Point cloud assessment method and device

Abstract

A point cloud assessment method and device are used for assessing the distribution stability of a point cloud so as to improve the point cloud target detection performance or the point cloud target tracking performance. The method comprises: acquiring a point cloud set of a target object (S401), the point cloud set comprising point cloud data of a plurality of position points on the surface of the target object collected in a first time period; further, constructing a sample probability density distribution function of the point cloud set (S402), the sample probability density distribution function being used for describing probabilities of distributing the point clouds by the plurality of position points at the first moment; and constructing a cumulative probability density distribution function of the point cloud set (S403), wherein the cumulative probability density distribution function is used for describing the probability of distributing the point clouds by the plurality of position points in the second time period; evaluating the stability of the point cloud set based on the similarity between the sample probability density distribution function and the cumulative probability density distribution function (S404); wherein the similarity is positively correlated with the stability.

Description

Point cloud evaluation method and device Technical Field

[0001] The present application relates to the field of detection technology, and in particular to a point cloud evaluation method and device. Background Art

[0002] Radars have target detection and tracking capabilities. By scanning a target object, they collect a point cloud of that object. The stability of the target's point cloud distribution significantly impacts target detection and tracking performance. For example, an unstable point cloud distribution can prevent the target detection algorithm from accurately identifying and locating the target object. Therefore, when performing point cloud target detection or tracking, radars must optimize the target detection algorithm based on the stability of the point cloud data to improve the radar's target detection and tracking performance.

[0003] However, in the current point cloud quality assessment system, the usability of a single point cloud can only be determined by judging whether it falls within the range of the target detection frame. It is impossible to evaluate the distribution stability of the point cloud, and it is difficult to determine whether the point cloud meets the point cloud stability requirements of the detection equipment.

[0004] Therefore, how to evaluate the distribution stability of point clouds is a technical problem that needs to be solved urgently in this field.

[0005] Summary of the Invention

[0006] The embodiments of the present application provide a point cloud evaluation method and device for evaluating the distribution stability of a point cloud, thereby improving the point cloud target detection performance or point cloud target tracking performance.

[0007] In a first aspect, an embodiment of the present application provides a point cloud evaluation method, the method comprising: obtaining a point cloud set of a target object, the point cloud set comprising point cloud data of a plurality of position points on the surface of the target object collected within a first time period; further, constructing a sample probability density distribution function of the point cloud set, the sample probability density distribution function being used to describe the probability of the point cloud being distributed at the first moment by the plurality of position points; and constructing a cumulative probability density distribution function of the point cloud set, the cumulative probability density distribution function being used to describe the probability of the point cloud being distributed at the plurality of position points within a second time period; and then evaluating the stability of the point cloud set based on the similarity between the sample probability density distribution function and the cumulative probability density distribution function; wherein the similarity is positively correlated with the stability.

[0008] The first time period includes the second time period, and the first moment is any moment in the second time period, or the first time period includes the first moment and the second time period, and the interval between the first moment and any moment in the second time period is less than the preset duration.

[0009] Among them, the stability of the point cloud set can be used to describe the distribution characteristics of the point cloud at multiple locations on the surface of the target object.

[0010] The method may be performed by an evaluation device, such as a detection device or a detection device capable of supporting the detection device in implementing the functions required by the method, such as a chip system. For example, if the detection device is a radar, such as a lidar or other radar, the evaluation device may be the radar, or a device disposed within the radar capable of supporting the radar in implementing the functions required by the method, such as a chip system or other functional module.

[0011] In an embodiment of the present application, considering that the current point cloud quality assessment system is unable to assess the distribution stability of the point cloud set of the target object, a sample probability density distribution function and a cumulative probability density distribution function of the point cloud set are constructed, and based on the similarity between the sample probability density distribution function and the cumulative probability density distribution function, the distribution stability of the point cloud set is assessed. In this way, it is helpful to understand the distribution stability of the point cloud collected by the detection device at different times in the first time period, and then the detection device can optimize and adjust the target detection algorithm according to the distribution characteristics of the point cloud, thereby improving the target tracking performance or target detection performance of the detection device. Alternatively, when the point cloud distribution is unstable, a detection device with stronger ranging and / or angle measurement capabilities can be used to collect the point cloud of the target object, thereby achieving accurate positioning and identification of the target object.

[0012] In one possible design, constructing the sample probability density distribution function of the point cloud includes determining the probability of each of the multiple locations distributing the point cloud at a first moment, and constructing the sample probability density distribution function of the point cloud set based on the probability of each location distributing the point cloud at the first moment. In this design, the probability of each location distributing the point cloud at the first moment can be calculated.

[0013] Optionally, the above sample probability density distribution function is as follows:

[0014] Where p(x, y) is the sample probability density distribution function, δ(·) is the Dirac generalized function, and P{·} represents the probability of the i-th position point in the multiple position points distributing the point cloud at the first moment; (x i ,y i ) represents the position coordinates of the i-th position point among multiple position points at the first moment; the value of i is a positive integer from 1 to M, and M is the total number of position points.

[0015] Optionally, when determining the probability of a point cloud being distributed at the first moment for each of a plurality of position points, the point cloud set can be rasterized to form a grid map comprising a plurality of grids, and one grid in the grid map corresponds to one of the plurality of position points, and the probability of a point cloud being distributed at the first moment for each grid in the grid map can be further statistically analyzed.

[0016] In one possible design, the above-mentioned second time period includes K moments; the above-mentioned construction of the cumulative probability density distribution function of the point cloud includes: determining the first cumulative probability density distribution function at the k-1 moment; wherein the value of k is greater than 1 and less than the value of K; determining the second cumulative probability density distribution function at the k moment based on the first cumulative probability density distribution function and the cumulative probability of the point cloud distributed by multiple position points at the k moment in the second time period; when the difference between the first cumulative probability density distribution function and the second cumulative probability density distribution function is greater than or equal to the second threshold, determining the second cumulative probability density distribution function at the k+1 moment based on the second cumulative probability density distribution function and the cumulative probability of the point cloud distributed by the multiple position points at the k+1 moment in the second time period; or, when the difference between the first cumulative probability density distribution function and the second cumulative probability density distribution function is less than the second threshold, using the first cumulative probability density distribution function as the cumulative probability density distribution function of the point cloud set. In this design, the cumulative probability of the distribution point cloud of multiple position points at time k in the second time period and the first cumulative probability density distribution function at time k-1 can be combined to determine the second cumulative probability density distribution function at time k, so that the cumulative probability density distribution function of the point cloud set finally determined can more accurately reflect the distribution characteristics of the point cloud set, thereby making the stability evaluation of the point cloud set more accurate.

[0017] In one possible design, the cumulative probability density distribution function of the point cloud set is as follows:

[0018] in, Q″{(x i ,y i )}=Q′{(x i ,y i )}+N;

[0019] Among them, q(x, y) is the cumulative probability density distribution function, δ(·) is the Dirac generalized function, Q{(x i ,y i )} represents the final cumulative probability of the distribution point cloud of the i-th position point among multiple position points from time 0 to time k in the second time period, (x i ,y i ) represents the position coordinates of the i-th position point in the second time period; Q′{(xi ,y i )} is used to represent the cumulative probability of the point cloud of the i-th position point in the second time period at the k-1 moment, Q″{(x i ,y i )} is used to represent the cumulative probability of the distribution point cloud of the i-th position point among multiple position points at time k in the second time period; N is a constant.

[0020] In one possible design, before evaluating the distribution stability of the point cloud set based on the similarity between the sample probability density distribution function and the cumulative probability density distribution function of the point cloud set, the method further includes: determining a relative divergence KLD between the sample probability density distribution function and the cumulative probability density distribution function of the point cloud set; and based on the KLD, determining the similarity between the sample probability density distribution function and the cumulative probability density distribution function of the point cloud set; wherein KLD is negatively correlated with the similarity. In this design, determining the similarity between the sample probability density distribution function and the cumulative probability density distribution function of the point cloud set based on KLD makes the evaluation of the similarity between the sample probability density distribution function and the cumulative probability density distribution function of the point cloud set more accurate, thereby making the stability evaluation of the point cloud set based on this similarity more accurate.

[0021] Among them, the negative correlation between KLD and similarity can be understood as the smaller the KLD, the greater the similarity; correspondingly, the better the stability of the point cloud set.

[0022] Optionally, the KLD, the sample probability density distribution function of the point cloud set, and the cumulative probability density distribution function of the point cloud set satisfy the following relationship:

[0023] Among them, D KL (p||q) is KLD, p(x, y) is the sample probability density distribution function, and q(x, y) is the cumulative probability density distribution function.

