Noise filtering method, electronic equipment, readable storage medium and vehicle

Abstract

The invention provides a noisy point filtering method, electronic equipment, a readable storage medium and a vehicle, and belongs to the technical field of intelligent automobiles, and the noisy point filtering method comprises the steps that point cloud data of a laser radar are acquired, the point cloud data comprise three-dimensional coordinates and target parameters of multiple points, and the target parameters comprise echo intensity and / or object reflectivity; determining noisy points in the point cloud data according to the three-dimensional coordinates and target parameters of the plurality of points; and filtering the noisy points in the point cloud data. According to the method of comprehensively identifying the noisy points according to the multiple features, the identification accuracy of the noisy points can be effectively improved, and then the noisy point filtering effect is improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent automobile technology, and in particular to a noise filtering method, an electronic device, a readable storage medium and a vehicle. Background Art

[0002] With the continuous development of science and technology, laser radar in the detection field has been widely used in scenarios such as intelligent robots and unmanned vehicles due to its advantages of high resolution, good concealment, and strong anti-interference ability. In vehicle application scenarios, dust will be generated during the driving process of the vehicle. The laser radar will produce dust noise due to the echo of dust, which will affect the accuracy of the data collected by the laser radar, and then have an adverse effect on the subsequent processing and analysis of the laser radar data.

[0003] At present, for noise points such as dust noise, rain and fog noise, the randomness of noise points is usually used to identify points that suddenly appear in adjacent multi-frame point cloud data according to their positions, and filter them out as noise points to improve the accuracy of LiDAR data. However, the method of identifying noise points by the single feature of the position of the noise points has the problem of low noise point identification accuracy, which in turn leads to low accuracy of noise point filtering based on this identification method. Summary of the invention

[0004] The embodiments of the present application provide a noise filtering method, an electronic device, a readable storage medium and a vehicle to solve the problem of low accuracy of current noise filtering.

[0005] In a first aspect, an embodiment of the present application provides a noise filtering method, comprising:

[0006] Acquire point cloud data of a laser radar, wherein the point cloud data includes three-dimensional coordinates of a plurality of points and target parameters, wherein the target parameters include echo intensity and / or object reflectivity;

[0007] Determining noise points in the point cloud data according to the three-dimensional coordinates of the plurality of points and the target parameters;

[0008] The noise points in the point cloud data are filtered out.

[0009] Optionally, determining the noise points in the point cloud data according to the three-dimensional coordinates and target parameters of the plurality of points includes:

[0010] According to the three-dimensional coordinates and target parameters of the multiple points, the discreteness of each point in the neighborhood area corresponding to each point is analyzed and processed, and the points corresponding to the neighborhood area with a discreteness greater than a first critical discreteness are determined as noise points in the point cloud data.

[0011] Optionally, analyzing and processing the discreteness of each point in a neighborhood area corresponding to each point according to the three-dimensional coordinates of the multiple points and the target parameters includes:

[0012] According to the three-dimensional coordinates and target parameters of each point, adjusting the aggregation degree of each point in the neighborhood area corresponding to each point;

[0013] The discreteness of each point in the adjusted neighborhood area corresponding to each point is analyzed and processed.

[0014] Optionally, adjusting the aggregation degree of each point in the neighborhood area corresponding to each point according to the three-dimensional coordinates and the target parameters of each point includes:

[0015] For each of the points, determining the absolute value of the distance difference between each of the target points and the point according to the three-dimensional coordinates of the point and the three-dimensional coordinates of each target point in the neighborhood area corresponding to the point, the target point being a point in the neighborhood area;

[0016] The target parameter of each target point is used to adjust the absolute value of the distance difference of each target point.

[0017] Optionally, analyzing and processing the discreteness of each point in the adjusted neighborhood area corresponding to each point, and determining the point corresponding to the neighborhood area having a discreteness greater than a first critical discreteness as a noise point in the point cloud data, includes:

[0018] For each of the points, determine, in a neighborhood area corresponding to the point, a target point whose adjusted absolute value of the distance difference is less than a first distance threshold as a target reference point;

[0019] When the proportion of the target reference points in the neighborhood area corresponding to the point is less than a first ratio threshold, the point is determined to be a noise point.

[0020] Optionally, the first distance threshold is a parameter used to determine whether the target point in the neighborhood area is a target reference point, and the first distance threshold corresponding to each point is positively correlated with the distance value of the point.

[0021] Optionally, the target parameter is the echo intensity; and the step of using the target parameter of each target point to adjust the absolute value of the distance difference of each target point includes:

[0022] Calculating the spatial attenuation weight of each target point according to the echo intensity of each target point and the distance coordinate value in the three-dimensional coordinates, wherein the spatial attenuation weight is positively correlated with the echo attenuation degree per unit distance;

[0023] The spatial attenuation weight of each target point is used to adjust the absolute value of the distance difference of each target point.

[0024] Optionally, the adopting of the spatial attenuation weight of each target point to adjust the absolute value of the distance difference of each target point includes: scaling the absolute value of the distance difference of each target point proportionally according to the spatial attenuation weight of each target point to obtain the adjusted absolute value of the distance difference of each target point.

[0025] Optionally, the target parameter is the reflectivity of the object; and the step of using the target parameter of each target point to adjust the absolute value of the distance difference of each target point includes:

[0026] Performing reverse amplification processing on the object reflectivity of each target point to obtain a reflection weight of each target point;

[0027] The reflection weight of each target point is used to adjust the absolute value of the distance difference of each target point.

[0028] Optionally, performing reverse amplification processing on the object reflectivity of each target point to obtain the reflection weight of each target point includes:

[0029] Setting the reflection weight of the target point whose object reflectivity is greater than or equal to the reflectivity threshold to 0;

[0030] The reflection weight of the target point whose reflectivity of the object is less than the reflectivity threshold is set to a value greater than 0.

[0031] Optionally, using the reflection weight of each target point to adjust the absolute value of the distance difference of each target point includes: scaling the absolute value of the distance difference of each target point proportionally according to the reflection weight of each target point to obtain the adjusted absolute value of the distance difference of each target point.

[0032] Optionally, determining the noise points in the point cloud data according to the three-dimensional coordinates and target parameters of the plurality of points further includes:

[0033] Determine points corresponding to the neighborhood area whose discreteness is less than or equal to the critical discreteness as valid points in the point cloud data;

[0034] According to the three-dimensional coordinates of the multiple points and the reflectivity of the object, the discreteness of each point in the neighborhood area corresponding to each valid point is analyzed and processed again, and the points corresponding to the neighborhood area with a discreteness greater than a second critical discreteness are determined as noise points in the point cloud data.

[0035] Optionally, determining the noise points in the point cloud data according to the three-dimensional coordinates and target parameters of the plurality of points further includes:

[0036] Determine points corresponding to the neighborhood area whose discreteness is less than or equal to the critical discreteness as valid points in the point cloud data;

[0037] According to the three-dimensional coordinates and echo intensities of the multiple points, the discreteness of each point in the neighborhood area corresponding to each valid point is analyzed and processed again, and the points corresponding to the neighborhood area with a discreteness greater than a second critical discreteness are determined as noise points in the point cloud data.

[0038] Optionally, determining the noise points in the point cloud data according to the three-dimensional coordinates and target parameters of the plurality of points further includes:

[0039] According to the collection order of each of the points in the point cloud data, each point in the neighborhood area corresponding to each of the points is determined.

[0040] Optionally, determining the noise points in the point cloud data according to the three-dimensional coordinates and target parameters of the plurality of points includes:

[0041] According to the distance value between each point and the laser radar in the three-dimensional coordinates, the point whose distance value is less than the maximum distance threshold is determined as a suspected noise point in the point cloud data;

[0042] Noise points among the suspected noise points are determined according to the three-dimensional coordinates of the multiple points and the target parameters.

[0043] In a second aspect, an embodiment of the present application provides a noise filtering device, comprising:

[0044] An acquisition module, used to acquire point cloud data of the laser radar, wherein the point cloud data includes three-dimensional coordinates of a plurality of points and target parameters, wherein the target parameters include echo intensity and / or object reflectivity;

[0045] A determination module, configured to determine noise points in the point cloud data according to the three-dimensional coordinates of the plurality of points and target parameters;

[0046] A filtering module is used to filter out the noise points in the point cloud data.

[0047] In a third aspect, an embodiment of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the noise filtering method described in any one of the first aspects when executing the computer program.

[0048] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the noise filtering method as described in any one of the first aspects of claim.

[0049] In a fifth aspect, an embodiment of the present application provides a vehicle, comprising the electronic device described in the third aspect, or the computer-readable storage medium described in the fourth aspect.

[0050] In a sixth aspect, an embodiment of the present application provides a computer program product, which includes computer instructions. When the computer instructions are executed on an electronic device, the electronic device executes the noise filtering method as described in any one of the first aspects of claim 1.

[0051] The noise filtering method, electronic device, readable storage medium and vehicle provided in the embodiments of the present application obtain the point cloud data of the laser radar to determine the noise points in the point cloud data according to the three-dimensional coordinates and target parameters of multiple points in the point cloud data, thereby filtering out the noise points in the point cloud data. Among them, the target parameters may include echo intensity and / or object reflectivity. Since the technical solution of the present application can comprehensively identify noise points based on at least two features of the points in the point cloud data. Therefore, compared with the solution of using a single feature for noise identification in the related art, it can effectively improve the recognition accuracy of noise points in the point cloud data, thereby improving the accuracy of noise filtering. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for describing the embodiments are briefly introduced below.

[0053] Figure 1 A schematic diagram of imaging of a laser radar provided in an embodiment of the present application;

[0054] Figure 2 A flowchart of a noise filtering method provided in an embodiment of the present application;

[0055] Figure 3 A scanning schematic diagram of a laser radar provided in an embodiment of the present application;

[0056] Figure 4 A flowchart of a noise filtering method provided in one embodiment of the present application;

[0057] Figure 5 A flowchart of a noise filtering method provided in another embodiment of the present application;

[0058] Figure 6 A flowchart of a noise filtering method provided in yet another embodiment of the present application;

[0059] Figure 7 A flowchart of a noise filtering method provided in yet another embodiment of the present application;

[0060] Figure 8 A flowchart of another noise filtering method provided in an embodiment of the present application;

[0061] Fig. 9 A noise filtering effect diagram provided in an embodiment of the present application;

[0062] Fig.10 A schematic diagram of a noise filtering device provided in an embodiment of the present application;

[0063] Fig.11 A block diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0064] The exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can also be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0065] In order to facilitate understanding of the technical solution of the present application, the principles involved in the technology of the present application are explained as follows.

