Method, system, electronic device and storage medium for optimizing sparse point cloud

Abstract

The application provides a sparse point cloud optimization method, system, electronic device and storage medium. The method comprises: performing first position estimation and velocity estimation on a first point cloud cluster detected in a first period; obtaining motion estimation of the first point cloud cluster based on the velocity estimation and a time difference between the first period and a second period; obtaining second position estimation of the first point cloud cluster in the second period based on the motion estimation and the first position estimation; and superimposing the first point cloud cluster and a second point cloud cluster detected in the second period based on the second position estimation to form a dense optimized point cloud cluster. The first point cloud cluster in a historical period is predicted in velocity, so as to estimate the position of the first point cloud cluster in the current period, and then combined with the second point cloud cluster in the current period, so that the density of the point cloud cluster in the current period is increased, and the problem of sparse point cloud cluster obtained by the sensor is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of optical detection, and in particular to a sparse point cloud optimization method and system, electronic equipment and storage medium. BACKGROUND

[0002] Millimeter wave radar uses radiated electromagnetic energy to measure targets in the sensor field of view, and has strong long-distance target detection capability. Compared with other automotive sensors, automotive radar can provide unique speed (Doppler) measurements of targets, still has good robustness in bad weather and strong light environments, and has relatively low cost, so it is called the most reliable sensor in automotive perception technology. With the development of radar technology and chip technology, millimeter wave radar is getting smaller and smaller, and can play a good performance in a multi-target complex environment, and better warn the driver, which makes people pay more and more attention to millimeter wave radar.

[0003] With the continuous expansion of millimeter wave radar from ADAS (Advanced Driving Assistance System) to L2, L3 and other high-level automatic driving applications, millimeter wave radar perception also requires higher detection accuracy, higher detection rate and lower false detection probability, and also provides more accurate information of the target. However, due to the limitation of the angle resolution of millimeter wave, the point cloud cluster distribution is relatively sparse, and the detection point cloud cluster distribution of the same target in different frames is relatively random, lacking clear contour features, and due to the influence of transverse detection accuracy, the target position accuracy is not high, which brings great challenges to the perception of millimeter wave radar.

[0004] In order to meet the accuracy requirements of high-level automatic driving for target attributes, high-precision detection under sparse millimeter wave radar point cloud cluster needs to be realized. In the prior art, target detection based on single-frame or single-cycle point cloud cluster is very easy to cause target missed detection and false detection due to the sparsity of point cloud cluster.

[0005] Therefore, a more reliable method is needed to solve the obstacle detection problem under sparse point cloud cluster. SUMMARY

[0006] In view of the problems in the prior art, the present application provides a sparse point cloud optimization method, system, electronic equipment and storage medium.

[0007] The sparse point cloud optimization method provided by the present application comprises:

[0008] performing first position estimation and velocity estimation on the first point cloud cluster detected in the first cycle;

[0009] obtaining a motion estimation of the first point cloud cluster based on the speed estimation and the time difference from the first period to the second period;

[0010] obtaining a second position estimation of the first point cloud cluster in the second period based on the motion estimation and the first position estimation;

[0011] superimposing the first point cloud cluster and a second point cloud cluster detected in the second period based on the second position estimation, to form a densely optimized point cloud cluster.

[0012] According to the optimization method of the sparse point cloud provided by the application, the method for obtaining the first point cloud cluster comprises:

[0013] generating a preliminary position estimation of the first point cloud cluster in the second period based on the motion estimation and the first position estimation;

[0014] obtaining a second position estimation of the first point cloud cluster in the second period based on the motion estimation and the first position estimation.

[0015] According to the optimization method of the sparse point cloud provided by the application, the self-position information of the detection subject in the time difference comprises translation and rotation caused by the change of the speed of the detection subject.

[0016] According to the optimization method of the sparse point cloud provided by the application, the method further comprises:

[0017] performing target detection based on the densely optimized point cloud cluster.