[0024] In one possible design, when the target object is a moving target, the multiple position points include a first position point and a second position point, and the second position point is a position point other than the first position point among the multiple position points; the cumulative probability of the distribution point cloud of the multiple position points at the k moment in the second time period includes a first cumulative probability and a second cumulative probability; wherein the first cumulative probability is the cumulative probability of the distribution point cloud of the first position point at the k moment in the second time period, and the second cumulative probability is the cumulative probability of the distribution point cloud of the second position point at the k moment in the second time period; accordingly, the above-mentioned determination of the second cumulative probability density distribution function at the k moment based on the first cumulative probability density distribution function and the cumulative probability of the distribution point cloud of the multiple position points at the k moment in the second time period includes: when the first cumulative probability is less than the second threshold, deleting the first cumulative probability from the cumulative probability of the distribution point cloud of the multiple position points at the k moment in the second time period to obtain the second cumulative probability; and determining the second cumulative probability density distribution function based on the second cumulative probability. In this design, taking into account the changing state of the moving target, if no distribution point cloud is counted at the first position for a long time, the probability of the distribution point cloud at the first position is low. Therefore, the probability corresponding to the first position needs to be excluded from the cumulative probability distribution, which improves the response of the cumulative probability distribution function to the changing state of the moving target and effectively ensures the effectiveness of the cumulative probability distribution function.

[0025] Optionally, the second cumulative probability can be obtained by deducting the first cumulative probability from the cumulative probability of the point cloud distribution of the multiple locations at time k in the second time period using a probability transfer function. In this way, to adapt to the point cloud distribution characteristics of a moving target, the introduction of a transfer function meets the needs of dynamically changing point cloud distributions, improves the responsiveness of the cumulative probability distribution function to the changing state of the moving target, and effectively ensures the effectiveness of the cumulative probability distribution function.

[0026] Optionally, the probability transfer function and the second cumulative probability density distribution function at time k satisfy the following relationship:

[0027] Among them, q k (x, y) is the second cumulative probability density distribution function at time k, δ(·) is the Dirac generalized function, Q{(x i ,y i )} represents the final cumulative probability of the distribution point cloud of the i-th position point among the multiple position points in the second time period, (x i ,y i ) represents the position coordinates of the i-th position point among the multiple position points within the second time period; T k|k-1 (x, y) is the probability transfer function.

[0028] Optionally, the probability transfer function is related to at least one of the target object's posture, velocity, or orientation. Thus, determining a corresponding probability transfer function based on at least one of the target object's posture, velocity, or orientation can better meet the requirements of dynamically changing point cloud distributions and further improve the responsiveness of the cumulative probability distribution function to the changing state of the moving target.

[0029] In one possible design, the method further includes: determining a first probability of a point cloud distribution for the i-th position among the plurality of positions at a first moment; determining a second probability of a point cloud distribution for the i-th position within a second time period; and evaluating the stability of the point cloud distribution for the i-th position within the first time period based on a difference between the first probability and the second probability; wherein the difference is positively correlated with the stability of the point cloud distribution for the i-th position within the first time period. In this design, the point cloud distribution stability of a position on the surface of the target object can be evaluated, and if the point cloud distribution for that position is unstable, the point cloud data for that position can be recollected.

[0030] In a second aspect, an embodiment of the present application further provides a point cloud evaluation device. As an example, the device includes: an acquisition module for acquiring a point cloud set of a target object, the point cloud set including point cloud data of multiple position points on the surface of the target object collected within a first time period; a processing module for constructing a sample probability density distribution function of the point cloud set, the sample probability density distribution function is used to describe the probability of multiple position points distributing point clouds at the first moment; the processing module is also used to construct a cumulative probability density distribution function of the point cloud set, the cumulative probability density distribution function is used to describe the probability of multiple position points distributing point clouds within a second time period; wherein, the first time period includes the second time period, the first moment is any moment within the second time period, or the interval between the first moment and any moment within the second time period is less than a preset time period; the processing module is also used to evaluate the stability of the point cloud set based on the similarity between the sample probability density distribution function and the cumulative probability density distribution function; wherein the similarity is positively correlated with the stability of the point cloud set.

[0031] In one possible design, the processing module is specifically used to: determine the probability of each position point among multiple position points distributing a point cloud at a first moment, and construct a sample probability density distribution function of the point cloud set based on the probability of each position point distributing a point cloud at the first moment.

[0032] In one possible design, the sample probability density distribution function of the point cloud set is as follows:

[0033] Among them, p(x, y) is the sample probability density distribution function, δ(·) is the Dirac generalized function, P{(x i ,y i)} represents the probability of the point cloud distributed at the first moment at the i-th position point among multiple position points; (x i ,y i ) represents the position coordinates corresponding to the i-th position point among multiple position points at the first moment; the value of i is a positive integer from 1 to M, and M is the total number of position points.

[0034] In one possible design, the second time period includes K moments; the processing module is specifically used to: determine the first cumulative probability density distribution function at moment k-1; wherein the value of k is greater than 1 and less than K; determine the second cumulative probability density distribution function at moment k based on the first cumulative probability density distribution function and the cumulative probability of the point cloud distributed by multiple position points at moment k in the second time period; when the difference between the first cumulative probability density distribution function and the second cumulative probability density distribution function is greater than or equal to a second threshold, determine the second cumulative probability density distribution function at moment k+1 based on the second cumulative probability density distribution function and the cumulative probability of the point cloud distributed by the multiple position points at moment k+1 in the second time period; or, when the difference between the first cumulative probability density distribution function and the second cumulative probability density distribution function is less than the second threshold, use the first cumulative probability density distribution function as the cumulative probability density distribution function of the point cloud set.

[0035] In one possible design, the cumulative probability density distribution function of the point cloud set is as follows:

[0036] in, Q″{(x i ,y i )}=Q′{(x i ,y i )}+N;

[0037] Among them, q(x, y) is the cumulative probability density distribution function, δ(·) is the Dirac generalized function, Q{(x i ,y i )} represents the final cumulative probability of the distribution point cloud of the i-th position point among multiple position points from time 0 to time k in the second time period, (x i ,y i ) represents the position coordinates of the i-th position point in the second time period; Q′{(x i ,y i )} is used to represent the cumulative probability of the point cloud of the i-th position point in the second time period at the k-1 moment, Q″{(x i ,y i )} is used to represent the cumulative probability of the distribution point cloud of the i-th position point among multiple position points at time k in the second time period; N is a constant.

[0038] In one possible design, the processing module is further used to: determine the relative divergence KLD between the sample probability density distribution function and the cumulative probability density distribution function; based on the KLD, determine the similarity between the sample probability density distribution function and the cumulative probability density distribution function; wherein, the KLD is negatively correlated with the similarity.

[0039] In one possible design, KLD, sample probability density distribution function, and cumulative probability density distribution function satisfy the following relationship:

[0040] Among them, D KL (p||q) is KLD, p(x, y) is the sample probability density distribution function, and q(x, y) is the cumulative probability density distribution function.

[0041] In one possible design, the target object is a moving target; the multiple position points include a first position point and a second position point, and the second position point is a position point other than the first position point among the multiple position points; the cumulative probability of the point cloud distributed at the k moment of the second time period of the multiple position points includes a first cumulative probability and a second cumulative probability; wherein the first cumulative probability is the cumulative probability of the point cloud distributed at the k moment of the second time period of the first position point, and the second cumulative probability is the cumulative probability of the point cloud distributed at the k moment of the second time period; accordingly, the processing module is specifically used to: when the first cumulative probability is less than a second threshold, delete the first cumulative probability from the cumulative probability of the point cloud distributed at the k moment of the multiple position points in the second time period to obtain the second cumulative probability; and determine the second cumulative probability density distribution function at the k moment based on the second cumulative probability.

[0042] Optionally, the second cumulative probability may be obtained by deleting the first cumulative probability from the cumulative probabilities of the point clouds distributed at the multiple locations at time k in the second time period through a probability transfer function.

[0043] Optionally, the probability transfer function and the second cumulative probability density distribution function at time k satisfy the following relationship:

[0044] Among them, q k (x, y) is the second cumulative probability density distribution function at time k, δ(·) is the Dirac generalized function, Q{(x i ,y i )} represents the final cumulative probability of the distribution point cloud of the i-th position point among the multiple position points in the second time period, (x i ,y i ) represents the position coordinates of the i-th position point among the multiple position points within the second time period; T k|k-1 (x, y) is the probability transfer function.

[0045] Optionally, the probability transfer function is related to at least one of a posture, a velocity, or an orientation of the target object.

[0046] In one possible design, the processing module is further used to: determine a first probability of a point cloud distribution at the i-th position point among multiple position points at a first moment; determine a second probability of a point cloud distribution at the i-th position point within a second time period; and evaluate the point cloud distribution stability of the i-th position point within the first time period based on a difference between the first probability and the second probability; wherein the difference is positively correlated with the distribution stability of the point cloud of the i-th position point.