[0066] Cluster noise points, such as vehicle dust, rain and fog, usually have the characteristics of relatively high point cloud density, low object reflectivity, and higher point cloud discreteness compared to real objects.

[0067] Among them, the relatively high point cloud density means that noise points such as vehicle dust, rain and fog usually appear in clusters, which makes them higher in density than sunlight noise points, raindrop noise points, etc., but relatively lower in density than real objects and ground lines. Since the density of cluster noise points is closer to the density of real objects, in an environment with cluster noise points such as vehicle dust, rain and fog, the difference in distance values ​​of each point cloud point is relatively small. Therefore, the difference in distance values ​​of each point cloud point can be magnified to distinguish between lower-density cluster noise points and higher-density real object points in the point cloud data.

[0068] Low object reflectivity means that noise points such as vehicle dust, rain and fog have low reflectivity compared to real objects. In this way, the reflectivity of a point cloud point can reflect to a certain extent whether the point cloud point is a noise point or a real object point. Therefore, the object reflectivity of the point cloud point can be used to amplify the distance value difference of each point cloud point to make the distance value difference of each real object point smaller; the distance value difference of each noise point such as vehicle dust, rain and fog is larger, which is more conducive to distinguishing between clumping noise points and real object points in point cloud data.

[0069] The higher degree of point cloud dispersion than real objects means that the number of noise points such as vehicle dust, rain and fog per unit area is less than the number of real object points per unit area. In this way, the point cloud data can be processed by region, and the number of point cloud points with small distance differences in a single region can be used to determine the degree of dispersion of the point cloud points in the region, and then distinguish between clumping noise points and real object points in the point cloud data.

[0070] For example, please refer to Figure 1 , which shows the effect image of the point cloud data obtained by the laser radar scanning the vehicle dust environment in the embodiment of the present application. Figure 1 As shown, the side of the vehicle has dust, which makes the side of the vehicle present a typical clustered object feature in the point cloud data. In addition, it can be clearly seen that the dust cluster noise point 101 has a high discrete feature, a low reflectivity feature, and a relatively high density feature. It should be noted that Figure 1 The brighter the color, the lower the reflectivity of the object.

[0071] Please refer to Figure 2 , which shows a flow chart of a noise filtering method provided by an embodiment of the present application. The noise filtering method can be applied to electronic devices. Optionally, the electronic device can be a vehicle computer, a wearable device, etc. Figure 2 As shown, the noise filtering methods include:

[0072] Step 201: Obtain point cloud data of a laser radar. The point cloud data includes three-dimensional coordinates of multiple points and target parameters.

[0073] In the embodiment of the present application, the point cloud data may include the three-dimensional coordinates and target parameters of each point (i.e., point cloud point). Among them, in the three-dimensional coordinates, the distance coordinate value in the direction (usually the z-axis direction) from the object point reflecting the laser to the laser radar is the distance value of the point.

[0074] The target reference includes echo intensity and / or object reflectivity. Among them, object reflectivity refers to the ratio of the radiant energy reflected from the surface of the object to the total radiant energy projected onto the surface of the object. It is an inherent property of the object, and different objects have different object reflectivities. Echo intensity refers to the intensity of the echo signal returned after the laser pulse of the lidar is reflected by the surface of the object, which can reflect the reflective characteristics of the surface of the object. When scanning with the same lidar, different objects have different echo intensities due to their different object reflectivities.

[0075] In an optional implementation, the laser radar may be a one-dimensional rotating mirror laser radar, which acquires point cloud data by line scanning. The acquisition order of the one-dimensional rotating mirror laser radar is: scanning by column, and dividing each column of points into x line scanning areas, each line scanning area including y points. The one-dimensional rotating mirror laser radar sequentially acquires data of each point in the x line scanning areas for each column of points, so that after the acquisition of the x line scanning areas is completed, the point cloud data of the column is obtained, and then after the acquisition of the x line scanning areas in each column is completed, a complete frame of point cloud data is obtained.

[0076] For example, Figure 3 As shown in the figure, taking x as 4 and y as 24 as an example, each column includes 96 points accordingly. The acquisition order of the one-dimensional rotating mirror laser radar is as follows: for the first column, after collecting the data of each point in the 4 line scanning areas in turn, the point cloud data of 96 points in the first column is obtained. Then, for the second column, after collecting the data of each point in the 4 line scanning areas in turn, the point cloud data of 96 points in the second column is obtained. And so on, until for the last column, after collecting the data of each point in the 4 line scanning areas in turn, a complete frame of point cloud data is obtained.

[0077] Under the acquisition order of the one-dimensional rotating mirror laser radar (i.e., the structure of the point cloud data), it is not difficult to obtain that the left neighboring point adjacent to the left side of each point in the entire frame is a point 95 points away from the point in the forward direction, arranged in the acquisition order in the point cloud data. Similarly, the left neighboring point adjacent to the right side of each point is a point 95 points away from the point in the backward direction, arranged in the acquisition order in the point cloud data. The upper neighboring point adjacent to the upper side of each point is the point before the point, arranged in the acquisition order in the point cloud data. The lower neighboring point adjacent to the lower side of each point is the point after the point, arranged in the acquisition order in the point cloud data.

[0078] Step 202: Determine noise points in the point cloud data according to the three-dimensional coordinates of the multiple points and the target parameters.

[0079] In some implementations of the present application, the noise points in the point cloud data can be determined based on the three-dimensional coordinates and target parameters of multiple points, combined with at least one of the following characteristics: relatively high location cloud density, low object reflectivity, and higher point cloud discreteness compared to real objects.

[0080] In an optional implementation, the process of the electronic device determining the noise points in the point cloud data according to the three-dimensional coordinates of the plurality of points and the target parameters may include step S0.

[0081] In step S10, based on the three-dimensional coordinates and target parameters of multiple points, the discrete degrees of each point in the neighborhood area corresponding to each point are analyzed and processed, and the points corresponding to the neighborhood area with a discrete degree greater than a first critical discrete degree are determined as noise points in the point cloud data.

[0082] Among them, the neighborhood area of ​​a point, also known as the neighborhood window, refers to an area of ​​a certain range centered on the point. The area includes the point and points near it. Points near the point are also called neighborhood points of the point. Optionally, the neighborhood area of ​​a point can be a p-row × q-column area. The p-row × q-column area refers to all points within the p-row and q-column range centered on the point. For example, the neighborhood area is a 3-row × 5-column area with a total of 15 points. Alternatively, the neighborhood area is a 5-row × 7-column area with a total of 35 points, etc.

[0083] Optionally, the electronic device may be provided with a first critical discrete degree. Furthermore, for each point in the point cloud data, the discrete degrees of each point in the neighborhood area corresponding to the single point are analyzed and processed according to the three-dimensional coordinates and target parameters of multiple points in the point cloud data to obtain the discrete degree corresponding to the single point. The discrete degree corresponding to each point in the point cloud data is determined by analogy. Since the discrete degree corresponding to a single point indicates the discrete degree of all points distributed in the neighborhood area of ​​the point. Therefore, combined with the feature that the point cloud discrete degree of noise points is higher than that of real objects, the points among the multiple points of the point cloud data whose discrete degree is greater than that corresponding to the first critical discrete degree can be determined as noise points in the point cloud data.

[0084] In some embodiments of the present application, the process of the electronic device analyzing and processing the degree of discreteness of each point in the neighborhood area corresponding to each point according to the three-dimensional coordinates of the multiple points and the target parameters may include steps S20 to S21.

[0085] In step S20, the aggregation degree of each point in the neighborhood area corresponding to each point is adjusted according to the three-dimensional coordinates of each point and the target parameters.

[0086] In step S21, the discreteness of each point in the adjusted neighborhood area corresponding to each point is analyzed and processed.

[0087] Optionally, for each point in the point cloud data, the electronic device can adjust the degree of aggregation of each point in the neighborhood area corresponding to the single point according to the three-dimensional coordinates and target parameters of multiple points in the point cloud data, and obtain the adjusted neighborhood area corresponding to the single point. Then, the discrete degree of each point in the adjusted neighborhood area corresponding to the single point is analyzed and processed to obtain the discrete degree corresponding to the single point, and then the discrete degree corresponding to each point in the point cloud data is determined. And so on to determine the discrete degree corresponding to each point in the point cloud data. Then, the electronic device can determine the point whose discrete degree is greater than the first critical discrete degree as a noise point in the point cloud data.

[0088] Since noise points have the characteristics of higher point cloud dispersion than real objects, they also have the characteristics of relatively high point cloud density. The point cloud density of real objects is also relatively high. For example, Figure 1 The point cloud density of the dust cluster noise point 101 is relatively close to the point cloud density of the vehicle. Therefore, in order to effectively distinguish between noise points and real object points, the aggregation degree of each point in the neighborhood area corresponding to each point in the point cloud data can be adjusted to amplify the point cloud density difference between the noise points and the real object points in the point cloud data, thereby facilitating the improvement of the analysis accuracy of the discrete degree of all points in the neighborhood area corresponding to each point.

[0089] In an optional implementation, the process in which the electronic device adjusts the degree of aggregation of each point in a neighborhood area corresponding to each point according to the three-dimensional coordinates of each point and the target parameters may include steps S30 and S31.

[0090] In step S30, for each point, the absolute value of the distance difference between each target point and the point is determined based on the three-dimensional coordinates of the point and the three-dimensional coordinates of each target point in the neighborhood area corresponding to the point.

[0091] In step S31, the target parameter of each target point is used to adjust the absolute value of the distance difference of each target point.

[0092] The target point corresponding to a point refers to a point in the neighborhood area of ​​the point. It is not difficult to understand that the target point of a single neighborhood area can be considered as a general term for all points included in the neighborhood area. The electronic device can sequentially perform steps S30 and S31 for each point to adjust the absolute value of the distance difference of each target point in the domain area corresponding to a single point in the point cloud data, thereby adjusting the absolute value of the distance difference of the target point in the domain area corresponding to each point in the point cloud data.

