according to claim 3, wherein The obtaining M weight coefficients according to the reflectivity characteristic values of the M data points, forming a weight coefficient set includes: When the reflectivity eigenvalue Ref(i) of the i-th data point meets the high reflectivity requirement, calculate the weight coefficient as 0.5 / Ref(i) 1 / 2 ; When the reflectivity eigenvalue Ref(i) of the i-th data point meets the low reflectivity requirement, calculate the weight coefficient as 1 / Ref(i) 1 / 2 .

5. The detection method according to claim 2, wherein The obtaining the plurality of candidate determination values according to the weighted distance difference set includes: Determining the average value of the X weighted distance differences with the smallest values in the weighted distance difference set as the first candidate determination value; Determining the average value of the Y weighted distance differences with the smallest values in the weighted distance difference set as the second candidate determination value; Determining the average value of the Z weighted distance differences with the smallest values in the weighted distance difference set as the third candidate determination value; wherein, X < Y < Z ≤ M and X, Y, Z are all positive integers.

6. The detection method according to claim 5, wherein the determining the target determination value of the current data point from the plurality of candidate determination values according to the characteristic parameters of the current data point and the characteristic parameters of the M data points includes: When the reflectivity characteristic values of at least two data points among the M data points meet the preset requirements, determining the target determination value as the first candidate determination value; When the reflectivity characteristic values of at least two data points among the M data points do not meet the preset requirements, determining the target determination value according to the characteristic parameters of the current data point.

7. The detection method according to claim 6, wherein The characteristic parameters further include a height value; the determining the target determination value according to the characteristic parameters of the current data point includes: When the height value of the current data point is less than or equal to the ground height threshold, determine the target determination value as the first candidate determination value; When the height value of the current data point is greater than the ground height threshold, determine the target determination value according to the distance value of the current data point.

8. The detection method according to claim 7, characterized in that, The determining the target determination value according to the distance value of the current data point includes: When the distance value of the current data point is within the first distance interval, determine the target determination value as the second candidate determination value; When the distance value of the current data point is within the second distance interval, determine the target determination value as the third candidate determination value; Wherein, the maximum value of the second distance interval is less than or equal to the minimum value of the first distance interval.

9. The detection method according to claim 7, wherein The determining the target determination value according to the characteristic parameters of the current data point further includes: Performing histogram statistics on the height values of the M data points according to N height intervals to obtain N statistical values; Determine the upper limit value of the height interval corresponding to the maximum value of the N statistical values as the ground height threshold; Wherein, N is an integer greater than 1.

10. The detection method according to claim 1, wherein The determining whether the current data point is a noise point according to the target determination value includes: When the target determination value is greater than or equal to the determination threshold, determine that the current data point is a noise point; When the target determination value is less than the determination threshold, determine that the current data point is not a noise point.

11. A detection device for lidar noise points, characterized in that, The device includes: An acquisition module, configured to acquire a plurality of candidate determination values according to the characteristic parameters of the current data point and the characteristic parameters of M data points within the neighborhood window of the current data point; wherein, the M data points are other data points within the neighborhood window of the current data point except the current data point, and M is an integer greater than 1; A target determination value determination module, configured to determine the target determination value of the current data point from the plurality of candidate determination values according to the characteristic parameters of the current data point and the characteristic parameters of the M data points; A noise point determination module, configured to determine whether the current data point is a noise point according to the target determination value.

12. A lidar, characterized in that, The lidar includes: A memory, configured to store executable program code; A processor, configured to call and run the executable program code from the memory, so that the lidar executes the method according to any one of claims 1 to 10.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and when the computer program is executed, the method according to any one of claims 1 to 10 is implemented.

Images