[0018] According to the optimization method of the sparse point cloud provided by the application, the method for obtaining the first point cloud cluster comprises:

[0019] detecting surrounding obstacles and targets by a millimeter radar wave in the first period, to obtain a plurality of frames of first point cloud information;

[0020] converting the plurality of frames of first point cloud information into a plurality of frames of second point cloud information in a vehicle body coordinate system;

[0021] forming a plurality of first point cloud clusters by clustering based on the plurality of frames of second point cloud information.

[0022] According to the optimization method of the sparse point cloud provided by the application, the period comprises a perception processing period of automatic driving.

[0023] The application further provides an optimization system of a sparse point cloud, and the system comprises:

[0024] a first estimation module configured to perform a first position estimation and a velocity estimation on a first point cloud cluster detected in a first period;

[0025] a second estimation module configured to obtain a motion estimation of the first point cloud cluster based on the velocity estimation and a time difference between the first period and a second period;

[0026] a third estimation module configured to obtain a second position estimation of the first point cloud cluster in the second period based on the motion estimation and the first position estimation;

[0027] an optimization module configured to superimpose the first point cloud cluster and a second point cloud cluster detected in the second period based on the second position estimation to form a densely optimized point cloud cluster.

[0028] The application also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the optimization method for sparse point cloud according to any one of the above.

[0029] The application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program is executable by a processor to implement the steps of the optimization method for sparse point cloud according to any one of the above.

[0030] The application also provides a computer program product comprising a computer program, wherein the computer program is executable by a processor to implement the steps of the optimization method for sparse point cloud according to any one of the above.

[0031] The optimization method for sparse point cloud, system, electronic device, and storage medium provided by the application estimate the position of the first point cloud cluster in the current period by predicting the velocity of the first point cloud cluster in the historical period, and then combine the first point cloud cluster with the second point cloud cluster in the current period, so as to increase the density of the point cloud cluster in the current period and solve the problem of sparse point cloud cluster obtained by the sensor. BRIEF DESCRIPTION OF DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the application or prior art, the following will briefly introduce the drawings needed in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and those skilled in the art can obtain other drawings according to these drawings without creative effort.

[0033] Figure 1 A flowchart of the optimization method for sparse point cloud provided by the application;

[0034] Figure 2A schematic diagram of a single-frame detection result of a forward millimeter wave radar in the prior art is shown in FIG. 1.

[0035] Figure 3 A schematic diagram of a multi-frame motion compensation detection result of a forward millimeter wave radar provided by an embodiment of the present application is shown in FIG. 2.

[0036] Figure 4 A schematic diagram of an optimization system of a sparse point cloud provided by the present application is shown in FIG. 3.

[0037] Figure 5 A schematic diagram of a physical structure of an electronic device provided by the present application is shown in FIG. 4. DETAILED DESCRIPTION

[0038] To make the objectives, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some of the embodiments of the present application, but not all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts fall within the scope of the present application.

[0039] The optimization method of a sparse point cloud provided by the present application will be described in detail below with reference to the accompanying drawings, specific embodiments and application scenarios.

[0040] Figure 1 A schematic diagram of a flow of the optimization method of a sparse point cloud provided by the present application is shown in FIG. 5. Figure 1 The optimization method of a sparse point cloud provided by the present application includes the following steps.

[0041] S100, performing first position estimation and speed estimation on a first point cloud cluster detected in a first period.

[0042] It should be noted that the detection subject is provided with a sensor, which can acquire a target in the periphery of the detection subject, and the target is in the form of a point cloud cluster.

[0043] Optionally, the first position refers to the position of the detected target relative to the detection subject,

[0044] Preferably, when the target and the detection subject are in the same plane, a polar coordinate system is used to identify the distance and azimuth angle of the target and the detection subject.

[0045] Optionally, the speed estimation refers to the speed of the target relative to the detection subject, and specifically, the method of speed estimation includes:

[0046] Obtaining the radial velocity distance and azimuth angle of a plurality of point clouds in the first point cloud cluster relative to the detection subject;

[0047] Based on the radial velocity, distance and azimuth angle of the plurality of point clouds relative to the detection subject, the velocity of the target relative to the detection subject is estimated by performing least square on the point clouds.