[0093] For ease of understanding, the following explanation introduces the definition of the current point, and each point in the point cloud data can be considered as the current point in turn. In this way, it can be understood that the electronic device considers a point in the point cloud data as the current point, and executes steps S30 and S31 for the current point to adjust the absolute value of the distance difference of each target point in the field area corresponding to the current point. Furthermore, the electronic device considers the next point in the point cloud data as the current point, and executes steps S30 and S31 for the current point to adjust the absolute value of the distance difference of each target point in the field area corresponding to the current point. By analogy, the electronic device can adjust the absolute value of the distance difference of the target point in the field area corresponding to each point in the point cloud data.

[0094] Optionally, the electronic device can determine each target point in the neighborhood of the current point, and obtain the distance value of the target point from the three-dimensional coordinates of each target point in the point cloud data. For each target point, the absolute value of the difference between the distance value of the target point and the distance value of the current point is calculated to obtain the absolute value of the distance difference of the target point, and then the absolute value of the distance difference between each target point and the current point is obtained.

[0095] For example, the electronic device can calculate a distance difference matrix based on the distance value of each target point in the neighborhood area of ​​the current point. The distance difference matrix includes the absolute value of the distance difference of each target point in the neighborhood area, and the position of the absolute value of the distance difference of the target point in the distance difference matrix is ​​the position of the target point in the neighborhood area.

[0096] In some embodiments of the present application, noise points in point cloud data are determined based on the three-dimensional coordinates and target parameters of multiple points, and the method further includes: determining each point in the neighborhood area corresponding to each point according to the acquisition order of each point in the point cloud data. Based on this, for the current point, the electronic device can first determine each target point in the neighborhood area of ​​the current point according to the acquisition order of each point in the point cloud data. Then, based on the distance value of each target point in the neighborhood area of ​​the current point, the absolute value of the distance difference between each target point and the current point is determined.

[0097] Optionally, when the laser radar is a one-dimensional rotating mirror laser radar, and the neighborhood area of ​​the point is an area of ​​p rows × q columns, the electronic device can determine, for the current point, the points in the point cloud data that are (x×y)×i-1 points forward, (x×y)×i points forward, (x×y)×i-2 points forward, the current point, the point before the current point, the point after the current point, and the points that are (x×y)×i-1 points backward, (x×y)×i points backward, and (x×y)×i-2 points backward from the current point as each target point in the neighborhood area of ​​the current point. Where i is a positive integer, and

[0098] For example, it is assumed that the neighborhood area of ​​a point is an area of ​​3 rows×5 columns, and the value of i is 1 and 2. Also, it is assumed that each column of the point cloud is 4×24 points (ie, 96 points).

[0099] For the current point, the electronic device can forward the point in the point cloud data that is 96×1-1 points away from the current point (the 2nd row and 2nd column in the neighborhood area), the point that is 96×1 points away from the current point (the 1st row and 2nd column in the neighborhood area), the point that is 96×1-2 points away from the current point (the 3rd row and 2nd column in the neighborhood area), the point that is 96×2-1 points away from the current point (the 2nd row and 1st column in the neighborhood area), the point that is 96×2 points away from the current point (the 1st row and 1st column in the neighborhood area), the point that is 96×2-2 points away from the current point (the 3rd row and 1st column in the neighborhood area), and the current point (the 2nd row and 3rd column in the neighborhood area). The previous point (the 1st row and 3rd column in the neighborhood area), the point after the current point (the 3rd row and 3rd column in the neighborhood area), and the point 96×1-1 points backward from the current point (the 2nd row and 4th column in the neighborhood area), 96×1 points forward (the 1st row and 4th column in the neighborhood area), 96×1-2 points forward (the 3rd row and 4th column in the neighborhood area), 96×2-1 points forward (the 2nd row and 5th column in the neighborhood area), 96×2 points forward (the 1st row and 5th column in the neighborhood area), and 96×2-2 points forward (the 3rd row and 5th column in the neighborhood area).

[0100] Of course, in some embodiments, the electronic device can also determine each target point in the neighborhood area of ​​each point according to the three-dimensional coordinates of each point. The embodiment of the present application does not limit the method of determining the target point in the neighborhood area.

[0101] Optionally, the electronic device can determine each target point in the neighborhood of the current point for the current point, and obtain the distance value of the target point from the three-dimensional coordinates of each target point in the point cloud data. For each target point, the absolute value of the difference between the distance value of the target point and the distance value of the current point is calculated to obtain the absolute value of the distance difference of the target point, and the target parameter of the target point is used to adjust the absolute value of the distance difference of the target point to obtain the adjusted absolute value of the distance difference of the target point, thereby obtaining the adjusted absolute value of the distance difference of each target point.

[0102] The process of adjusting the absolute value of the distance difference of each target point by using the target parameter of each target point may include: scaling the absolute value of the distance difference of each target point according to the target parameter of each target point to obtain the adjusted absolute value of the distance difference of each target point. The scaling ratio of the absolute value of the distance difference of each target point may be determined according to the speed at which the target parameter of the target point changes with the distance value. For example, the faster the target parameter of the target point changes with the distance value, the higher the scaling ratio of the absolute value of the distance difference of the target point.

[0103] In some embodiments of the present application, the electronic device analyzes and processes the degree of discreteness of each point in the adjusted neighborhood area corresponding to each point, and determines the points corresponding to the neighborhood area with a degree of discreteness greater than a first critical discreteness as noise points in the point cloud data. The process may include steps S40 to S41.

[0104] In step S40, for each point, a target point whose adjusted absolute value of the distance difference is less than a first distance threshold in a neighborhood area corresponding to the point is determined as a target reference point.

[0105] In step S41, when the proportion of the number of target reference points in the neighborhood area corresponding to the point is less than a first ratio threshold, the point is determined to be a noise point.

[0106] Similarly, the electronic device can execute steps S40 and S41 in sequence for each point to determine the target reference point in the field area corresponding to the single point, and then determine whether the single point is a noise point by judging whether the proportion of the target reference points is less than the first ratio threshold, so as to determine whether each point in the point cloud data is a noise point.

[0107] For ease of understanding, the following explanation also introduces the definition of the current point, and each point in the point cloud data can be considered as the current point in turn. In this way, it can be understood that the electronic device considers a point in the point cloud data as the current point, and executes steps S40 and S41 for the current point to determine the target reference point in the field area corresponding to the current point, and then determines whether the point before the point is a noise point by judging whether the number of target reference points accounts for less than the first ratio threshold. Furthermore, the electronic device considers the next point in the point cloud data as the current point, and executes steps S40 and S41 for the current point to determine the target reference point in the field area corresponding to the current point, and then determines whether the point before the point is a noise point by judging whether the number of target reference points accounts for less than the first ratio threshold. By analogy, the electronic device can realize the judgment of whether each point in the point cloud data is a noise point, and obtain all noise points in the point cloud data.

[0108] Optionally, the electronic device can compare the absolute value of the adjusted distance difference of the target point with the first distance threshold (FeaTh1) for each target point in the neighborhood area of ​​the current point, and determine that the target point is a target reference point when the first distance difference is less than the first distance threshold; and determine that the target point is a first invalid point when the first distance difference is greater than or equal to the first distance threshold, so as to determine all noise points in the neighborhood area of ​​the current point.

[0109] Among them, the first distance threshold is a parameter used to determine whether the target point in the neighborhood area is a target reference point. In some embodiments, the first distance threshold corresponding to each point is positively correlated with the distance value of the point. That is, for points with different distance values, the corresponding first distance thresholds for determining the target reference point are different, and the larger the distance value of the point, the larger the corresponding first distance threshold.

[0110] In the embodiment of the present application, in the neighborhood area of ​​the current point, if the absolute value of the adjusted distance difference of the target point is smaller, it indicates that the distance between the target point and the current point is closer, and the relative concentration of the two is higher. Then, if the proportion of target points with a high concentration with the current point in the neighborhood area is smaller, it can be shown that the degree of dispersion of each point in the neighborhood area is greater; conversely, if the proportion of target points with a high concentration with the current point in the neighborhood area is larger, it can be shown that the degree of dispersion of each point in the neighborhood area is smaller. In this way, according to the characteristic that the degree of dispersion of the point cloud of the noise point is higher than that of the real object, when the proportion of target points with a high concentration in the neighborhood area of ​​the current point is relatively small, the current point can be considered as a noise point; conversely, when the proportion of target points with a high concentration in the neighborhood area of ​​the current point is relatively large, the current point can be considered as a real object point.

[0111] Based on this, the first distance threshold can be used to screen the target points in the neighborhood area of ​​the current point whose adjusted absolute value of the distance difference is less than the first distance threshold, so as to screen the target points that are closer to the current point, that is, the target reference points. In this way, the electronic device can determine whether the target points in the neighborhood area of ​​the current point are mainly concentrated around the current point by judging whether the proportion of the number of target reference points in the neighborhood area of ​​the current point is less than the first ratio threshold, that is, determine whether the cloud point discreteness of the neighborhood area is high, so that when the proportion of the number of target reference points in the neighborhood area is less than the first ratio threshold, indicating that the cloud point discreteness of the neighborhood area of ​​the current point is high, the current point is determined to be a noise point. On the contrary, when the proportion of the number of target reference points in the neighborhood area is greater than or equal to the first ratio threshold, indicating that the cloud point discreteness of the neighborhood area of ​​the current point is low, it is determined that the current point is not a noise point, that is, a real object point.

[0112] Optionally, the first distance threshold of the neighborhood area of ​​each point can be a dynamic threshold that is proportional to the distance value of the point. In this way, when the distance value of the current point is smaller, the first distance threshold of the neighborhood area of ​​the current point is smaller, so that the number of target reference points that are closer to the current point can be screened out, which can more effectively reflect the degree of cloud point discreteness.

[0113] In an optional implementation, the process of the electronic device determining whether the proportion of the number of target reference points in the neighborhood area is less than a first proportion threshold may include:

[0114] The electronic device can count the number of target reference points in the neighborhood area of ​​the current point, and determine the size relationship between the number of target reference points and a first quantity threshold. The first quantity threshold can be determined based on the total number of target points in the neighborhood area. When the number of target reference points is greater than or equal to the first quantity threshold, the electronic device determines that the number ratio of the target reference points in the neighborhood area is greater than or equal to the first ratio threshold. When the number of target reference points is less than the first quantity threshold, the electronic device determines that the number ratio of the target reference points in the neighborhood area is less than the first ratio threshold.