[0047] In a third aspect, an embodiment of the present application provides a point cloud evaluation device, comprising at least one processor and an interface, wherein the processor is used to call and run a computer program from the interface. When the processor executes the computer program, the method described in the first aspect or any possible design of the first aspect can be implemented.

[0048] In a fourth aspect, embodiments of the present application provide a terminal comprising the apparatus described in the second aspect. Optionally, the terminal may be an onboard device, a vehicle, a monitoring controller, an unmanned aerial vehicle (UAV), a robot, a roadside unit, or the like. Alternatively, the terminal may be a smart device for smart home, smart manufacturing, or other applications requiring target detection or tracking.

[0049] In the fifth aspect, an embodiment of the present application provides a chip system, which includes: a processor and an interface, wherein the processor is used to call and run a computer program from the interface. When the processor executes the computer program, it can implement the method described in the above-mentioned first aspect or any possible design of the first aspect.

[0050] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program for executing the method described in the first aspect or any possible design of the first aspect.

[0051] In the seventh aspect, an embodiment of the present application also provides a computer program product, including a computer program or instructions, which, when executed, can implement the method described in the above-mentioned first aspect or any possible design of the first aspect.

[0052] The technical effects that can be achieved in the second to seventh aspects mentioned above can be referred to the technical effects that can be achieved in the first aspect mentioned above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] FIG1 is a schematic diagram of a point cloud envelope provided in an embodiment of the present application;

[0054] FIG2 is a schematic diagram of a point cloud evaluation scenario provided by an embodiment of the present application;

[0055] FIG3 is a schematic diagram of a point cloud data collection scenario provided by an embodiment of the present application;

[0056] FIG4 is a flow chart of a point cloud evaluation method according to an embodiment of the present application;

[0057] FIG5 is a schematic diagram of another point cloud evaluation scenario provided by an embodiment of the present application;

[0058] FIG6 is a third flow chart of the point cloud evaluation method provided in an embodiment of the present application;

[0059] FIG7 is a fourth flow chart of the point cloud evaluation method provided in an embodiment of the present application;

[0060] FIG8 is a schematic diagram of a point cloud evaluation device according to an embodiment of the present application;

[0061] FIG9 is a second schematic diagram of the point cloud evaluation device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

[0063] Below, some terms used in the embodiments of the present application are explained to facilitate understanding by those skilled in the art.

[0064] 1) Point cloud: A collection of point data on an object's surface measured by a detection device is called a point cloud. A point cloud is a collection of points obtained by obtaining the spatial coordinates of each sampling point on the target object's surface. It is also known as a massive point collection of the target object's surface characteristics.

[0065] In some embodiments, a point cloud is a set of vectors in a three-dimensional coordinate system. These vectors are usually expressed in the form of x, y, and z three-dimensional coordinates, and are generally used to represent the outer surface shape of a target object. In addition, in addition to the geometric position information represented by (x, y, z), a point cloud can also represent the color information of a point in red, green, blue (RGB) format, grayscale value, depth, object reflection surface intensity, etc. In other embodiments, a point cloud can be represented in the form of two-dimensional coordinates in a two-dimensional coordinate system, and the two-dimensional coordinates can be used to represent the spatial position coordinates of the point cloud. In the embodiments of the present application, the spatial position coordinates of the point cloud are mainly introduced by taking the two-dimensional coordinates of the point cloud, i.e., (x, y) coordinates, as an example.

[0066] 2) Point cloud envelope: This effectively simplifies the edge of a point cloud and can greatly assist in subsequent operations such as target tracking. Generally speaking, the main forms of point cloud envelopes include convex polygons (also called convex polygon envelopes), ellipses (also called elliptical envelopes), and bounding boxes (also called bounding boxes, also called bounding box envelopes). Point cloud envelopes can include convex polygon envelopes (as shown in Figure 1 (a)), elliptical envelopes (as shown in Figure 1 (b)), and bounding box envelopes (as shown in Figure 1 (c)). A bounding box can be understood as an imaginary outer frame surrounding the object being detected. In digital image processing, a bounding box can be the coordinates of a rectangular frame that completely encloses a digital image when placed on a page, canvas, screen, or other similar two-dimensional background. Simply put, a bounding box approximates a complex geometric object using a slightly larger geometric body with simpler characteristics.

[0067] 3) Ground truth frame: The target detection frame used by the detection device when collecting the target object's point cloud. This ground truth frame is obtained by smoothing the target detection frame generated by the high-precision detection device during target detection based on historical detection information. It should be understood that a high-precision detection device can be understood as a detection device with stronger angle and range measurement capabilities than the current detection device. For example, if the current detection device is a millimeter-wave radar, the high-precision detection device can be a lidar.

[0068] 4) A sample probability density distribution function, which describes the probability of a point cloud being distributed at a first spatial location at a first moment. Here, "first moment" can be understood as any moment when the detection device collects a point cloud of the target object.

[0069] 5) A cumulative probability density distribution function, which is used to describe the probability of distributing a point cloud at a first spatial location within a first time period. The "first time period" can be understood as any time period during which the detection device collects point clouds of the target object.

[0070] 6) Relative Divergence (KLD), also known as relative entropy, describes the similarity between the sample probability density distribution function and the cumulative probability density distribution function. In this embodiment, the smaller the KLD, the greater the similarity between the sample probability density distribution function and the cumulative probability density distribution function, and the more stable the point cloud distribution of the target object.

[0071] 7) Point cloud target tracking can be understood as the detection device tracking the target object's motion trajectory by collecting the target object's point cloud.

[0072] 8) Target detection can be understood as the detection device detecting the specific position and / or speed of the target object.

[0073] 9) Stationary targets are defined as objects whose position and orientation do not change, such as buildings on a road.

[0074] 10) Moving targets can be understood as target objects whose position and direction may change, for example, a moving vehicle.

[0075] 11) An evaluation device, which may be deployed in a detection device, such as a sensor, such as a radar (e.g., a light detection and ranging (lidar) or a millimeter-wave radar). Alternatively, the sensor may be a sensor mounted on a radar and used to collect a point cloud of a target object.

[0076] 12) In the embodiments of the present application, "at least one" refers to one or more, and "more" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: the existence of A alone, the existence of A and B at the same time, and the existence of B alone, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or plural.

[0077] Furthermore, unless otherwise indicated, ordinal numbers such as "first" and "second" in the embodiments of this application are used to distinguish multiple objects and are not used to limit the size, shape, content, sequence, timing, priority, or importance of the multiple objects. For example, the first moment and the second moment are only used to distinguish different moments and do not indicate a difference in priority or importance between the two moments.

[0078] The quality of point cloud has a significant impact on the target detection performance or target tracking performance of the detection equipment. Therefore, after the detection equipment collects the point cloud of the target object, the quality of the point cloud needs to be evaluated.

[0079] As shown in Figure 2, in some technical solutions, only the truth box provided by high-precision equipment is used to evaluate the quality of the point cloud. Specifically, in Figure 2, whether a single point cloud is qualified is determined by judging whether the point cloud is distributed within the truth box and the allowable error range around the truth box. If the point cloud is distributed within the truth box or the allowable error range of the truth box, the point cloud is considered qualified; conversely, if the point cloud is not distributed within the truth box or the allowable error range of the truth box, the point cloud is considered unqualified. It can be seen that in this technical solution, it is impossible to evaluate the distribution stability of the point cloud, and it is difficult to judge whether the point cloud meets the point cloud stability requirements of the detection equipment.

[0080] Other technical solutions also use the truth box provided by high-precision equipment to evaluate the quality of the point cloud. Specifically, the quality of the point cloud is evaluated by examining the overlap area or intersection over union (IOU) between the point cloud envelope and the truth box. The larger the overlap area or IOU, the better the point cloud quality; the smaller the overlap area or IOU, the worse the point cloud quality. However, this technical solution still has the problem of being unable to evaluate the distribution stability of the point cloud.

[0081] In summary, the current point cloud quality assessment method is too rough and simple, unable to evaluate the distribution stability of the point cloud, and difficult to determine whether the point cloud meets the point cloud stability requirements of the detection equipment.

[0082] In view of this, the present application provides a point cloud evaluation method, in which a sample probability density distribution function and a cumulative probability density distribution function of a point cloud set of a target object are constructed, and based on the similarity between the sample probability density distribution function and the cumulative probability density distribution function, the distribution stability of the point cloud of the target object is evaluated. In this way, it is helpful to understand the distribution stability of the point cloud collected by the detection device at different times, and then the detection device can optimize and adjust the target detection algorithm according to the distribution characteristics of the point cloud set, thereby improving the target tracking performance or target detection performance of the detection device. Alternatively, when the point cloud distribution is unstable, a detection device with stronger ranging and / or angle measurement capabilities can be used to collect the point cloud of the target object, thereby achieving accurate positioning and identification of the target object.