[0048] Optionally, according to the obtained first point cloud cluster, the position, size and velocity information of the target represented by the first point cloud cluster can be obtained. The specific method is as follows: assuming that there are K point clouds in the first point cloud cluster, each point cloud has distance r, azimuth angle θ and radial velocity v r Since the attributes of the target represented by the K point clouds are unique, least square is performed on the K point clouds to estimate the real velocity information of the target, and the reference formula is as follows:

[0049]

[0050] At the same time, according to the position and distribution of the center point of the first point cloud cluster, the position and size information of the target are obtained.

[0051] Further, the calculated v x , v y (the velocity of the target in the x and y directions) is assigned to each point cloud in the cluster.

[0052] S200, based on the velocity estimation and the time difference between the first period and the second period, the motion estimation of the first point cloud cluster is obtained.

[0053] It should be noted that the second period and the first period do not necessarily follow the chronological order, and are not necessarily two adjacent periods.

[0054] Optionally, the second period is the next period immediately after the first period, and the time difference is equal to the span of one period.

[0055] Optionally, the motion estimation refers to the relative displacement of the detected target relative to the detection subject under the previous velocity estimation, on the premise that the detection subject maintains the original motion state unchanged.

[0056] S300, based on the motion estimation and the first position estimation, a second position estimation of the first point cloud cluster in the second period is obtained.

[0057] S400, based on the second position estimation, the first point cloud cluster and the second point cloud cluster detected in the second period are superimposed to form a dense optimized point cloud cluster.

[0058] It should be noted that the second point cloud cluster and the first point cloud cluster come from the detection of the same target.

[0059] The embodiment estimates the position of the first point cloud cluster in the current period by predicting the speed of the first point cloud cluster in the historical period, and combines the first point cloud cluster with the second point cloud cluster in the current period, so that the density of the point cloud cluster in the current period is increased, and the problem of sparse point cloud cluster obtained by the sensor is solved.

[0060] Further, on the basis of the foregoing embodiment, in another embodiment, the embodiment provides a sparse point cloud optimization method, based on motion estimation and first position estimation, obtaining a second position estimation of the first point cloud cluster in a second period, comprising:

[0061] Based on the motion estimation and the first position estimation, a preliminary position estimation of the first point cloud cluster in the second period is generated.

[0062] Based on the detection subject's own pose information in the time difference and the preliminary position estimation, a second position estimation of the first point cloud cluster in the second period is obtained.

[0063] It should be noted that the foregoing embodiment only considers that the target has a first relative displacement with the detection subject under the predicted speed, and the embodiment considers that the speed (vector) of the detection subject changes in the process of the time difference, resulting in a second relative displacement of the actual position of the detection subject in the second period compared with the (assumed) virtual position of the detection subject in the second period when the speed of the detection subject is constant, and the first relative displacement and the second relative displacement are combined to determine the displacement of the first point cloud cluster relative to the detection subject in the second period, so as to further estimate the second position based on the original first position.

[0064] Optionally, the first point cloud cluster is first motion compensated, including: according to the time difference Δt of the point cloud arrival time and the current time, combining the speed vx, vy of the point cloud calculated and assigned in the foregoing, the preliminary position estimation x', y' of the point cloud of the compensation target motion is calculated, and the reference formula is as follows:

[0065] x' = x + v x · Δt

[0066] y' = y + v y · Δt

[0067] Then, combined with the motion information of the ego vehicle, the deviation caused by the motion of the ego vehicle is compensated to obtain the second position estimation x'', y'' of the target under the current motion state of the ego vehicle, and the reference formula is as follows:

[0068]

[0069] Wherein, T represents the relative motion of the ego vehicle.

[0070] The embodiment considers detecting the motion change of the subject, so that the second position predicted by the first point cloud cluster is more accurate.

[0071] Further, on the basis of the foregoing embodiment, in another embodiment, the embodiment provides a sparse point cloud optimization method, the self pose information includes translation and rotation caused by the speed change of the detection subject.

[0072] It should be noted that the translation refers to the aforementioned second relative displacement, and the rotation is the rotation of the detection subject around the self-axis, such as counterclockwise rotation of the detection subject, which causes the azimuth angle of the first point cloud cluster to rotate clockwise around the detection subject compared with the detection subject.