[0115] In another optional implementation, the electronic device may also count the number of target reference points in the neighborhood area of ​​the current point, and determine the ratio of the number of target reference points to the total number of target points in the neighborhood area as the number ratio of the target reference points in the neighborhood area. Determine the size of the number ratio of the target reference points and the first ratio threshold. When the number ratio of the target reference points is greater than or equal to the first ratio threshold, the electronic device determines that the current point is a real object point. When the number ratio of the target reference points is less than the first ratio threshold, the electronic device determines that the current point is a noise point.

[0116] Step 203: Filter out noise points in the point cloud data.

[0117] In the embodiment of the present application, after the electronic device performs step 202 to determine all noise points in the point cloud data, it filters out the noise points in the point cloud data.

[0118] Optionally, the electronic device may replace the data of the noise points in the point cloud data with the data of its neighborhood points to filter out the noise points in the point cloud data. Alternatively, the electronic device may also replace the data of the noise points in the point cloud data with the median of each target point in its neighborhood area to filter out the noise points in the point cloud data. It should be noted that the embodiments of the present application do not limit the method of filtering out noise points in the point cloud data.

[0119] In some embodiments of the present application, the process of determining the noise points in the point cloud data according to the three-dimensional coordinates and target parameters of the multiple points in step 202 includes: determining the points with a distance value less than a maximum distance threshold as suspected noise points in the point cloud data according to the distance value of the three-dimensional coordinates of each point to the laser radar; and determining the noise points in the suspected noise points according to the three-dimensional coordinates and target parameters of the multiple points. The implementation method of determining the noise points in the suspected noise points according to the three-dimensional coordinates and target parameters of the multiple points can refer to the implementation method of the aforementioned step 202, which will not be repeated here.

[0120] Since noise points are usually located in front of real object points. Therefore, after acquiring the point cloud data, the electronic device can first determine whether the distance value of the point is less than the maximum distance threshold for each point, so that when the distance value is less than the maximum distance threshold, indicating that the point is relatively close to the laser radar, the point is determined to be a suspected noise point, and all suspected noise points in the point cloud data are obtained, and then the noise points in the suspected noise points are determined. When the distance value of the point is greater than or equal to the maximum distance threshold, indicating that the point is relatively far from the laser radar, the point can be directly determined as a real object point without executing the subsequent noise point judgment steps, thereby improving the efficiency of noise point recognition.

[0121] In summary, the noise filtering method provided in the embodiment of the present application obtains the point cloud data of the laser radar to determine the noise points in the point cloud data according to the three-dimensional coordinates and target parameters of multiple points in the point cloud data, thereby filtering out the noise points in the point cloud data. Among them, the target parameters may include echo intensity and / or object reflectivity. Since the technical solution of the present application can comprehensively identify noise points based on at least two features of the points in the point cloud data. Therefore, compared with the solution of using a single feature for noise identification in the related art, it can effectively improve the recognition accuracy of noise points in the point cloud data, thereby improving the accuracy of noise filtering.

[0122] Embodiment 1

[0123] This embodiment takes the target parameter as echo intensity as an example to introduce the implementation process of the noise filtering method in detail. Figure 4 , which shows a flow chart of a noise filtering method provided in an embodiment of the present application. The noise filtering method can be applied to electronic devices. Optionally, the electronic device can be a car machine, a wearable device, etc. Figure 4 As shown, the noise filtering methods include:

[0124] Step 401: Obtain point cloud data of the laser radar. The point cloud data includes three-dimensional coordinates and echo intensity of multiple points.

[0125] The explanation and implementation of this step can refer to the explanation and implementation of the aforementioned step 201, and will not be elaborated in detail in the embodiment of the present application.

[0126] Step 402: Analyze and process the discreteness of each point in the neighborhood area corresponding to each point according to the three-dimensional coordinates and echo strengths of the multiple points.

[0127] Step 403: Determine the points corresponding to the neighborhood area whose discreteness is greater than the first critical discreteness as noise points in the point cloud data.

[0128] The explanation and implementation of step 402 to step 403 may refer to the explanation and implementation of the aforementioned step 202 when the target parameter is echo intensity. It should be noted that the first embodiment is different from the aforementioned method embodiment in the following ways.

[0129] Optionally, the process in which the electronic device analyzes and processes the discreteness of each point in the neighborhood area corresponding to each point according to the three-dimensional coordinates and echo strengths of the multiple points may include steps S50 and S51.

[0130] In step S50, the aggregation degree of each point in the neighborhood area corresponding to each point is adjusted according to the three-dimensional coordinates and the echo intensity of each point.

[0131] In step S51, the discreteness of each point in the adjusted neighborhood area corresponding to each point is analyzed and processed.

[0132] Optionally, for each point in the point cloud data, the electronic device can adjust the degree of aggregation of each point in the neighborhood area corresponding to the single point according to the three-dimensional coordinates and echo intensity of multiple points in the point cloud data, and obtain the adjusted neighborhood area corresponding to the single point. Then, the discrete degree of each point in the adjusted neighborhood area corresponding to the single point is analyzed and processed to obtain the discrete degree corresponding to the single point, and then the discrete degree corresponding to each point in the point cloud data is determined. And so on to determine the discrete degree corresponding to each point in the point cloud data. Then, the electronic device can determine the point whose discrete degree is greater than the first critical discrete degree as a noise point in the point cloud data.

[0133] Since noise points have the characteristics of higher point cloud dispersion than real objects, they also have the characteristics of relatively high point cloud density. The point cloud density of real objects is also relatively high. For example, Figure 1 The point cloud density of the dust cluster noise point 101 is relatively close to the point cloud density of the vehicle. Therefore, in order to effectively distinguish between noise points and real object points, the aggregation degree of each point in the neighborhood area corresponding to each point in the point cloud data can be adjusted to amplify the point cloud density difference between the noise points and the real object points in the point cloud data, thereby facilitating the improvement of the analysis accuracy of the discrete degree of all points in the neighborhood area corresponding to each point.

[0134] In an optional implementation, the process in which the electronic device adjusts the degree of aggregation of each point in a neighborhood area corresponding to each point according to the three-dimensional coordinates and the echo intensity of each point may include steps S60 and S61.

[0135] In step S60, for each point, the absolute value of the distance difference between each target point and the point is determined according to the three-dimensional coordinates of the point and the three-dimensional coordinates of each target point in the neighborhood area corresponding to the point.

[0136] In step S61, the absolute value of the distance difference of each target point is adjusted using the echo intensity of each target point.

[0137] The electronic device may sequentially perform step S60 and step S61 for each point to adjust the absolute value of the distance difference between each target point in the domain area corresponding to the single point in the point cloud data, thereby adjusting the absolute value of the distance difference between the target points in the domain area corresponding to each point in the point cloud data. The electronic device may determine the absolute value of the distance difference between each target point and the point according to the three-dimensional coordinates of the point and the three-dimensional coordinates of each target point in the neighborhood area corresponding to the point, which may be implemented by referring to the implementation in the aforementioned step 202, and this embodiment will not be described in detail here.

[0138] In some embodiments of the present application, the electronic device uses the echo intensity of each target point to adjust the absolute value of the distance difference of each target point, which may include the following steps S70 and S71.

[0139] In step S70, the spatial attenuation weight of each target point is calculated according to the echo intensity of each target point and the distance coordinate value in the three-dimensional coordinates.

[0140] Among them, the spatial attenuation weight is positively correlated with the echo attenuation degree of the target point per unit distance. That is, the greater the echo attenuation degree per unit distance, the faster the echo intensity changes with the distance value, and the greater the spatial attenuation weight,

[0141] Optionally, the electronic device may store a spatial attenuation function. The electronic device may input the echo intensity and distance value of each target point into the spatial attenuation function, and calculate the spatial attenuation weight of each target point. The spatial attenuation function may be a function with the echo intensity and distance value of the point as independent variables and the spatial attenuation weight as the dependent variable. The spatial attenuation function is defined according to the echo intensity and distance value of the point, and is a function used to indicate the echo attenuation degree of each point per unit distance.

[0142] In an optional case, the spatial attenuation function may be a function obtained by fitting the relationship between the echo intensity and the distance value of multiple points. Further optionally, the spatial attenuation function of the neighborhood area of ​​each point may be different. The electronic device may define the spatial attenuation function of the neighborhood area according to the current echo intensity and distance value of each point in the neighborhood area.

[0143] In step S71, the spatial attenuation weight of each target point is used to adjust the absolute value of the distance difference of each target point.

[0144] Optionally, the electronic device may, for each target point in the neighborhood area of ​​the current point, proportionally scale the absolute value of the distance difference of the target point according to the spatial attenuation weight of the target point to obtain the adjusted absolute value of the distance difference of the target point, so as to obtain the adjusted absolute value of the distance difference of each target point. In some embodiments, the electronic device may, for each target point in the neighborhood area of ​​the current point, calculate the product of the spatial attenuation weight of the target point and the absolute value of its distance difference to obtain the adjusted absolute value of the distance difference of the target point.

[0145] For example, the electronic device can establish a spatial attenuation matrix of the neighborhood area of ​​the current point based on the spatial attenuation weight of each target point in the neighborhood area of ​​the current point. The spatial attenuation matrix is ​​a two-dimensional matrix, which includes the spatial attenuation weight of each target point in the neighborhood area, and the position of the spatial attenuation weight of the target point in the spatial attenuation matrix is ​​the position of the target point in the neighborhood area.

[0146] The electronic device calculates the spatial attenuation weight of the neighborhood area of ​​the current point and the product of the distance difference matrix to obtain the first characteristic matrix of the neighborhood area of ​​the current point. The first characteristic matrix includes the absolute value of the adjusted distance difference of each target point in the neighborhood area, and the position of the absolute value of the adjusted distance difference of the target point in the first characteristic matrix is ​​the position of the target point in the neighborhood area.

[0147] Step 404: Filter out noise points in the point cloud data.

[0148] The explanation and implementation of this step can refer to the explanation and implementation of the aforementioned step 203, and will not be elaborated in detail in this embodiment of the present application.