[0083] It should be understood that the detection device in the embodiment of the present application can be understood as a three-dimensional point cloud scanning device with a point cloud acquisition function, such as a radar.

[0084] The technical solutions provided in the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0085] First, the application scenarios to which the embodiments of the present application are applicable are introduced.

[0086] For example, Figure 3 shows a schematic diagram of an application scenario of a point cloud assessment method provided by an embodiment of the present application. In Figure 3, the detection device takes a laser radar 100 as an example. The laser radar 100 can be installed at the front of the vehicle. During the vehicle's driving process, the laser radar 100 can emit a laser signal. After the laser signal hits the object in front of the vehicle, it will be reflected by the target object (such as pedestrian a, road facility b, vehicle c, etc.), and the reflected echo signal can be received by the laser radar 100, and then the laser radar 100 can form a point cloud of the target object based on the echo signal.

[0087] Furthermore, the lidar 100 can construct a sample probability density distribution function and a cumulative probability density distribution function based on the point cloud of the target object. The sample probability density distribution function is used to describe the probability of the point cloud being distributed at a first spatial location at a first moment, and the cumulative probability density distribution function is used to describe the probability of the point cloud being distributed at the first spatial location within a first time period. Furthermore, the lidar 100 can evaluate the distribution stability of the target object's point cloud based on the similarity between the sample probability density distribution function and the cumulative probability density distribution function. This helps understand the distribution stability of the point clouds collected by the lidar 100 at different moments, and the lidar 100 can then optimize and adjust the target detection algorithm based on the distribution characteristics of the point cloud, thereby improving the target tracking performance or target detection performance of the lidar 100.

[0088] It should be understood that the above-mentioned laser radar 100 can be one of a mechanical laser radar, a liquid laser radar, a pure solid-state laser radar or a hybrid solid-state laser radar (also called a semi-solid-state laser radar), or it can also be other types of laser radars, and the embodiments of the present application do not make specific limitations on this.

[0089] In other possible embodiments, the laser radar 100 can also be installed on a ship or a machine.

[0090] The above introduces the application scenarios applicable to the embodiments of the present application. The following introduces the point cloud evaluation method provided by the embodiments of the present application.

[0091] FIG4 exemplarily illustrates a flow chart corresponding to a point cloud evaluation method provided in an embodiment of the present application, which is applicable to a point cloud evaluation device (e.g., the laser radar 100 shown in FIG3 ). As shown in FIG4 , the method includes:

[0092] S401, obtaining a point cloud set of a target object.

[0093] In an embodiment of the present application, the point cloud set includes point cloud data of a plurality of locations on the surface of the target object collected during a first time period. The first time period can be understood as the time period during which the point cloud data of the target object is collected. For example, the first time period can be any one of 1 hour, 30 minutes, 15 minutes, etc.

[0094] In the embodiment of the present application, obtaining the point cloud of the target object includes but is not limited to the following implementations:

[0095] In embodiment 1, the point cloud evaluation device has a point cloud acquisition function, and the point cloud evaluation device can directly scan the target object to obtain the point cloud of the target object.

[0096] In implementation mode 2, the point cloud evaluation device does not have a point cloud acquisition function. The point cloud evaluation device may initiate a request to a server; the server receives the request and sends the point cloud of the target object to the point cloud evaluation device.

[0097] In implementation mode 3, the point cloud evaluation device does not have a point cloud acquisition function. The point cloud evaluation device can initiate a request to the server; the server receives the request and sends an acquisition instruction to the detection device, and then the detection device obtains the point cloud of the target object according to the acquisition instruction and sends the point cloud of the target object to the point cloud evaluation device.

[0098] S402, constructing a sample probability density distribution function of the point cloud set.

[0099] The sample probability density distribution function is used to describe the probability of a point cloud being distributed at a first moment by multiple location points.

[0100] Accordingly, in one possible embodiment, the point cloud evaluation device may construct a sample probability density distribution function for a point cloud by determining the probability of a point cloud being distributed at each of a plurality of locations at a first moment, and constructing a sample probability density distribution function for the point cloud set based on the probability of a point cloud being distributed at each location at the first moment. In this way, the probability of a point cloud being distributed at each spatial location at the first moment can be calculated.

[0101] Optionally, the sample probability density distribution function of the point cloud constructed by the point cloud evaluation device may be as follows:

[0102] Wherein, p(x, y) is the sample probability density distribution function, δ(·) is the Dirac generalized function, and P{·} represents the probability of the i-th spatial position among the multiple spatial positions distributing the point cloud at the first moment; (x i ,y i ) represents the position coordinates of the i-th spatial position among the multiple spatial positions at the first moment; the value of i is a positive integer from 1 to M, and M is the total number of position points.

[0103] Optionally, when determining the probability of a point cloud being distributed at the first moment for each of a plurality of position points, the point cloud set can be rasterized to form a grid map comprising a plurality of grids, and one grid in the grid map corresponds to one of the plurality of position points, and the probability of a point cloud being distributed at the first moment for each grid in the grid map can be further statistically analyzed.

[0104] For example, please refer to Figure 5. After the point cloud evaluation device obtains the point cloud of the target object, it rasterizes the point cloud according to the position information of the target object to obtain multiple grids (one grid corresponds to one position point) to form a corresponding grid map; further, the point cloud evaluation device can also obtain the true value frame provided by the high-precision detection equipment, and screen the multiple grids in the grid map based on the true value frame and the error allowable range of the truth value frame; then, the point cloud evaluation device counts the probability of the point cloud distributed in each of the multiple grids within the true value frame and the error allowable range of the truth value frame, and then combines the Dirac generalized function to determine the sample probability density distribution function of each grid (that is, determine the sample probability density distribution function of each position point among the multiple position points).

[0105] For example, the point cloud of the target object includes point cloud 1, point cloud 2, and point cloud 3; the position coordinates of position point 1 are (x1, y1), the position coordinates of position point 2 are (x2, y2), and the position coordinates of position point 3 are (x3, y3). Point cloud 1 is distributed in position point 1 at the first moment, point cloud 2 is distributed in position point 2 at the first moment, and point cloud 3 is distributed in position point 3 at the first moment. It is considered that the probability of position point 1 distributing the point cloud at the first moment is 1, the probability of position point 2 distributing the point cloud at the first moment is 1, and the probability of position point 3 distributing the point cloud at the first moment is 1. Furthermore, the point cloud evaluation device can determine the sample probability density distribution function of position point 1, position point 2, and position point 3 in combination with the values ​​of position point 1, position point 2, and position point 3 in the Dirac generalized function.

[0106] S403: Construct a cumulative probability density distribution function of the point cloud set.

[0107] In an embodiment of the present application, the cumulative probability density distribution function is used to describe the probability of a point cloud being distributed at multiple locations within a second time period. The first time period includes the second time period, and the first moment is any moment within the second time period. Alternatively, the first time period includes the first moment and the second time period, and the interval between the first moment and any moment within the second time period is less than a preset duration. The "second time period" can be understood as any time period within the first time period, and the preset duration can be, for example, 3 minutes or 5 minutes, and the embodiment of the present application does not impose any specific limitation thereto.

[0108] In an embodiment of the present application, when the above-mentioned second time period includes K moments, the process of the point cloud evaluation device constructing the cumulative probability density distribution function of the point cloud can be: determining the first cumulative probability density distribution function at moment k-1; wherein the value of k is greater than 1 and less than the value of K; determining the second cumulative probability density distribution function at moment k based on the first cumulative probability density distribution function and the cumulative probability of the point cloud distributed by multiple position points at moment k in the second time period; when the difference between the first cumulative probability density distribution function and the second cumulative probability density distribution function is greater than or equal to the second threshold, determining the second cumulative probability density distribution function at moment k+1 based on the second cumulative probability density distribution function and the cumulative probability of the point cloud distributed by the multiple position points at moment k+1 in the second time period; or, when the difference between the first cumulative probability density distribution function and the second cumulative probability density distribution function is less than the second threshold, using the first cumulative probability density distribution function as the cumulative probability density distribution function of the point cloud set. In this way, the second cumulative probability density distribution function at moment k is determined by combining the cumulative probability of the distribution point cloud of multiple position points at moment k in the second time period and the first cumulative probability density distribution function at moment k-1, so that the cumulative probability density distribution function of the point cloud set finally determined can more accurately reflect the distribution characteristics of the point cloud set, thereby making the stability evaluation of the point cloud set more accurate.

[0109] Optionally, the point cloud evaluation device can determine the difference between the first cumulative probability density distribution function and the second cumulative probability density distribution function by determining the relative divergence (Kullback-Leibler divergence, KLD) between the first cumulative probability density distribution function and the second cumulative probability density distribution function; the larger the KLD, the smaller the difference between the first cumulative probability density distribution function and the second cumulative probability density distribution function.