[0073] The embodiment refines the change of the relative distance and azimuth of the first point cloud cluster caused by the motion change of the detection subject, so that the second position predicted by the first point cloud cluster is more accurate.

[0074] Further, on the basis of the foregoing embodiment, in another embodiment, the embodiment provides a sparse point cloud optimization method, the method further comprises:

[0075] Based on the dense optimized point cloud cluster, the target detection is performed.

[0076] Optionally, in each period around the detection subject, a plurality of point cloud clusters are acquired by detection, and the number of targets of the detection subject is determined based on the dense point cloud cluster by a clustering method, and the contour shape and size of the target are further determined.

[0077] The embodiment realizes fast target detection by the clustering method using the point cloud cluster. Due to the aforementioned migration method of the first point cloud cluster combining target prediction motion and detection subject motion, the first point cloud cluster can be moved to the second period to accurately coincide with the second point cloud cluster. If the above scheme is not used, the first point cloud cluster may be moved to the second period, which is not in the actual target frame corresponding thereto, resulting in missed detection and false detection.

[0078] Further, on the basis of the foregoing embodiment, in another embodiment, the embodiment provides a sparse point cloud optimization method, the method for acquiring the first point cloud cluster comprises:

[0079] In the first period, a plurality of first point cloud information is acquired by detecting the surrounding obstacles by the millimeter radar wave.

[0080] The plurality of first point cloud information is converted into a plurality of second point cloud information in the vehicle body coordinate system.

[0081] Based on the plurality of second point cloud information, a plurality of first point cloud clusters are formed by clustering.

[0082] Optionally, the cycle refers to the perception processing cycle of autonomous driving. Preferably, the perception processing cycle of autonomous driving is 10Hz. At this cycle, better target detection results can be achieved.

[0083] Optionally, the target includes other vehicles in the vicinity of the vehicle in autonomous driving.

[0084] Optionally, multiple millimeter-wave radars around the autonomous vehicle obtain M frames of radar point clouds within one perception processing cycle. The coordinate system of each frame of point cloud is transformed using millimeter-wave radar calibration information to obtain M frames of point clouds in the vehicle coordinate system.

[0085] Optionally, the M-frame point clouds are stitched together to obtain a 360° point cloud. DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering is then performed on these point clouds. DBSCAN clustering, also known as density-based clustering, generally works as follows: for each point in the search space, a suitable distance metric is used to determine the number of neighboring targets. If the number of targets within a point's neighborhood exceeds a certain threshold, a cluster is formed, and the corresponding target point is considered the core point. If a core point is located in the cluster of another core point, the two points are considered directly density-reachable. Multiple points connected by direct density reachability are considered density-reachable. Starting from a core point, the cluster grows continuously towards the density-reachable region, eventually obtaining a maximum region including the core point and its neighboring targets, forming the final cluster. Target points that do not meet the core point definition and are not included in a cluster are considered noise points. Through clustering, the point cloud cluster of the target in the current sensing processing cycle can be obtained, which is the set of detection points of the millimeter-wave radar on each object.

[0086] This embodiment discloses specific application scenarios of point cloud densification and how to acquire and process point cloud clusters. When used for autonomous driving, it can achieve more accurate target detection, thereby serving the autonomous driving system.

[0087] The technical effects of the method in this embodiment are compared with those of the prior art as follows:

[0088] Figure 2 This is a schematic diagram of a single-frame detection result from a forward-facing millimeter-wave radar in the prior art, such as... Figure 2 As shown, the black box represents the target detected by the vehicle-launched millimeter-wave radar, the dashed box represents the target's actual location and size, the dots represent the millimeter-wave radar detection points, and the rectangles based on the dots represent the detection results obtained from a single frame of point cloud. It can be seen that due to the sparseness of the point cloud, the single-frame detection results result in inaccurate target center position and target size. Furthermore, the target on the far right, having only one point, cannot be clustered, leading to missed detections.

[0089] Figure 3 A forward millimeter wave radar multi-frame motion compensation detection result schematic diagram provided by the embodiment of the present application is shown in the figure. Figure 3 Compared with the prior art, Figure 2 the position and length-width precision of the detection result of the embodiment are greatly improved, and at the same time, due to multi-frame accumulation, the rightmost target point cloud density is improved and can be detected by the radar. The missed detection caused by sparse point cloud is reduced.