[0149] In some embodiments of the present application, before analyzing and processing the degree of discreteness of each point in the neighborhood area corresponding to each point according to the three-dimensional coordinates and echo intensity of the multiple points in step 402, the noise point filtering method also includes: according to the distance value of the three-dimensional coordinates of each point, determining the points with distance values ​​less than the maximum distance threshold as suspected noise points in the point cloud data; and determining the noise points among the suspected noise points according to the three-dimensional coordinates and target parameters of the multiple points.

[0150] Since noise points are usually located in front of real object points. Therefore, after acquiring the point cloud data, the electronic device can first determine whether the distance value of the point is less than the maximum distance threshold for each point, so that when the distance value is less than the maximum distance threshold, indicating that the point is relatively close to the laser radar, the point is determined to be a suspected noise point, and all suspected noise points in the point cloud data are obtained, and then the noise points in the suspected noise points are determined. When the distance value of the point is greater than or equal to the maximum distance threshold, indicating that the point is relatively far from the laser radar, the point can be directly determined as a real object point without executing the subsequent noise point judgment steps, thereby improving the efficiency of noise point recognition.

[0151] In summary, the noise point filtering method provided in the embodiment of the present application obtains the point cloud data of the laser radar to determine the noise points in the point cloud data according to the three-dimensional coordinates and echo intensity of multiple points in the point cloud data, thereby filtering out the noise points in the point cloud data. Since the technical solution of the present application can comprehensively identify the noise points according to at least two features of the points in the point cloud data. Therefore, compared with the solution of using a single feature for noise point identification in the related art, it can effectively improve the recognition accuracy of noise points in the point cloud data, thereby improving the accuracy of noise point filtering.

[0152] Embodiment 2

[0153] This embodiment takes the reflectivity of an object as an example to introduce the implementation process of the noise filtering method in detail. Figure 5 , which shows a flow chart of a noise filtering method provided in an embodiment of the present application. The noise filtering method can be applied to electronic devices. Optionally, the electronic device can be a car machine, a wearable device, etc. Figure 5 As shown, the noise filtering methods include:

[0154] Step 501: Obtain point cloud data of the laser radar. The point cloud data includes the three-dimensional coordinates of multiple points and the reflectivity of the object.

[0155] The explanation and implementation of this step can refer to the explanation and implementation of the aforementioned step 201, and will not be elaborated in detail in the embodiment of the present application.

[0156] Step 502: Analyze and process the discreteness of each point in the neighborhood area corresponding to each point according to the three-dimensional coordinates of the multiple points and the reflectivity of the object.

[0157] Step 503: Determine the points corresponding to the neighborhood area whose discreteness is greater than the first critical discreteness as noise points in the point cloud data.

[0158] The explanation and implementation of step 502 to step 503 can refer to the explanation and implementation of the aforementioned step 202 when the target parameter is the reflectivity of the object. It should be noted that the first embodiment is different from the aforementioned method embodiment in the following ways.

[0159] Optionally, the process in which the electronic device analyzes and processes the discreteness of each point in the neighborhood area corresponding to each point according to the three-dimensional coordinates of the multiple points and the reflectivity of the object may include steps S80 to S81.

[0160] In step S80, the aggregation degree of each point in the neighborhood area corresponding to each point is adjusted according to the three-dimensional coordinates of each point and the reflectivity of the object.

[0161] In step S81, the discreteness of each point in the adjusted neighborhood area corresponding to each point is analyzed and processed.

[0162] Optionally, for each point in the point cloud data, the electronic device can adjust the degree of aggregation of each point in the neighborhood area corresponding to the single point according to the three-dimensional coordinates of multiple points in the point cloud data and the reflectivity of the object, and obtain the adjusted neighborhood area corresponding to the single point. Then, the discrete degree of each point in the adjusted neighborhood area corresponding to the single point is analyzed and processed to obtain the discrete degree corresponding to the single point, and then the discrete degree corresponding to each point in the point cloud data is determined. And so on to determine the discrete degree corresponding to each point in the point cloud data. Then, the electronic device can determine the point whose discrete degree is greater than the first critical discrete degree as a noise point in the point cloud data.

[0163] In an optional implementation, the process in which the electronic device adjusts the degree of aggregation of each point in a neighborhood area corresponding to each point according to the three-dimensional coordinates of each point and the reflectivity of the object may include steps S90 to S91.

[0164] In step S90, for each point, the absolute value of the distance difference between each target point and the point is determined based on the three-dimensional coordinates of the point and the three-dimensional coordinates of each target point in the neighborhood area corresponding to the point.

[0165] In step S91, the object reflectivity of each target point is used to adjust the absolute value of the distance difference of each target point.

[0166] The electronic device may sequentially perform step S90 and step S91 for each point to adjust the absolute value of the distance difference between each target point in the domain area corresponding to the single point in the point cloud data, thereby adjusting the absolute value of the distance difference between the target points in the domain area corresponding to each point in the point cloud data. The electronic device may determine the absolute value of the distance difference between each target point and the point according to the three-dimensional coordinates of the point and the three-dimensional coordinates of each target point in the neighborhood area corresponding to the point, which may be implemented by referring to the implementation in the aforementioned step 202, and will not be described in detail in this embodiment.

[0167] In some embodiments of the present application, the electronic device uses the object reflectivity of each target point, and the process of adjusting the absolute value of the distance difference of each target point may include steps S100 and S101.

[0168] In step S100, the object reflectivity of each target point is reversely amplified to obtain the reflection weight of each target point.

[0169] In the embodiment of the present application, the reverse amplification process refers to determining the reflection weight of a target point having a relatively small object reflectivity value to be a relatively large value.

[0170] In an optional implementation, the electronic device performs reverse amplification processing on the object reflectivity of each target point, and the process of obtaining the reflection weight of each target point may include: setting the reflection weight of the target point whose object reflectivity is greater than or equal to the reflectivity threshold (RefTh) to 0, and setting the reflection weight of the target point whose object reflectivity is less than the reflectivity threshold (RefTh) to a value greater than 0. Optionally, the reflection weight of the target point whose object reflectivity is less than the reflectivity threshold (RefTh) is set to a positive integer at least equal to 10.

[0171] In one example, the electronic device may set the reflection weight of the target point whose object reflectivity is greater than or equal to the reflectivity threshold to 0, and set the reflection weight of the target point whose object reflectivity is less than the reflectivity threshold to 10. In another example, the electronic device may set the reflection weight of the target point whose object reflectivity is greater than or equal to the reflectivity threshold to 0, and set the reflection weight of the target point whose object reflectivity is less than the reflectivity threshold to a target value. The target value is greater than or equal to 10 and is negatively correlated with the object reflectivity. That is, on the basis that the object reflectivity of the target point is less than the reflectivity threshold, the smaller the value is, the greater the reflection weight of the target point is.

[0172] In another optional implementation, the electronic device performs reverse amplification processing on the object reflectivity of each target point to obtain the reflection weight of each target point, which may include: calculating the reflection weight of each target point according to the negative correlation function and the object reflectivity of each target point. The negative correlation function may be a function with the object reflectivity as an independent variable and the reflection weight as a dependent variable, and is used to indicate the negative correlation between the reflection weight and the object reflectivity.

[0173] It should be noted that, in some embodiments, the negative correlation function may be a custom function. Alternatively, the negative correlation function may be a function obtained by fitting the object reflectivity of multiple points with a preset reflection weight. For example, the negative correlation function may be a linear function with a negative slope, and the intercept of the linear function on the y-axis is greater than or equal to 10.

[0174] In step S101, the reflection weight of each target point is used to adjust the absolute value of the distance difference of each target point.

[0175] Optionally, the electronic device may, for each target point in the neighborhood area of ​​the current point, proportionally scale the absolute value of the distance difference of the target point according to the reflection weight of the target point to obtain the adjusted absolute value of the distance difference of the target point, so as to obtain the adjusted absolute value of the distance difference of each target point. In some embodiments, the electronic device may, for each target point in the neighborhood area of ​​the current point, calculate the product of the reflection weight of the target point and the absolute value of the distance difference to obtain the adjusted absolute value of the distance difference of the target point, so as to obtain the adjusted absolute value of the distance difference of each target point.

[0176] For example, the electronic device can establish a reflectivity weight matrix of the neighborhood area of ​​the current point based on the reflection weight of each target point in the neighborhood area of ​​the current point. The reflectivity weight matrix is ​​a two-dimensional matrix, which includes the reflection weight of each target point in the neighborhood area, and the position of the reflection weight of the target point in the reflectivity weight matrix is ​​the position of the target point in the neighborhood area.

[0177] The electronic device calculates the product of the reflectivity weight matrix and the distance difference matrix of the neighborhood area of ​​the current point to obtain a second characteristic matrix of the neighborhood area of ​​the current point. The second characteristic matrix includes the absolute value of the adjusted distance difference of each target point in the neighborhood area, and the position of the absolute value of the adjusted distance difference of the target point in the second characteristic matrix is ​​the position of the target point in the neighborhood area.

[0178] In this way, we can use the property that the object reflectivity of noise points is lower than the object reflectivity of real objects. On the basis of predicting whether the target point is a noise point through the object reflectivity, the absolute value of the distance difference of the target points suspected to be noise points in the neighborhood area is increased, and the absolute value of the distance difference of the target points suspected to be real object points in the neighborhood area is reduced, thereby amplifying the distance value difference of each target point in the neighborhood area, so as to amplify the point cloud density difference between the noise points and the real object points in the point cloud data, and facilitate the subsequent use of the distance value difference of each target point in the neighborhood area of ​​the current point for noise point judgment.

[0179] Step 504: Filter out noise points in the point cloud data.

[0180] The explanation and implementation of this step can refer to the explanation and implementation of the aforementioned step 203, and will not be elaborated in detail in this embodiment of the present application.

[0181] In some embodiments of the present application, before analyzing and processing the degree of discreteness of each point in the neighborhood area corresponding to each point according to the three-dimensional coordinates of the multiple points and the reflectivity of the object in step 502, the noise point filtering method also includes: determining the points with distance values ​​less than the maximum distance threshold as suspected noise points in the point cloud data according to the distance value of the three-dimensional coordinates of each point; and determining the noise points among the suspected noise points according to the three-dimensional coordinates of the multiple points and the target parameters.