[0110] In one possible design, the cumulative probability density distribution function of the point cloud set is as follows:

[0111] in, Q″{(x i ,y i )}=Q′{(x i ,y i )}+N;

[0112] Where q(x, y) is the cumulative probability density distribution function, δ(·) is the Dirac generalized function, (x i ,y i ) represents the position coordinates of the i-th position point in the second time period, Q{(x i ,y i)} represents the final cumulative probability of the distribution point cloud of the i-th position point in the multiple position points from time 0 to time k in the second time period; Q′{(x i ,y i )} is used to represent the cumulative probability of the point cloud of the i-th position point in the second time period at the k-1 moment, Q″{(x i ,y i )} is used to represent the cumulative probability of the distribution point cloud of the i-th position point among multiple position points at time k in the second time period; N is a constant.

[0113] Exemplarily, the second time period is 3 minutes, and N is 1. After the point cloud evaluation device obtains the point cloud of the target object within these 3 minutes, it rasterizes the point cloud according to the position information of the target object to obtain multiple grids (one grid corresponds to one spatial position) to form a corresponding grid map; further, the point cloud evaluation device can determine the probability of the point cloud being distributed at different times in each of the multiple grids within these 3 minutes, and then combine the Dirac generalized function to determine the cumulative probability density distribution function of each grid within these 3 minutes (that is, the cumulative probability density distribution function of each spatial position in the multiple spatial positions). For example, the position coordinates of spatial position 1 are (x1, y1), the position coordinates of spatial position 2 are (x2, y2), and the position coordinates of spatial position 3 are (x3, y3); among them, the probability of spatial position 1 distributing a point cloud in the first minute of these 3 minutes is 1 / 3, the probability of spatial position 2 distributing a point cloud in the first moment of these 3 minutes is 1 / 3, and the probability of spatial position 3 distributing a point cloud in the first moment of these 3 minutes is 1 / 3; if spatial position 1, spatial position 2, and spatial position 3 all distribute point clouds in the second moment of these 3 minutes, then the probability of spatial position 1 distributing a point cloud in the second moment of these 3 minutes is (1 / 3+1) / ((1 / 3+1)*3 ), the probability that spatial position 2 distributes a point cloud at the second moment in these 3 minutes is (1 / 3+1) / ((1 / 3+1)*3), and the probability that spatial position 3 distributes a point cloud at the second moment in these 3 minutes is (1 / 3+1) / ((1 / 3+1)*3); if spatial position 1 and spatial position 2 both distribute point clouds at the third moment in these 3 minutes, but spatial position 3 does not distribute a point cloud at the third moment in these 3 minutes, then the probability that spatial position 1 and spatial position 2 distribute point clouds at the third moment in these 3 minutes is (1 / 3+1+1) / ((1 / 3+1+1)*2), and the probability that spatial position 3 distributes a point cloud at the third moment in the third minute of these 3 minutes is 0.

[0114] For ease of understanding, FIG6 shows another flow chart of the point cloud evaluation method provided by an embodiment of the present application. In FIG6, the cumulative probability density distribution function at a moment corresponds to a heat map; accordingly, as shown in FIG6, the process of the point cloud evaluation device constructing the cumulative probability density distribution function of the point cloud set includes: rasterizing the point cloud set to obtain a grid map; performing statistics on the probability of each grid distribution point cloud in the grid map to form a heat map at moment k-1 and a heat map at moment k; determining the KLD between the heat map at moment k-1 and the heat map at moment k, and when the KLD is less than a preset threshold, using the heat map at moment k-1 as the cumulative probability density distribution function of the point cloud set; further, combining the sample probability density distribution function at the first moment and the cumulative probability density distribution function of the point cloud set, the stability of the point cloud set can be evaluated.

[0115] S404 : Evaluate the stability of the point cloud set based on the similarity between the sample probability density distribution function and the cumulative probability density distribution function.

[0116] In the present embodiment, the term "point cloud stability" refers to the distribution of point clouds across multiple locations on the surface of a target object. A more stable point cloud is defined as the consistency of the point cloud distribution across these locations at different times within the same time period. A stable point cloud means the target object is a stable reflection source, resulting in a high-quality point cloud that meets the stability requirements of point cloud data used by detection equipment for target detection or tracking.

[0117] It should be understood that when the point cloud evaluation device determines the similarity between the sample probability density distribution function and the cumulative probability density distribution function, it compares the sample probability density distribution function and the cumulative probability density distribution function of multiple position points (i.e., the same position range).

[0118] In one possible embodiment, before evaluating the distribution stability of the point cloud set based on the similarity between the sample probability density distribution function and the cumulative probability density distribution function of the point cloud set, the method further includes: determining the relative divergence KLD between the sample probability density distribution function and the cumulative probability density distribution function of the point cloud set; and based on the KLD, determining the similarity between the sample probability density distribution function and the cumulative probability density distribution function of the point cloud set; wherein KLD is negatively correlated with the similarity. The negative correlation between KLD and similarity can be understood as the smaller the KLD, the greater the similarity; accordingly, the better the stability of the point cloud set. In this way, the similarity evaluation of the sample probability density distribution function and the cumulative probability density distribution function of the point cloud set can be made more accurate, thereby making the stability evaluation of the point cloud set based on this similarity more accurate.

[0119] Optionally, the KLD, the sample probability density distribution function of the point cloud set, and the cumulative probability density distribution function of the point cloud set satisfy the following relationship:

[0120] Among them, D KL (p||q) is KLD, p(x, y) is the sample probability density distribution function, and q(x, y) is the cumulative probability density distribution function; the smaller the KLD, the greater the similarity between the sample probability density distribution function and the cumulative probability density distribution function, and the better the stability of the point cloud set of the target object.

[0121] It should be understood that the above formula 3 is only an example, and the above formula 3 can also be modified as follows:

[0122] Here, c can be a constant.

[0123] The point cloud distribution characteristics of moving targets also change dynamically. The continuous accumulation of point clouds by the point cloud evaluation device will cause significant tailing, resulting in statistical deviations in the cumulative distribution. Therefore, in the embodiments of this application, taking into account the changing state of the moving target, if a spatial location has no distributed point cloud counted for an extended period of time, the probability of a distributed point cloud at that location is low, and therefore the probability corresponding to that location needs to be excluded from the cumulative probability distribution.

[0124] In one possible embodiment, when the target object is a moving target, the multiple location points include a first location point and a second location point, the second location point being a location point other than the first location point in the multiple location points; the cumulative probability of the point cloud distributed at time k in the second time period of the multiple location points includes a first cumulative probability and a second cumulative probability; wherein the first cumulative probability is the cumulative probability of the point cloud distributed at time k in the second time period, and the second cumulative probability is the cumulative probability of the point cloud distributed at time k in the second time period; accordingly, determining the second cumulative probability density distribution function at time k based on the first cumulative probability density distribution function and the cumulative probability of the point cloud distributed at time k in the second time period includes: when the first cumulative probability is less than a second threshold, deleting the first cumulative probability from the cumulative probability of the point cloud distributed at time k in the multiple location points in the second time period to obtain a second cumulative probability; and determining the second cumulative probability density distribution function based on the second cumulative probability. In this way, the probability corresponding to the first location point (i.e., the first cumulative probability) is excluded from the cumulative probability distribution, thereby improving the response of the cumulative probability distribution function to the changing state of the moving target and effectively ensuring the validity of the cumulative probability distribution function.

[0125] Optionally, the point cloud evaluation device may utilize a probability transfer function to remove the first cumulative probability from the cumulative probability of the point cloud distribution of the multiple locations at time k in the second time period to obtain the second cumulative probability. Thus, by introducing a transfer function to adapt to the point cloud distribution characteristics of a moving target, the need for dynamic changes in the point cloud distribution is met, the cumulative probability distribution function's responsiveness to the changing state of the moving target is improved, and the effectiveness of the cumulative probability distribution function is effectively ensured.

[0126] Optionally, the probability transfer function and the second cumulative probability density distribution function at time k satisfy the following relationship:

[0127] Among them, q k (x, y) is the second cumulative probability density distribution function at time k, δ(·) is the Dirac generalized function, Q{(x i ,y i )} represents the final cumulative probability of the distribution point cloud of the i-th position point among the multiple position points in the second time period, (x i ,y i ) represents the position coordinates of the i-th position point among the multiple position points within the second time period; T k|k-1 (x, y) is the probability transfer function.

[0128] For example, the transfer function may be as follows: k|k-1 (x, y) = e -τ(x，y)t

[0129] Optionally, the probability transfer function is related to at least one of the target object's posture, velocity, or orientation. Thus, determining the corresponding probability transfer function based on at least one of the target object's posture, velocity, or orientation can better meet the requirements of dynamically changing point cloud distributions and further improve the response of the cumulative probability distribution function to the changing state of the moving target.

