Radar data pre-labeling method based on clustering

Abstract

The invention discloses a radar data pre-labeling method based on clustering. The method comprises the following steps: reading radar data; performing clustering processing on the radar data; target labels are edited based on clustering results, and the clustered normal targets are labeled; and taking a clustering result and a labeling result as a real labeling result. According to the radar data pre-labeling method based on clustering, radar data labeling is carried out based on clustering, the labeling precision of 3D radar data can be improved to be complete and just surrounds a target area, laser radar labeling can be rapidly achieved, and the workload is reduced by 90% or above; the problems of low radar data labeling precision, non-uniform rules and low radar data labeling speed can be solved; automatic labeling and model algorithm training can be realized, and pre-labeling, manual correction and algorithm training can be automatically realized; and multi-mode automatic labeling can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving, and more particularly, to a method for pre-labeling radar data based on clustering. Background Art

[0002] In autonomous driving and robotic systems, deep learning algorithms based on radar data are usually used for perception. To achieve this goal, a large number of relatively accurate radar data annotation samples are required for algorithm training. However, since radar data is data distributed in three-dimensional space, it is difficult to perform data annotation, especially for millimeter-wave radar data. At the same time, due to the large differences in spatial perception ability among different people, it is difficult to form a relatively unified annotation method, which brings great difficulties in model training.

[0003] Currently, one annotation method is to manually analyze radar data and draw target boxes. Manual drawing of radar target boxes has the following problems: there are large differences in spatial perception ability among different people, and it is difficult to form a unified annotation specification; the manual annotation workload is large and the operation is inconvenient. Another annotation method is to perform data pre-labeling through inference based on existing algorithm models. Performing data inference pre-labeling through existing models has the following problems: the current radar models generally have a low recall rate and it is difficult to perform effective annotation; radar models perform data inference pre-labeling, which has high requirements for model performance, usually has large deviations, and is difficult to correct.

[0004] Therefore, there is an urgent need for a method for pre-labeling radar data based on clustering. Summary of the Invention

[0005] The object of the present invention is to provide a method for pre-labeling radar data based on clustering to solve the above problems in the prior art, which can quickly achieve lidar annotation and reduce the workload by more than 90%.

[0006] The present invention provides a method for pre-labeling radar data based on clustering, which includes:

[0007] Read radar data;

[0008] Perform clustering processing on the radar data;

[0009] Edit target labels based on the clustering results and perform label processing on the normal targets after clustering;

[0010] Use the clustering results and the results of label processing as the true annotation results.

[0011] In the method for pre-labeling radar data based on clustering as described above, preferably, the radar data includes data collected by lidar and / or millimeter-wave radar.

[0012] The above-mentioned clustering-based radar data pre-annotation method, wherein, preferably, the reading of radar data specifically includes:

[0013] Load the radar data frame and only retain the X, Y, and Z dimension information of the radar data.

[0014] The above-mentioned clustering-based radar data pre-annotation method, wherein, preferably, the clustering process of the radar data specifically includes:

[0015] Perform clustering on the read radar data and draw the bounding box of the target area after clustering.

[0016] The above-mentioned clustering-based radar data pre-annotation method, wherein, preferably, the editing of the target label based on the clustering result and the tagging process of the normal target after clustering specifically include:

[0017] Tag the clustering targets that meet the expectations after clustering, and eliminate the clustering targets that do not meet the expectations.

[0018] The above-mentioned clustering-based radar data pre-annotation method, wherein, preferably, the clustering targets that meet the expectations include vehicles and pedestrians.

[0019] The above-mentioned clustering-based radar data pre-annotation method, wherein, preferably, after the clustering process and before the editing of the target label, the clustering-based radar data pre-annotation method further includes:

[0020] Eliminate abnormal targets.

[0021] The above-mentioned clustering-based radar data pre-annotation method, wherein, preferably, the elimination of abnormal targets specifically includes:

[0022] Eliminate the targets that are not of concern after clustering and the results of abnormal clustering.

[0023] The present invention provides a clustering-based radar data pre-annotation method. By performing radar data annotation based on clustering, the annotation accuracy of 3D radar data can be improved to completely and precisely surround the target area, enabling rapid lidar annotation and reducing the workload by more than 90%; it can solve the problems of low radar data annotation accuracy, inconsistent rules, and slow radar data annotation speed; it can achieve automated annotation and model algorithm training, and can automatically realize pre-annotation, manual correction, and algorithm training; it can achieve multi-modal automatic annotation. Description of the Drawings

[0024] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described below in conjunction with the drawings, wherein:

[0025] Figure 1 This is a flowchart of an embodiment of the radar data pre-annotation method based on clustering provided by the present invention. Detailed implementation manners

[0026] Now, various exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. The description of the exemplary embodiments is merely illustrative and in no way limits the present disclosure or its application or use. The present disclosure can be implemented in many different forms and is not limited to the embodiments described herein. These embodiments are provided to make the present disclosure thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that: unless otherwise specifically stated, the relative arrangements of components and steps, the compositions of materials, numerical expressions, and numerical values set forth in these embodiments should be construed as merely exemplary and not as limitations.