[0182] Since noise points are usually located in front of real object points. Therefore, after acquiring the point cloud data, the electronic device can first determine whether the distance value of the point is less than the maximum distance threshold for each point, so that when the distance value is less than the maximum distance threshold, indicating that the point is relatively close to the laser radar, the point is determined to be a suspected noise point, and all suspected noise points in the point cloud data are obtained, and then the noise points in the suspected noise points are determined. When the distance value of the point is greater than or equal to the maximum distance threshold, indicating that the point is relatively far from the laser radar, the point can be directly determined as a real object point without executing the subsequent noise point judgment steps, thereby improving the efficiency of noise point recognition.

[0183] In summary, the noise point filtering method provided in the embodiment of the present application obtains the point cloud data of the laser radar to determine the noise points in the point cloud data according to the three-dimensional coordinates of multiple points in the point cloud data and the reflectivity of the object, thereby filtering out the noise points in the point cloud data. Since the technical solution of the present application can comprehensively identify the noise points according to at least two features of the points in the point cloud data. Therefore, compared with the solution of using a single feature for noise point identification in the related art, it can effectively improve the recognition accuracy of noise points in the point cloud data, thereby improving the accuracy of noise point filtering.

[0184] Embodiment 3

[0185] This embodiment takes the target parameters including echo intensity and object reflectivity as an example to introduce the implementation process of the noise filtering method in detail. Figure 6 , which shows a flow chart of a noise filtering method provided in an embodiment of the present application. The noise filtering method can be applied to electronic devices. Optionally, the electronic device can be a car machine, a wearable device, etc. Figure 6 As shown, the noise filtering methods include:

[0186] Step 601: Obtain point cloud data of the laser radar. The point cloud data includes the three-dimensional coordinates of multiple points, echo intensity, and object reflectivity.

[0187] The explanation and implementation of this step can refer to the explanation and implementation of the aforementioned step 201, and will not be elaborated in detail in the embodiment of the present application.

[0188] Step 602: Analyze and process the discreteness of each point in the neighborhood area corresponding to each point according to the three-dimensional coordinates and echo strengths of the multiple points.

[0189] Step 603: Determine the points corresponding to the neighborhood area whose discreteness is greater than the first critical discreteness as noise points in the point cloud data.

[0190] Step 604: Determine the points corresponding to the neighborhood area whose discreteness is less than or equal to the first critical discreteness as valid points in the point cloud data.

[0191] Step 605: According to the three-dimensional coordinates of the multiple points and the reflectivity of the object, the discreteness of each point in the neighborhood area corresponding to each valid point is analyzed again.

[0192] Step 606: Determine the points corresponding to the neighborhood area whose discreteness is greater than the second critical discreteness as noise points in the point cloud data.

[0193] The explanation and implementation of steps 602 to 604 may refer to the explanation and implementation of the aforementioned step 202 when the target parameter is the echo intensity, and the explanation and implementation of steps 402 to 403 in the aforementioned embodiment 1. The explanation and implementation of steps 605 to 606 may refer to the explanation and implementation of the aforementioned step 202 when the target parameter is the object reflectivity, and the explanation and implementation of steps 502 to 503 in the aforementioned embodiment 2. This embodiment of the application will not be elaborated on.

[0194] It should be noted that, different from the aforementioned method side embodiment, after the electronic device uses the echo intensity to analyze and process the discreteness of each point in the neighborhood area corresponding to a point in the point cloud data to determine that the point is a valid point (i.e., a real object point), the electronic device can also use the object reflectivity to analyze and process the discreteness of each point in the neighborhood area corresponding to the valid point again to determine again whether the valid point is a noise point. Using the echo intensity and object reflectivity of the point to perform secondary verification processing on the noise point can effectively ensure the accuracy of the noise point confirmation, and thus ensure the noise point filtering effect.

[0195] Step 607: Filter out noise points in the point cloud data.

[0196] The explanation and implementation of this step can refer to the explanation and implementation of the aforementioned step 203, and will not be elaborated in detail in this embodiment of the present application.

[0197] In some embodiments of the present application, before analyzing and processing the degree of discreteness of each point in the neighborhood area corresponding to each point according to the three-dimensional coordinates and echo intensity of the multiple points in step 602, the noise point filtering method also includes: according to the distance value of the three-dimensional coordinates of each point, determining the points with distance values ​​less than the maximum distance threshold as suspected noise points in the point cloud data; and determining the noise points among the suspected noise points according to the three-dimensional coordinates and target parameters of the multiple points.

[0198] Since noise points are usually located in front of real object points. Therefore, after acquiring the point cloud data, the electronic device can first determine whether the distance value of the point is less than the maximum distance threshold for each point, so that when the distance value is less than the maximum distance threshold, indicating that the point is relatively close to the laser radar, the point is determined to be a suspected noise point, and all suspected noise points in the point cloud data are obtained, and then the noise points in the suspected noise points are determined. When the distance value of the point is greater than or equal to the maximum distance threshold, indicating that the point is relatively far from the laser radar, the point can be directly determined as a real object point without executing the subsequent noise point judgment steps, thereby improving the efficiency of noise point recognition.

[0199] In the embodiment of the present application, the technical solution of the present application can use the five features of the three-dimensional coordinates of the cloud point, the echo intensity and the object reflectivity to comprehensively identify the noise points. Compared with the solution of using a single feature for noise point identification in the related art, it can effectively improve the recognition accuracy of noise points in the point cloud data, thereby improving the accuracy of noise point filtering. In addition, the technical solution of the present application also uses the echo intensity and the object reflectivity to perform secondary verification processing of the noisy points, thereby further ensuring the recognition accuracy of the noise points and improving the noise point filtering effect.

[0200] Embodiment 4

[0201] This embodiment takes the target parameters including echo intensity and object reflectivity as an example to introduce the implementation process of the noise filtering method in detail. Figure 7 , which shows a flow chart of a noise filtering method provided in an embodiment of the present application. The noise filtering method can be applied to electronic devices. Optionally, the electronic device can be a car machine, a wearable device, etc. Figure 7 As shown, the noise filtering methods include:

[0202] Step 701: Obtain point cloud data of the laser radar. The point cloud data includes the three-dimensional coordinates of multiple points, echo intensity, and object reflectivity.

[0203] The explanation and implementation of this step can refer to the explanation and implementation of the aforementioned step 201, and will not be elaborated in detail in the embodiment of the present application.

[0204] Step 702: Analyze and process the discreteness of each point in the neighborhood area corresponding to each point according to the three-dimensional coordinates of the multiple points and the reflectivity of the object.

[0205] Step 703: Determine the points corresponding to the neighborhood area whose discreteness is greater than the first critical discreteness as noise points in the point cloud data.

[0206] Step 704: Determine the points corresponding to the neighborhood area whose discreteness is less than or equal to the first critical discreteness as valid points in the point cloud data.

[0207] Step 705: According to the three-dimensional coordinates and echo strengths of the multiple points, the discreteness of each point in the neighborhood area corresponding to each valid point is analyzed again.

[0208] Step 706: Determine the points corresponding to the neighborhood area whose discreteness is greater than the second critical discreteness as noise points in the point cloud data.

[0209] The explanation and implementation of steps 702 to 704 can refer to the explanation and implementation of the aforementioned step 202 when the target parameter is the object reflectivity, and the explanation and implementation of steps 502 to 503 in the aforementioned embodiment 2. The explanation and implementation of steps 705 to 706 can refer to the explanation and implementation of the aforementioned step 202 when the target parameter is the echo intensity, and the explanation and implementation of steps 402 to 403 in the aforementioned embodiment 1. This embodiment of the application will not be elaborated on.

[0210] It should be noted that, different from the aforementioned method side embodiment, after the electronic device uses the object reflectivity to analyze and process the discreteness of each point in the neighborhood area corresponding to a point in the point cloud data to determine that the point is a valid point (i.e., a real object point), the electronic device can also use the echo intensity to analyze and process the discreteness of each point in the neighborhood area corresponding to the valid point again to determine whether the valid point is a noise point. Using the echo intensity and object reflectivity of the point to perform secondary verification processing on the noise point can effectively ensure the accuracy of the noise point confirmation, and thus ensure the noise point filtering effect.

[0211] Step 707: Filter out noise points in the point cloud data.

[0212] The explanation and implementation of this step can refer to the explanation and implementation of the aforementioned step 203, and will not be elaborated in detail in this embodiment of the present application.

[0213] In some embodiments of the present application, before analyzing and processing the degree of discreteness of each point in the neighborhood area corresponding to each point according to the three-dimensional coordinates of the multiple points and the reflectivity of the object in step 702, the noise point filtering method also includes: according to the distance value of the three-dimensional coordinates of each point, determining the points with distance values ​​less than the maximum distance threshold as suspected noise points in the point cloud data; and determining the noise points among the suspected noise points according to the three-dimensional coordinates of the multiple points and the target parameters.

[0214] Since noise points are usually located in front of real object points. Therefore, after acquiring the point cloud data, the electronic device can first determine whether the distance value of the point is less than the maximum distance threshold for each point, so that when the distance value is less than the maximum distance threshold, indicating that the point is relatively close to the laser radar, the point is determined to be a suspected noise point, and all suspected noise points in the point cloud data are obtained, and then the noise points in the suspected noise points are determined. When the distance value of the point is greater than or equal to the maximum distance threshold, indicating that the point is relatively far from the laser radar, the point can be directly determined as a real object point without executing the subsequent noise point judgment steps, thereby improving the efficiency of noise point recognition.

[0215] In the embodiment of the present application, the technical solution of the present application can use the five features of the three-dimensional coordinates of the cloud point, the echo intensity and the object reflectivity to comprehensively identify the noise points. Compared with the solution of using a single feature for noise point identification in the related art, it can effectively improve the recognition accuracy of noise points in the point cloud data, thereby improving the accuracy of noise point filtering. In addition, the technical solution of the present application also uses the echo intensity and the object reflectivity to perform secondary verification processing of the noisy points, thereby further ensuring the recognition accuracy of the noise points and improving the noise point filtering effect.

[0216] In order to facilitate understanding of the technical solution of the present application, the noise filtering method provided in the embodiment of the present application is further schematically illustrated by the following example. Figure 8 As shown, the noise filtering methods include:

[0217] Step 801: Receive point cloud data acquired by a one-dimensional rotating mirror laser radar using a line scanning method, where the point cloud data includes three-dimensional coordinates of multiple points, echo intensity, and object reflectivity.