[0130] For example, FIG7 shows a schematic diagram of evaluating the point cloud of a moving target. After the point cloud evaluation device obtains the point cloud set of the target object, it rasterizes the point cloud set of the target object based on the target true value coordinate system (such as the target vehicle body coordinate system) provided by the high-precision detection equipment; the point cloud from time 0 to time k-1 is accumulated on the same heat map in a binary filtering manner, and each grid corresponds to the probability of a distributed point cloud, and forms a cumulative heat map at time k-1; further, the point cloud evaluation device transfers (i.e., excludes) the probability of some grids based on the probability transfer function, and continuously accumulates to obtain the cumulative heat map at time k, and uses the cumulative heat map at time k as the cumulative probability density distribution function; then, the point cloud evaluation device evaluates the stability of the point cloud set of the target object based on the cumulative probability density distribution function and the sample probability density distribution function.

[0131] In some other embodiments, the point cloud evaluation device may further determine a first probability of a point cloud distribution at the i-th position among multiple positions on the surface of the target object at a first moment; determine a second probability of a point cloud distribution at the i-th position within a second time period; and evaluate the stability of the point cloud distribution at the i-th position within the first time period based on the difference between the first probability and the second probability; wherein the difference is positively correlated with the stability of the point cloud distribution at the i-th position within the first time period. In this way, the point cloud distribution stability of a position on the surface of the target object can be evaluated, and if the point cloud distribution at the position is unstable, the point cloud data for the position can be recollected.

[0132] In some other embodiments, the point cloud evaluation device can also evaluate the speed change characteristics of the point cloud of the target object, making the quality evaluation of the point cloud more comprehensive, and then intuitively judging whether the speed stability of the point cloud meets the speed stability requirements of the detection equipment, which helps to improve the target detection performance and target tracking performance of the detection equipment.

[0133] For example, after the point cloud evaluation device acquires the point cloud of the target object, it can construct the sample probability density distribution function and cumulative probability density distribution function of the point cloud as shown below: Q″{(x i ,y i , v i )}=Q′{(x i ,y i , v i )}+N;

[0134] Wherein, p(x, y, v) is the sample probability density distribution function, δ(·) is the Dirac generalized function, and P{·} represents the probability of the i-th spatial position among the multiple spatial positions distributing the point cloud at the first moment; (x i ,y i ) represents the position coordinate of the i-th spatial position among the multiple spatial positions at the first moment; v i represents the velocity of the point cloud at the i-th spatial position among the multiple spatial positions at the first moment;

[0135] Wherein, q(x, y, v) is the cumulative probability density distribution function, Q{·} represents the probability that the i-th spatial position among the multiple spatial positions distributes the point cloud within the first time period, (x i ,y i , v i ) represents the position coordinates and velocity of the i-th spatial position in the plurality of spatial positions within the first time period; Q′{(x i ,y i , v i )} is used to represent the probability of the i-th spatial position among the multiple spatial positions distributing the point cloud at the second moment in the first time period, Q″{(x i ,y i , v i )} is used to represent the probability of the point cloud distribution of the i-th spatial position among the multiple spatial positions at the third moment in the first time period, and the second moment is an adjacent moment of the third moment; N is a constant.

[0136] Similarly, the point cloud evaluation device can evaluate the velocity stability of the point cloud based on the similarity between the sample probability density distribution function and the cumulative probability density distribution function. In one possible embodiment, the point cloud evaluation device can determine the relative divergence KLD between the sample probability density distribution function and the cumulative probability density distribution function according to the following formula:

[0137] Among them, D KL (p||q) is KLD, p(x, y, v) is the sample probability density distribution function, and q(x, y, v) is the cumulative probability density distribution function; the smaller the KLD, the greater the similarity between the sample probability density distribution function and the cumulative probability density distribution function, and the better the speed stability of the point cloud of the target object.

[0138] According to the aforementioned method, FIG8 is a schematic diagram of the structure of a point cloud evaluation device provided in an embodiment of the present application. As shown in FIG8 , the device can be an electronic device, or a chip or circuit, such as a chip or circuit that can be provided in an electronic device. The electronic device can be a device with processing capabilities, such as the laser radar described in FIG3 .

[0139] Furthermore, the point cloud evaluation device 801 may further include a bus system, wherein the processor 802 , the memory 804 , and the transceiver 803 may be connected via the bus system.

[0140] It should be understood that the processor 802 may be a chip. For example, the processor 802 may be a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on chip (SoC), a central processor unit (CPU), a network processor (NP), a digital signal processor (DSP), a microcontroller unit (MCU), a programmable logic device (PLD), or other integrated chips.

[0141] During implementation, each step of the above method can be completed by hardware integrated logic circuits in processor 802 or by software instructions. The steps of the method disclosed in conjunction with the embodiments of the present application can be directly embodied as being executed by a hardware processor, or can be executed by a combination of hardware and software modules in processor 802. The software module can be located in a storage medium mature in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in memory 804, and processor 802 reads the information in memory 804 and, in conjunction with its hardware, completes the steps of the above method.

[0142] It should be noted that the processor 802 in the embodiments of the present application can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method embodiment can be completed by hardware integrated logic circuits in the processor or software instructions. The above processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component. The various methods, steps, and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of the present application can be directly embodied as being executed by a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium mature in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0143] It is understood that the memory 804 in the embodiment of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus RAM (DR RAM). It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0144] When the point cloud evaluation device 801 corresponds to an electronic device, the point cloud evaluation device may include a processor 802, a transceiver 803, and a memory 804. The memory 804 is used to store instructions, and the processor 802 is used to execute the instructions stored in the memory 804 to implement the relevant solutions in the above method.

[0145] [Corrected on 29.03.2024 according to Rule 91] Based on the above embodiments and the same concept, the embodiments of the present application further provide a point cloud evaluation device, as shown in Figure 9, the device 900 includes: an acquisition module 901, used to acquire a point cloud set of a target object, the point cloud set including point cloud data of multiple position points on the surface of the target object collected within a first time period; a processing module 902, used to construct a sample probability density distribution function of the point cloud set, the sample probability density distribution function is used to describe the probability of the point cloud being distributed at the first moment by the multiple position points; and to construct a cumulative probability density distribution function of the point cloud set, the cumulative probability density distribution function is used to describe the probability of the point cloud being distributed at the multiple position points within a second time period; wherein the first time period includes the second time period, the first moment is any moment within the second time period, or the interval between the first moment and any moment within the second time period is less than a preset time period; further, the processing module 902 is further used to evaluate the stability of the point cloud set based on the similarity between the sample probability density distribution function and the cumulative probability density distribution function; wherein the similarity is positively correlated with the stability of the point cloud distribution.

[0146] In one possible design, the processing module 902 is specifically used to determine the probability of each of the multiple position points distributing the point cloud at the first moment, and based on the probability of each position point distributing the point cloud at the first moment, construct a sample probability density distribution function of the point cloud set.

[0147] In one possible design, the sample probability density distribution function of the point cloud set is as follows:

[0148] Among them, p(x, y) is the sample probability density distribution function, δ(·) is the Dirac generalized function, P{(x i ,y i )} represents the probability of the point cloud distributed at the first moment at the i-th position point among multiple position points; (x i ,y i ) represents the position coordinates corresponding to the i-th position point among multiple position points at the first moment; the value of i is a positive integer from 1 to M, and M is the total number of position points.

[0149] In one possible design, the second time period includes K moments; the processing module 902 is specifically used to: determine the first cumulative probability density distribution function at moment k-1; wherein the value of k is greater than 1 and less than the value of K; determine the second cumulative probability density distribution function at moment k based on the first cumulative probability density distribution function and the cumulative probability of the point cloud distributed by multiple position points at moment k in the second time period; when the difference between the first cumulative probability density distribution function and the second cumulative probability density distribution function is greater than or equal to a second threshold, determine the second cumulative probability density distribution function at moment k+1 based on the second cumulative probability density distribution function and the cumulative probability of the point cloud distributed by the multiple position points at moment k+1 in the second time period; or, when the difference between the first cumulative probability density distribution function and the second cumulative probability density distribution function is less than the second threshold, use the first cumulative probability density distribution function as the cumulative probability density distribution function of the point cloud set.