[0090] The embodiment of the present application innovatively applies the target speed regression method to the historical perception processing period point cloud, obtains the real speed direction of the point cloud cluster, and thus predicts the motion of the point cloud in the past period of time according to the time difference of the point cloud, accurately compensates the target motion. Then, using the own pose information, the points are subjected to motion compensation of self-vehicle translation and rotation, and through the above two steps, the state of the point cloud in the past period of time at the current time and pose is obtained, so as to be accumulated together with the point cloud obtained at the current time, and jointly participate in the target detection at the current time.

[0091] The sparse point cloud optimization system provided by the present application is described below. The sparse point cloud optimization system described below can be correspondingly referred to the sparse point cloud optimization method described above.

[0092] Figure 4 A sparse point cloud optimization system provided by the present application is shown in the figure. Figure 4 The sparse point cloud optimization system provided by the present application is shown in the figure. The sparse point cloud optimization system provided by the present application comprises:

[0093] The first estimation module performs first position estimation and speed estimation on the first point cloud cluster detected in the first period;

[0094] The second estimation module obtains the motion estimation of the first point cloud cluster based on the speed estimation and the time difference from the first period to the second period;

[0095] The third estimation module obtains the second position estimation of the first point cloud cluster in the second period based on the motion estimation and the first position estimation;

[0096] The optimization module superimposes the first point cloud cluster and the second point cloud cluster detected in the second period based on the second position estimation, to form a dense optimized point cloud cluster.

[0097] The embodiment estimates the position of the first point cloud cluster in the current period by predicting the speed of the first point cloud cluster in the historical period, and then combines the first point cloud cluster with the second point cloud cluster in the current period, so that the density of the point cloud cluster in the current period is increased, and the problem of sparse point cloud cluster obtained by the sensor is solved.

[0098] Figure 5 An entity structure diagram of an electronic device provided by the present application is shown in Figure 5 As shown, the electronic device can include a processor 510, a communications interface 520, a memory 530, and a communications bus 540, wherein the processor 510, the communications interface 520, and the memory 530 complete mutual communication through the communications bus 540. The processor 510 can invoke a logical instruction in the memory 530 to execute an optimization method for a sparse point cloud, the method comprising:

[0099] performing a first position estimation and a velocity estimation on a first point cloud cluster detected in a first period;

[0100] obtaining a motion estimation of the first point cloud cluster based on the velocity estimation and a time difference from the first period to a second period;

[0101] obtaining a second position estimation of the first point cloud cluster in the second period based on the motion estimation and the first position estimation;

[0102] superimposing the first point cloud cluster and a second point cloud cluster detected in the second period based on the second position estimation to form a densely optimized point cloud cluster.

[0103] In addition, the logical instruction in the memory 530 described above can be implemented in the form of a software functional unit and sold or used as an independent product, which can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or parts of the present application that essentially contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0104] On the other hand, the present application also provides a computer program product, which includes a computer program stored on a non-transitory computer readable storage medium, and the computer program includes program instructions, when the program instructions are executed by a computer, the computer can execute the optimization method for a sparse point cloud provided by the above-mentioned methods, the method comprising:

[0105] performing a first position estimation and a velocity estimation on a first point cloud cluster detected in a first period;

[0106] performing a motion estimation of the first point cloud cluster based on the velocity estimation and a time difference between the first period and a second period;

[0107] performing a second position estimation of the first point cloud cluster in the second period based on the motion estimation and the first position estimation;

[0108] superimposing the first point cloud cluster and a second point cloud cluster detected in the second period based on the second position estimation to form a densely optimized point cloud cluster.

[0109] In yet another aspect, the present application also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the above-provided method for optimizing a sparse point cloud, the method comprising:

[0110] performing a first position estimation and a velocity estimation on a first point cloud cluster detected in a first period;

[0111] performing a motion estimation of the first point cloud cluster based on the velocity estimation and a time difference between the first period and a second period;

[0112] performing a second position estimation of the first point cloud cluster in the second period based on the motion estimation and the first position estimation;

[0113] superimposing the first point cloud cluster and a second point cloud cluster detected in the second period based on the second position estimation to form a densely optimized point cloud cluster.