[0218] Step 802: for each point (i.e., current point), determine the 3-row × 5-column neighborhood area of ​​the current point in three-dimensional space according to the acquisition order of each point in the point cloud data, and obtain each point in the neighborhood window (i.e., target point) in the neighborhood area.

[0219] Step 803: Obtain the distance value of the current point (ie, the coordinate value of the z-axis in the three-dimensional position coordinates), the echo intensity, and the object reflectivity.

[0220] Step 804: Determine whether the distance value of the current point is less than the maximum distance threshold value DisThMax. If the distance value of the current point is greater than or equal to the maximum distance threshold value DisThMax, execute step 805. If the distance value of the current point is less than the maximum distance threshold value DisThMax, execute step 806.

[0221] Step 805: Determine whether the current point is a real object point.

[0222] Step 806: Obtain the depth value, echo intensity and object reflectivity of each point in the neighborhood window in the 3-row×5-column neighborhood area.

[0223] Step 807: Calculate the spatial attenuation weight of each point in the neighborhood window in the neighborhood area of ​​the current point according to the echo intensity and distance value of each point in the neighborhood window in the neighborhood area.

[0224] Step 808: Calculate a distance difference matrix based on the distance value of the current point and the distance values ​​of points in each neighborhood window in the neighborhood area of ​​the current point.

[0225] Step 809, calculate the spatial attenuation weight of the neighborhood area of ​​the current point and the product of the distance difference matrix to obtain the first characteristic matrix of the neighborhood area of ​​the current point. The first characteristic matrix includes the first distance difference of each point in the neighborhood window in the neighborhood area. The first distance difference refers to the absolute value of the distance difference obtained by adjusting the absolute value of the distance difference of the points in the neighborhood window using the spatial attenuation weight of the points in the neighborhood window.

[0226] Step 810: Determine whether the first distance difference of the points in the neighborhood window in the first feature matrix is ​​less than the first distance threshold FeaTh1. If the first distance difference is less than the first distance threshold, execute step 811. If the first distance difference is greater than or equal to the first distance threshold, execute step 812.

[0227] Step 811: Determine the point in the neighborhood window as a valid point in the neighborhood window (ie, the target reference point).

[0228] Step 812: Determine the points within the neighborhood window as invalid points of the neighborhood window.

[0229] Step 813: Accumulate the number of valid points in the neighborhood window to obtain the number Num1 of valid points in the neighborhood window (ie, the number of target reference points).

[0230] Step 814: Determine whether the number of valid points in the window is greater than the valid point number threshold (i.e., the first number threshold). If the number of valid points in the window is less than or equal to the valid point number threshold, execute step 815. If the number of valid points in the window is greater than the valid point number threshold, execute step 816.

[0231] Step 815: directly determine the current point as a dust noise point.

[0232] Step 816: preliminarily determine the current point as a real object point (i.e., determine the current point as a valid point), and perform reflectivity weight information determination on the preliminarily determined real object point. The reflectivity weight information determination specifically includes:

[0233] Step 817, defining a reflectance weight matrix, specifically including: setting the reflection weights of points in the neighborhood window whose object reflectance is greater than or equal to the reflectance threshold RefTh in the neighborhood area of ​​the current point to 0, and setting the reflection weights of points in the neighborhood window whose object reflectance is less than the reflectance threshold to 10, so as to obtain a reflectance weight matrix.

[0234] Step 818, calculate the reflectivity weight matrix and the distance difference matrix product of the neighborhood area of ​​the current point to obtain the second characteristic matrix (also called the reflectivity information weight matrix) of the neighborhood area of ​​the current point. The second characteristic matrix includes the second distance difference of each target point in the neighborhood area. The second distance difference refers to the absolute value of the adjusted distance difference obtained by adjusting the absolute value of the distance difference of the points in the neighborhood window using the reflection weight of the points in the neighborhood window.

[0235] Step 819: Determine whether the second distance difference of the points in the neighborhood window in the second feature matrix is ​​less than the second distance threshold FeaTh2. If the second distance difference is less than the second distance threshold, execute step 820. If the second distance difference is greater than or equal to the second distance threshold, execute step 821.

[0236] Step 820: Determine the point in the neighborhood window as a valid point in the neighborhood window (ie, the target reference point).

[0237] Step 821: Determine the points within the neighborhood window as invalid points of the neighborhood window.

[0238] Step 822: Accumulate the number of valid points in the neighborhood window to obtain the number Num2 of valid points in the neighborhood window (ie, the number of target reference points).

[0239] Step 823: Determine whether the number of valid points in the window is greater than the valid point number threshold (i.e., the second number threshold). If the number of valid points in the window is less than or equal to the valid point number threshold, execute step 815 to determine the current point as a dust noise point. If the number of valid points in the window is greater than the valid point number threshold, execute step 805 to determine the current point as a real object point.

[0240] Step 824: Filter out dust noise in the point cloud data.

[0241] Will Figure 1 The point cloud data of the effect diagram shown in the figure can achieve quality optimization of the point cloud data after the above calculation process using the solution of this application. The effect diagram of the optimized point cloud data is as follows Fig. 9 As shown, Fig. 9 In the figure, the arrow 901 in the box points to Figure 1 The result after filtering out the medium and high density dust noise cluster 101. Figure 1By comparison, it can be seen that the high-density dust noise cluster 101 is effectively filtered out, the noise residual rate in the point cloud data is less than 1%, and the number of remaining noise points is less than 5. At the same time, real objects such as vehicles, ground lines, and walls are effectively retained.

[0242] In summary, the noise filtering method provided in the embodiment of the present application obtains the point cloud data of the laser radar to determine the noise points in the point cloud data according to the three-dimensional coordinates and target parameters of multiple points in the point cloud data, thereby filtering out the noise points in the point cloud data. Among them, the target parameters may include echo intensity and / or object reflectivity. Since the technical solution of the present application can comprehensively identify noise points based on at least two features of the points in the point cloud data. Therefore, compared with the solution of using a single feature for noise identification in the related art, it can effectively improve the recognition accuracy of noise points in the point cloud data, thereby improving the accuracy of noise filtering.

[0243] Please refer to Fig.10 , which shows a block diagram of a noise filtering device provided in an embodiment of the present application. Fig.10 As shown, the noise filtering device 1000 includes: an acquisition module 1001 , a determination module 1002 and a filtering module 1003 .

[0244] An acquisition module 1001 is used to acquire point cloud data of a laser radar, where the point cloud data includes three-dimensional coordinates of multiple points and target parameters, where the target parameters include echo intensity and / or object reflectivity;

[0245] A determination module 1002, for determining noise points in the point cloud data according to the three-dimensional coordinates of the plurality of points and the target parameters;

[0246] The filtering module 1003 is used to filter out noise points in the point cloud data.

[0247] In an embodiment of the present application, by acquiring the point cloud data of the laser radar, the noise points in the point cloud data are determined according to the three-dimensional coordinates and target parameters of multiple points in the point cloud data, thereby filtering out the noise points in the point cloud data. Among them, the target parameters may include echo intensity and / or object reflectivity. Since the technical solution of the present application can comprehensively identify noise points based on at least two features of the points in the point cloud data. Therefore, compared with the solution of using a single feature for noise point identification in the related art, the recognition accuracy of noise points in the point cloud data can be effectively improved, thereby improving the accuracy of noise point filtering.

[0248] Optionally, the determination module 1002 is also used to: analyze and process the degree of discreteness of each point in the neighborhood area corresponding to each point based on the three-dimensional coordinates and target parameters of the multiple points, and determine the points corresponding to the neighborhood area whose degree of discreteness is greater than the first critical discreteness as noise points in the point cloud data.

[0249] Optionally, the determination module 1002 is further configured to:

[0250] According to the three-dimensional coordinates and target parameters of each point, adjust the aggregation degree of each point in the neighborhood area corresponding to each point;

[0251] The discrete degree of each point in the adjusted neighborhood area corresponding to each point is analyzed and processed.

[0252] Optionally, the determination module 1002 is further configured to:

[0253] For each point, determine the absolute value of the distance difference between each target point and the point according to the three-dimensional coordinates of the point and the three-dimensional coordinates of each target point in the neighborhood area corresponding to the point. The target point is the point in the neighborhood area.

[0254] The target parameters of each target point are used to adjust the absolute value of the distance difference of each target point.

[0255] Optionally, the determination module 1002 is further configured to:

[0256] For each point, determine the target point in the neighborhood area corresponding to the point, whose adjusted distance difference absolute value is less than the first distance threshold as the target reference point;

[0257] When the proportion of the number of target reference points in the neighborhood area corresponding to the point is less than the first ratio threshold, the point is determined to be a noise point.

[0258] Optionally, the first distance threshold is a parameter used to determine whether a target point in the neighborhood area is a target reference point, and the first distance threshold corresponding to each point is positively correlated with the distance value of the point.

[0259] Optionally, the target parameter is echo intensity; the determination module 1002 is further used to:

[0260] The spatial attenuation weight of each target point is calculated according to the echo intensity of each target point and the distance coordinate value in the three-dimensional coordinates. The spatial attenuation weight is positively correlated with the echo attenuation degree per unit distance. The spatial attenuation weight of each target point is used to adjust the absolute value of the distance difference of each target point.

[0261] Optionally, the determination module 1002 is further configured to: proportionally scale the absolute value of the distance difference of each target point according to the spatial attenuation weight of each target point to obtain an adjusted absolute value of the distance difference of each target point.

[0262] Optionally, the target parameter is the reflectivity of the object; the determination module 1002 is further used to:

[0263] The object reflectivity of each target point is reversely amplified to obtain the reflection weight of each target point; the reflection weight of each target point is used to adjust the absolute value of the distance difference of each target point.

[0264] Optionally, the determination module 1002 is further configured to:

[0265] The reflection weight of the target point whose object reflectivity is greater than or equal to the reflectivity threshold is set to 0; the reflection weight of the target point whose object reflectivity is less than the reflectivity threshold is set to a value greater than 0.

[0266] Optionally, the determination module 1002 is further configured to: scale the absolute value of the distance difference of each target point according to the reflection weight of each target point to obtain an adjusted absolute value of the distance difference of each target point.