[0150] In one possible design, the cumulative probability density distribution function of the point cloud set is as follows:

[0151] in, Q″{(x i ,y i )}=Q′{(x i ,y i )}+N;

[0152] Among them, q(x, y) is the cumulative probability density distribution function, δ(·) is the Dirac generalized function, Q{(x i ,y i )} represents the final cumulative probability of the distribution point cloud of the i-th position point among multiple position points from time 0 to time k in the second time period, (x i ,y i ) represents the position coordinates of the i-th position point in the second time period; Q′{(x i ,y i )} is used to represent the cumulative probability of the point cloud of the i-th position point in the second time period at the k-1 moment, Q″{(x i ,y i )} is used to represent the cumulative probability of the distribution point cloud of the i-th position point among multiple position points at time k in the second time period; N is a constant.

[0153] In one possible design, the processing module 902 is further used to: determine the relative divergence KLD between the sample probability density distribution function and the cumulative probability density distribution function; based on the KLD, determine the similarity between the sample probability density distribution function and the cumulative probability density distribution function; wherein, the KLD is negatively correlated with the similarity.

[0154] In one possible design, KLD, sample probability density distribution function, and cumulative probability density distribution function satisfy the following relationship:

[0155] Among them, D KL (p||q) is KLD, (x, y) is the sample probability density distribution function, and q(x, y) is the cumulative probability density distribution function.

[0156] In one possible design, the target object is a moving target; the multiple position points include a first position point and a second position point, and the second position point is a position point other than the first position point among the multiple position points; the cumulative probability of the distribution point cloud of the multiple position points at the k moment in the second time period includes a first cumulative probability and a second cumulative probability; wherein the first cumulative probability is the cumulative probability of the distribution point cloud of the first position point at the k moment in the second time period, and the second cumulative probability is the cumulative probability of the distribution point cloud of the second position point at the k moment in the second time period; accordingly, the processing module 902 is specifically used to: when the first cumulative probability is less than the second threshold, delete the first cumulative probability from the cumulative probability of the distribution point cloud of the multiple position points at the k moment in the second time period to obtain the second cumulative probability; and determine the second cumulative probability density distribution function at the k moment based on the second cumulative probability.

[0157] Optionally, the second cumulative probability may be obtained by deleting the first cumulative probability from the cumulative probabilities of the point clouds distributed at the multiple locations at time k in the second time period through a probability transfer function.

[0158] Optionally, the probability transfer function and the second cumulative probability density distribution function at time k satisfy the following relationship:

[0159] Among them, q k (x, y) is the second cumulative probability density distribution function at time k, δ(·) is the Dirac generalized function, Q{(x i ,y i )} represents the final cumulative probability of the distribution point cloud of the i-th position point among the multiple position points in the second time period, (x i ,y i ) represents the position coordinates of the i-th position point among the multiple position points within the second time period; T k|k-1 (x, y) is the probability transfer function.

[0160] Optionally, the probability transfer function is related to at least one of a posture, a velocity, or an orientation of the target object.

[0161] In one possible design, the processing module 902 is further used to: determine a first probability of a point cloud distribution at the i-th position point among multiple position points at a first moment; determine a second probability of a point cloud distribution at the i-th position point within a second time period; and evaluate the point cloud distribution stability of the i-th position point within the first time period based on a difference between the first probability and the second probability; wherein the difference is positively correlated with the distribution stability of the point cloud of the i-th position point.

[0162] According to the method provided in an embodiment of the present application, an embodiment of the present application provides a point cloud evaluation device, comprising at least one processor and an interface, wherein the processor is used to call and run a computer program from the interface, and when the processor executes the computer program, the above method can be implemented.

[0163] According to the method provided in the embodiments of the present application, the embodiments of the present application provide a terminal including the above-mentioned point cloud evaluation device. Optionally, the terminal can be an onboard device, a vehicle, a monitoring controller, an unmanned aerial vehicle, a robot, a roadside unit, etc. Alternatively, the terminal can be an intelligent device such as a smart home or smart manufacturing device that requires target detection or tracking.

[0164] According to the method provided in an embodiment of the present application, an embodiment of the present application provides a chip system, which includes: a processor and an interface, the processor is used to call and run a computer program from the interface, and when the processor executes the computer program, the above method can be implemented.

[0165] According to the method provided by the embodiment of the present application, the embodiment of the present application provides a computer-readable storage medium, which has a computer program for executing the above method.

[0166] According to the method provided in the embodiment of the present application, the embodiment of the present application also provides a computer program product, including a computer program or instructions, which can implement the above method when the computer program or instructions are executed.

[0167] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0168] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the present application. It should be understood that each flow and / or box in the flow chart and / or block diagram, as well as the combination of the flow chart and / or box in the flow chart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for implementing the functions specified in one or more flow charts and / or one or more boxes in the block diagram.

[0169] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0170] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0171] Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of protection of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. A point cloud evaluation method, characterized in that: The method comprises: Acquire a point cloud set of a target object, the point cloud set comprising point cloud data of a plurality of position points on the surface of the target object collected within a first time period; Constructing a sample probability density distribution function of the point cloud set, wherein the sample probability density distribution function is used to describe the probability of the point clouds of the multiple position points being distributed at a first moment; Constructing a cumulative probability density distribution function of the point cloud set, the cumulative probability density distribution function is used to describe the probability of the point cloud being distributed at the plurality of locations within a second time period; wherein the first time period includes the second time period, the first moment is any moment within the second time period, or the first time period includes the first moment and the second time period, and the interval between the first moment and any moment within the second time period is less than a preset duration; Based on the similarity between the sample probability density distribution function and the cumulative probability density distribution function, the stability of the point cloud set is evaluated; wherein the similarity is positively correlated with the stability.

2. The method according to claim 1, characterized in that Constructing a sample probability density distribution function of the point cloud set, including: Determine the probability of each of the plurality of position points distributing a point cloud at the first moment; Based on the probability of each position point distributing the point cloud at the first moment, a sample probability density distribution function of the point cloud set is constructed.

3. The method according to claim 2, characterized in that The sample probability density distribution function is as follows: Where p(x, y) is the sample probability density distribution function, δ(·) is the Dirac generalized function, P{(x i ,y i )} represents the probability of the i-th position point among the multiple position points distributing the point cloud at the first moment; (x i ,y i ) represents the position coordinates corresponding to the i-th position point among the multiple position points at the first moment; the value of i is a positive integer from 1 to M, and M is the total number of the position points.

4. The method according to claim 1, characterized in that: The second time period includes K moments; Constructing a cumulative probability density distribution function of the point cloud set, including: Determine the first cumulative probability density distribution function at time k-1; wherein the value of k is greater than 1 and less than K; Determine a second cumulative probability density distribution function at time k according to the first cumulative probability density distribution function and the cumulative probability of the distribution point clouds of the plurality of location points at time k in the second time period; When the difference between the first cumulative probability density distribution function and the second cumulative probability density distribution function is greater than or equal to a second threshold, determining a second cumulative probability density distribution function at time k+1 according to the second cumulative probability density distribution function and the cumulative probability of the distribution point cloud of the plurality of position points at time k+1 in the second time period; or, When the difference between the first cumulative probability density distribution function and the second cumulative probability density distribution function is less than a second threshold, the first cumulative probability density distribution function is used as the cumulative probability density distribution function of the point cloud set.

5. The method according to claim 4, characterized in that The cumulative probability density distribution function of the point cloud set is as follows: in, Q″{(x i ,y i )}=Q′{(x i ,y i )}+N; Wherein, q(x, y) is the cumulative probability density distribution function, δ(·) is the Dirac generalized function, Q{·} represents the final cumulative probability of the i-th position point among the multiple position points in the distribution point cloud in the second time period, (x i ,y i ) represents the position coordinates of the i-th position point among the plurality of position points in the second time period; Q′{(x i ,y i )} is used to represent the cumulative probability of the distribution point cloud of the i-th position point among the multiple position points at the k-1 moment in the second time period, Q″{(x i ,y i )} is used to represent the cumulative probability of the distribution point cloud of the i-th position point among the multiple position points at the kth moment in the second time period; N is a constant.

6. The method according to any one of claims 1 to 5, characterized in that: Before evaluating the stability of the point cloud set based on the similarity between the sample probability density distribution function and the cumulative probability density distribution function, the method further includes: Determine the relative divergence KLD between the sample probability density distribution function and the cumulative probability density distribution function; Based on the KLD, the similarity between the sample probability density distribution function and the cumulative probability density distribution function is determined; wherein the KLD is negatively correlated with the similarity.

7. The method according to claim 6, characterized in that The KLD, the sample probability density distribution function, and the cumulative probability density distribution function satisfy the following relationship: Among them, D KL (p||q) is the KLD, p(x, y) is the sample probability density distribution function, and q(x, y) is the cumulative probability density distribution function.