[0114] The apparatus embodiments described above are merely illustrative, wherein the units shown as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the present embodiment scheme according to actual needs. Those skilled in the art can understand and implement it without creative labor.

[0115] Those skilled in the art can clearly understand the technical solutions of the various embodiments from the above description of the embodiments, and the various embodiments can be implemented by means of software with the necessary general hardware platforms, and of course, can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that makes a contribution, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, and the like, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0116] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features therein; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method of optimizing a sparse point cloud, the method comprising: The method comprises: performing first position estimation and velocity estimation on a first point cloud cluster detected by a millimeter wave radar in a first period; the velocity estimation comprises: performing least squares on a plurality of point clouds in the first point cloud cluster based on the radial velocity, distance and azimuth angle of the plurality of point clouds relative to a detection subject, to obtain a velocity of a target represented by the first point cloud cluster relative to the detection subject, and assigning the velocity to the plurality of point clouds; based on the velocity estimation and a time difference between the first period and a second period, obtaining a motion estimation of the first point cloud cluster; based on the motion estimation and the first position estimation, obtaining a second position estimation of the first point cloud cluster in the second period, comprising: based on the motion estimation and the first position estimation, generating a preliminary position estimation of the first point cloud cluster in the second period; based on the preliminary position estimation and self-pose information of the detection subject within the time difference, obtaining the second position estimation of the first point cloud cluster in the second period; the self-pose information includes translation and rotation caused by changes in the speed of the detection subject itself; based on the second position estimation, superimposing the first point cloud cluster and a second point cloud cluster detected by the millimeter wave radar in the second period to form a densely optimized point cloud cluster.

2. The method of claim 1, wherein, The method further comprises: based on the densely optimized point cloud cluster, performing target detection.

3. The method of optimizing a sparse point cloud according to any one of claims 1-2, wherein, The method for obtaining the first point cloud cluster comprises: in the first period, detecting surrounding obstacles by a millimeter wave radar wave to obtain a plurality of frames of first point cloud information; convert the plurality of frames of first point cloud information into a plurality of frames of second point cloud information in a vehicle body coordinate system; based on the plurality of frames of second point cloud information, form a plurality of first point cloud clusters through clustering.

4. The method of claim 3, wherein, The period includes a perception processing period of autonomous driving.

5. An optimization system for a sparse point cloud, the system comprising: The system comprises: a first estimation module, which performs first position estimation and velocity estimation on a first point cloud cluster detected by a millimeter wave radar in a first period; the velocity estimation comprises: performing least squares on a plurality of point clouds in the first point cloud cluster based on the radial velocity, distance and azimuth angle of the plurality of point clouds relative to a detection subject, to obtain a velocity of a target represented by the first point cloud cluster relative to the detection subject, and assigning the velocity to the plurality of point clouds; a second estimation module, which obtains a motion estimation of the first point cloud cluster based on the velocity estimation and a time difference between the first period and a second period; a third estimation module, which obtains a second position estimation of the first point cloud cluster in the second period based on the motion estimation and the first position estimation, comprising: based on the motion estimation and the first position estimation, generating a preliminary position estimation of the first point cloud cluster in the second period; based on the preliminary position estimation and self-pose information of the detection subject within the time difference, obtaining the second position estimation of the first point cloud cluster in the second period; the self-pose information includes translation and rotation caused by changes in the speed of the detection subject itself; an optimization module to superimpose the first point cloud cluster and a second point cloud cluster detected by the millimeter wave radar within the second period based on the second position estimate to form a densely optimized point cloud cluster.

6. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor, when executing the program, implements the steps of the optimization method of a sparse point cloud as claimed in any one of claims 1-4.

7. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the optimization method of a sparse point cloud as claimed in any one of claims 1-4.

8. A computer program product comprising a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the optimization method of a sparse point cloud as claimed in any one of claims 1-4. The computer program, when executed by the processor, implements the steps of the optimization method of a sparse point cloud as claimed in any one of claims 1-4.

Images