[0267] Optionally, the determination module 1002 is further configured to:

[0268] The points corresponding to the neighborhood area whose discreteness is less than or equal to the critical discreteness are determined as valid points in the point cloud data;

[0269] According to the three-dimensional coordinates of multiple points and the reflectivity of the object, the discrete degree of each point in the neighborhood area corresponding to each valid point is analyzed and processed again, and the points corresponding to the neighborhood area with a discrete degree greater than the second critical discrete degree are determined as noise points in the point cloud data.

[0270] Optionally, the determination module 1002 is further configured to: determine points corresponding to the neighborhood area whose discreteness is less than or equal to the critical discreteness as valid points in the point cloud data;

[0271] According to the three-dimensional coordinates and echo intensity of multiple points, the discrete degree of each point in the neighborhood area corresponding to each valid point is analyzed and processed again, and the points corresponding to the neighborhood area with a discrete degree greater than the second critical discrete degree are determined as noise points in the point cloud data.

[0272] Optionally, the determination module 1002 is further used to: determine each point in the neighborhood area corresponding to each point according to the collection order of each point in the point cloud data.

[0273] Optionally, the determination module 1002 is further configured to:

[0274] According to the distance value between each point and the laser radar in the three-dimensional coordinates, the points whose distance value is less than the maximum distance threshold are determined as suspected noise points in the point cloud data;

[0275] Determine noise points among suspected noise points according to 3D coordinates of multiple points and target parameters.

[0276] In summary, the noise filtering device provided in the embodiment of the present application obtains the point cloud data of the laser radar to determine the noise points in the point cloud data according to the three-dimensional coordinates and target parameters of multiple points in the point cloud data, thereby filtering out the noise points in the point cloud data. Among them, the target parameters may include echo intensity and / or object reflectivity. Since the technical solution of the present application can comprehensively identify noise points based on at least two features of the points in the point cloud data. Therefore, compared with the solution of using a single feature for noise identification in the related art, it can effectively improve the recognition accuracy of noise points in the point cloud data, thereby improving the accuracy of noise filtering.

[0277] The present application also provides an electronic device, such as Fig.11 As shown, it includes a processor 1101 , a communication interface 1102 , a memory 1103 and a communication bus 1104 , wherein the processor 1101 , the communication interface 1102 , and the memory 1103 communicate with each other via the communication bus 1104 .

[0278] The memory 1103 is used to store computer programs.

[0279] When the processor 1101 is used to execute the program stored in the memory 1103, the following steps are implemented: obtaining the point cloud data of the laser radar, the point cloud data including the three-dimensional coordinates and target parameters of multiple points, the target parameters including the echo intensity and / or the object reflectivity; determining the noise points in the point cloud data according to the three-dimensional coordinates and target parameters of the multiple points; filtering out the noise points in the point cloud data.

[0280] The processor 1101 may also implement other steps in the above-mentioned noise filtering method, which will not be described in detail here.

[0281] The communication bus mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0282] The communication interface is used for communication between the above electronic device and other devices.

[0283] The memory may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0284] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0285] In another embodiment provided in the present application, a computer-readable storage medium is provided, in which instructions are stored. When the computer-readable storage medium is run on a computer, the computer executes the noise filtering method described in the above embodiment.

[0286] In another embodiment provided in the present application, a computer program product including instructions is also provided. When the computer program product is run on a computer, the computer is enabled to execute the noise filtering method described in the above embodiment.

[0287] In another embodiment provided by the present application, a vehicle is provided, wherein the vehicle comprises: the electronic device provided by the embodiment of the present application, or the computer-readable storage medium provided by the embodiment of the present application.

[0288] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive Solid State Disk (SSD)), etc.

[0289] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0290] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. For the embodiments of the device, electronic device, computer-readable storage medium and computer program products containing instructions, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0291] The above description is only a preferred embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application are included in the protection scope of the present application.

Claims

1. A noise filtering method, characterized in that: include: Acquire point cloud data of a laser radar, wherein the point cloud data includes three-dimensional coordinates of a plurality of points and target parameters, wherein the target parameters include echo intensity and / or object reflectivity; Determining noise points in the point cloud data according to the three-dimensional coordinates of the plurality of points and the target parameters; The noise points in the point cloud data are filtered out.

2. The method according to claim 1, characterized in that The step of determining the noise points in the point cloud data according to the three-dimensional coordinates of the plurality of points and the target parameters includes: According to the three-dimensional coordinates and target parameters of the multiple points, the discreteness of each point in the neighborhood area corresponding to each point is analyzed and processed, and the points corresponding to the neighborhood area with a discreteness greater than a first critical discreteness are determined as noise points in the point cloud data.

3. The method according to claim 2, characterized in that The analyzing and processing of the discreteness of each point in the neighborhood area corresponding to each point according to the three-dimensional coordinates of the multiple points and the target parameters includes: According to the three-dimensional coordinates and target parameters of each point, adjusting the aggregation degree of each point in the neighborhood area corresponding to each point; The discreteness of each point in the adjusted neighborhood area corresponding to each point is analyzed and processed.

4. The method according to claim 3, characterized in that The step of adjusting the aggregation degree of each point in the neighborhood area corresponding to each point according to the three-dimensional coordinates and the target parameters of each point includes: For each of the points, determining the absolute value of the distance difference between each of the target points and the point according to the three-dimensional coordinates of the point and the three-dimensional coordinates of each target point in the neighborhood area corresponding to the point, the target point being a point in the neighborhood area; The target parameter of each target point is used to adjust the absolute value of the distance difference of each target point.

5. The method according to claim 4, characterized in that The analyzing and processing of the discreteness of each point in the adjusted neighborhood area corresponding to each point, and determining the points corresponding to the neighborhood area having a discreteness greater than a first critical discreteness as noise points in the point cloud data, comprises: For each of the points, determine, in a neighborhood area corresponding to the point, a target point whose adjusted absolute value of the distance difference is less than a first distance threshold as a target reference point; When the proportion of the target reference points in the neighborhood area corresponding to the point is less than a first ratio threshold, the point is determined to be a noise point.

6. The method according to claim 5, characterized in that The first distance threshold is a parameter used to determine whether the target point in the neighborhood area is a target reference point. The first distance threshold corresponding to each point is positively correlated with the distance value of the point.

7. The method according to claim 4, characterized in that The target parameter is the echo intensity; and the step of using the target parameter of each target point to adjust the absolute value of the distance difference of each target point includes: Calculating the spatial attenuation weight of each target point according to the echo intensity of each target point and the distance coordinate value in the three-dimensional coordinates, wherein the spatial attenuation weight is positively correlated with the echo attenuation degree per unit distance; The spatial attenuation weight of each target point is used to adjust the absolute value of the distance difference of each target point.

8. The method according to claim 7, characterized in that The adopting the spatial attenuation weight of each target point to adjust the absolute value of the distance difference of each target point comprises: According to the spatial attenuation weight of each target point, the absolute value of the distance difference of each target point is scaled proportionally to obtain the adjusted absolute value of the distance difference of each target point.

9. The method according to claim 4, characterized in that The target parameter is the reflectivity of the object; and the step of using the target parameter of each target point to adjust the absolute value of the distance difference of each target point includes: Performing reverse amplification processing on the object reflectivity of each target point to obtain a reflection weight of each target point; The reflection weight of each target point is used to adjust the absolute value of the distance difference of each target point.

10. The method according to claim 9, characterized in that The reverse amplification process is performed on the object reflectivity of each target point to obtain the reflection weight of each target point, including: Setting the reflection weight of the target point whose object reflectivity is greater than or equal to the reflectivity threshold to 0; The reflection weight of the target point whose reflectivity of the object is less than the reflectivity threshold is set to a value greater than 0.

11. The method according to claim 9 or 10, characterized in that: The adopting the reflection weight of each target point to adjust the absolute value of the distance difference of each target point comprises: According to the reflection weight of each target point, the absolute value of the distance difference of each target point is scaled proportionally to obtain the adjusted absolute value of the distance difference of each target point.

12. The method according to claim 7, characterized in that The step of determining the noise points in the point cloud data according to the three-dimensional coordinates of the plurality of points and the target parameters further comprises: Determine points corresponding to the neighborhood area whose discreteness is less than or equal to the critical discreteness as valid points in the point cloud data; According to the three-dimensional coordinates of the multiple points and the reflectivity of the object, the discreteness of each point in the neighborhood area corresponding to each valid point is analyzed and processed again, and the points corresponding to the neighborhood area with a discreteness greater than a second critical discreteness are determined as noise points in the point cloud data.

13. The method according to claim 9, characterized in that The step of determining the noise points in the point cloud data according to the three-dimensional coordinates of the plurality of points and the target parameters further comprises: Determine points corresponding to the neighborhood area whose discreteness is less than or equal to the critical discreteness as valid points in the point cloud data; According to the three-dimensional coordinates and echo intensities of the multiple points, the discreteness of each point in the neighborhood area corresponding to each valid point is analyzed and processed again, and the points corresponding to the neighborhood area with a discreteness greater than a second critical discreteness are determined as noise points in the point cloud data.

14. The method according to claim 2, characterized in that The step of determining the noise points in the point cloud data according to the three-dimensional coordinates of the plurality of points and the target parameters further comprises: According to the collection order of each of the points in the point cloud data, each point in the neighborhood area corresponding to each of the points is determined.

15. The method according to claim 1, characterized in that The step of determining the noise points in the point cloud data according to the three-dimensional coordinates of the plurality of points and the target parameters includes: According to the distance value between each point and the laser radar in the three-dimensional coordinates, the point whose distance value is less than the maximum distance threshold is determined as a suspected noise point in the point cloud data; Noise points among the suspected noise points are determined according to the three-dimensional coordinates of the multiple points and the target parameters.

16. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the noise filtering method according to any one of claims 1 to 15 is implemented.

17. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the noise filtering method according to any one of claims 1 to 15 are implemented.

18. A vehicle, characterized in that: include: The electronic device of claim 16, or the computer-readable storage medium of claim 17.

19. A computer program product, characterized in that The computer program product comprises computer instructions, and when the computer instructions are executed on an electronic device, the electronic device executes the noise filtering method according to any one of claims 1 to 15.

Images