8. The method according to claim 4, characterized in that The target object is a moving target; the multiple position points include a first position point and a second position point, and the second position point is a position point other than the first position point among the multiple position points; the cumulative probability of the point cloud distributed at the k moment of the second time period of the multiple position points includes a first cumulative probability and a second cumulative probability; wherein the first cumulative probability is the cumulative probability of the point cloud distributed at the k moment of the second time period of the first position point, and the second cumulative probability is the cumulative probability of the point cloud distributed at the k moment of the second time period of the second position point; Determining a second cumulative probability density distribution function at time k according to the first cumulative probability density distribution function and the cumulative probability of the distribution point clouds of the plurality of position points at time k in the second time period includes: When the first cumulative probability is less than a second threshold, the first cumulative probability is deleted from the cumulative probability of the distribution point cloud of the multiple position points at time k in the second time period to obtain the second cumulative probability; A second cumulative probability density distribution function at time k is determined according to the second cumulative probability.

9. The method according to claim 8, characterized in that Deleting the first cumulative probability from the cumulative probability of the distribution point cloud of the multiple location points at time k in the second time period to obtain the second cumulative probability includes: The second cumulative probability is obtained by deleting the first cumulative probability from the cumulative probabilities of the point clouds distributed at the multiple locations at time k in the second time period through a probability transfer function.

10. The method according to claim 9, characterized in that The probability transfer function and the second cumulative probability density distribution function at time k satisfy the following relationship: Among them, q k (x, y) is the second cumulative probability density distribution function at time k, δ(·) is the Dirac generalized function, Q{(x i ,y i )} represents the final cumulative probability of the distribution point cloud of the i-th position point among multiple position points from time 0 to time k in the second time period, (x i ,y i ) represents the position coordinates of the i-th position point among the multiple position points in the second time period; T k|k-1 (x, y) is the probability transfer function.

11. The method according to claim 9 or 10, characterized in that: The probability transfer function is related to at least one of a posture, a velocity, or an orientation of the target object.

12. The method according to any one of claims 1 to 11, characterized in that: The method further comprises: Determine a first probability that an i-th position point among the plurality of position points distributes a point cloud at the first moment; Determine a second probability that the i-th position point distributes the point cloud within the second time period; Based on the difference between the first probability and the second probability, the point cloud distribution stability of the i-th position point in the first time period is evaluated; wherein the difference is positively correlated with the distribution stability of the point cloud of the i-th position point.

13. A point cloud evaluation device, characterized in that: The device comprises: An acquisition module, configured to acquire a point cloud set of a target object, wherein the point cloud set includes point cloud data of a plurality of position points on the surface of the target object collected within a first time period; A processing module, used to construct a sample probability density distribution function of the point cloud set, wherein the sample probability density distribution function is used to describe the probability of the point cloud being distributed at the first moment by the plurality of position points; The processing module is further used to construct a cumulative probability density distribution function of the point cloud set, and the cumulative probability density distribution function is used to describe the probability of the point cloud being distributed at the plurality of positions within a second time period; wherein the first time period includes the second time period, and the first moment is any moment within the second time period, or the first time period includes the first moment and the second time period, and the interval between the first moment and any moment within the second time period is less than a preset duration; The processing module is further used to evaluate the stability of the point cloud set based on the similarity between the sample probability density distribution function and the cumulative probability density distribution function; wherein the similarity is positively correlated with the stability.

14. The device according to claim 13, characterized in that The processing module is specifically used for: Determine the probability of each of the plurality of position points distributing a point cloud at the first moment; Based on the probability of each position point distributing the point cloud at the first moment, a sample probability density distribution function of the point cloud set is constructed.

15. The device according to claim 14, characterized in that The sample probability density distribution function is as follows: Where p(x, y) is the sample probability density distribution function, δ(·) is the Dirac generalized function, P{(x i ,y i )} represents the probability of the i-th position point among the multiple position points distributing the point cloud at the first moment; (x i ,y i ) represents the position coordinates corresponding to the i-th position point among the multiple position points at the first moment; the value of i is a positive integer from 1 to M, and M is the total number of the position points.

16. The device according to claim 13, characterized in that The second time period includes K moments; The processing module is specifically used for: Determine the first cumulative probability density distribution function at time k-1; wherein the value of k is greater than 1 and less than K; Determine a second cumulative probability density distribution function at time k according to the first cumulative probability density distribution function and the cumulative probability of the distribution point clouds of the plurality of location points at time k in the second time period; When the difference between the first cumulative probability density distribution function and the second cumulative probability density distribution function is greater than or equal to a second threshold, determining a second cumulative probability density distribution function at time k+1 according to the second cumulative probability density distribution function and the cumulative probability of the distribution point cloud of the plurality of position points at time k+1 in the second time period; or, When the difference between the first cumulative probability density distribution function and the second cumulative probability density distribution function is less than a second threshold, the first cumulative probability density distribution function is used as the cumulative probability density distribution function of the point cloud set.

17. The device according to claim 16, characterized in that The cumulative probability density distribution function of the point cloud set is as follows: in, Q″{(x i ,y i )}=Q′{(x i ,y i )}+N; Wherein, q(x, y) is the cumulative probability density distribution function, δ(·) is the Dirac generalized function, Q{·} represents the final cumulative probability of the i-th position point among the multiple position points in the distribution point cloud in the second time period, (x i ,y i ) represents the position coordinates of the i-th position point among the plurality of position points in the second time period; Q′{(x i ,y i )} is used to represent the cumulative probability of the distribution point cloud of the i-th position point among the multiple position points at the k-1 moment in the second time period, Q″{(x i ,y i )} is used to represent the cumulative probability of the distribution point cloud of the i-th position point among the multiple position points at the kth moment in the second time period; N is a constant.

18. The device according to any one of claims 13 to 17, characterized in that: The processing module is further used for: Determine the relative divergence KLD between the sample probability density distribution function and the cumulative probability density distribution function; Based on the KLD, the similarity between the sample probability density distribution function and the cumulative probability density distribution function is determined; wherein the KLD is negatively correlated with the similarity.

19. The device according to claim 18, characterized in that The KLD, the sample probability density distribution function, and the cumulative probability density distribution function satisfy the following relationship: Among them, D KL (p||q) is the KLD, p(x, y) is the sample probability density distribution function, and q(x, y) is the cumulative probability density distribution function.

20. The device according to claim 16, characterized in that The target object is a moving target; the multiple position points include a first position point and a second position point, and the second position point is a position point other than the first position point among the multiple position points; the cumulative probability of the point cloud distributed at the k moment of the second time period of the multiple position points includes a first cumulative probability and a second cumulative probability; wherein the first cumulative probability is the cumulative probability of the point cloud distributed at the k moment of the second time period of the first position point, and the second cumulative probability is the cumulative probability of the point cloud distributed at the k moment of the second time period of the second position point; The processing module is specifically used for: When the first cumulative probability is less than a second threshold, the first cumulative probability is deleted from the cumulative probability of the distribution point cloud of the multiple position points at time k in the second time period to obtain the second cumulative probability; A second cumulative probability density distribution function at time k is determined according to the second cumulative probability.

21. The device according to claim 20, characterized in that The processing module is specifically used for: The second cumulative probability is obtained by deleting the first cumulative probability from the cumulative probabilities of the point clouds distributed at the multiple locations at time k in the second time period through a probability transfer function.

22. The device according to claim 21, characterized in that The probability transfer function and the second cumulative probability density distribution function at time k satisfy the following relationship: Among them, q k (x, y) is the second cumulative probability density distribution function at time k, δ(·) is the Dirac generalized function, Q{(x i ,y i )} represents the final cumulative probability of the distribution point cloud of the i-th position point among multiple position points from time 0 to time k in the second time period, (x i ,y i ) represents the position coordinates of the i-th position point among the multiple position points in the second time period; T k|k-1 (x, y) is the probability transfer function.

23. The device according to claim 20 or 21, characterized in that The probability transfer function is related to at least one of a posture, a velocity, or an orientation of the target object.

24. The device according to any one of claims 13 to 23, characterized in that The processing module is further used for: Determine a first probability that an i-th position point among the plurality of position points distributes a point cloud at the first moment; Determine a second probability that the i-th position point distributes the point cloud within the second time period; Based on the difference between the first probability and the second probability, the point cloud distribution stability of the i-th position point in the first time period is evaluated; wherein the difference is positively correlated with the distribution stability of the point cloud of the i-th position point.

25. A point cloud evaluation device, characterized in that: comprising at least one processor and an interface; The at least one processor is used to call and run a computer program from the interface, and when the at least one processor executes the computer program, the method according to any one of claims 1 to 12 is implemented.

26. A chip system, characterized in that: The chip system includes: at least one processor and an interface; The at least one processor is used to call and run a computer program from the interface, and when the at least one processor executes the computer program, the method according to any one of claims 1 to 12 is implemented.

27. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a computer, the computer is enabled to execute the method according to any one of claims 1 to 12.

28. A terminal, characterized in that: The terminal includes a point cloud evaluation device as described in any one of claims 13-24.

Images