[0027] The terms "first", "second", and the like used in the present disclosure do not denote any order, quantity, or importance, but are merely used to distinguish different parts. Words such as "including" or "comprising" mean that the elements before such words cover the elements listed after such words, and do not exclude the possibility of also covering other elements. Terms such as "upper" and "lower" are only used to represent relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0028] In the present disclosure, when it is described that a specific component is located between a first component and a second component, there may or may not be an intermediate component between the specific component and the first component or the second component. When it is described that a specific component is connected to other components, the specific component may be directly connected to the other components without an intermediate component, or may not be directly connected to the other components but have an intermediate component.

[0029] All terms used in the present disclosure (including technical terms or scientific terms) have the same meaning as understood by those of ordinary skill in the art to which the present disclosure pertains, unless otherwise specifically defined. It should also be understood that terms defined in a general dictionary, such as those, should be interpreted as having a meaning consistent with their meaning in the context of the relevant art, and should not be interpreted in an idealized or overly formal sense, unless specifically defined as such here.

[0030] Technologies, methods, and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and devices should be regarded as part of the specification.

[0031] As Figure 1 shown, in the actual execution process of the radar data pre-annotation method based on clustering provided in this embodiment, it specifically includes the following steps:

[0032] Step S1, read radar data.

[0033] Specifically, load the radar data frame and only retain the X, Y, and Z dimension information of the radar data. Among them, the radar data includes data collected by lidar and / or millimeter-wave radar. The present invention supports the data reading and parsing of lidar and millimeter-wave radar from different manufacturers.

[0034] Step S2, perform clustering processing on the radar data.

[0035] Specifically, perform clustering processing on the read radar data and draw the bounding box of the target area after clustering. In some embodiments of the present invention, the bounding box of the target area is accurately drawn according to the clustering result, so that the annotation accuracy of 3D radar data can be improved to completely and exactly surround the target area.

[0036] Step S3, edit the target label based on the clustering result and perform label processing on the normal targets after clustering.

[0037] Specifically, perform label processing on the clustering targets that meet the expectations after clustering, and eliminate the clustering targets that do not meet the expectations. Among them, the clustering targets that meet the expectations include vehicles and pedestrians.

[0038] Furthermore, in some embodiments of the present invention, after the step S2 and before the step S3, the radar data pre-annotation method based on clustering further includes:

[0039] Step S2-3, eliminate abnormal targets.

[0040] Specifically, eliminate the targets that are not concerned after clustering and the results of abnormal clustering.

[0041] Step S4, use the clustering result and the label processing result as the true annotation result.

[0042] The radar data pre-annotation method based on clustering provided by the embodiments of the present invention annotates radar data based on clustering, can improve the annotation accuracy of 3D radar data to completely and exactly surround the target area, can quickly implement lidar annotation, and reduce the workload by more than 90%; can solve the problems of low radar data annotation accuracy, inconsistent rules, and slow radar data annotation speed; can achieve automatic annotation and model algorithm training, and can automatically implement pre-annotation, manual correction, and algorithm training; can achieve multi-modal automatic annotation.

[0043] Thus far, the embodiments of the present disclosure have been described in detail. To avoid obscuring the concept of the present disclosure, some details well known in the art have not been described. Those skilled in the art can fully understand how to implement the technical solutions disclosed herein based on the above description.

[0044] Although some specific embodiments of the present disclosure have been described in detail by way of examples, those skilled in the art should understand that the above examples are only for illustration purposes and not for limiting the scope of the present disclosure. Those skilled in the art should understand that the above embodiments can be modified or partial technical features can be equivalently replaced without departing from the scope and spirit of the present disclosure. The scope of the present disclosure is defined by the appended claims.

Claims

1. A radar data pre-labeling method based on clustering, characterized in that: include: Read radar data; Performing clustering processing on the radar data; Edit target labels based on clustering results and label normal targets after clustering; The clustering results and labeling results are regarded as the actual annotation results.

2. The radar data pre-labeling method based on clustering according to claim 1, characterized in that: The radar data includes data collected by laser radar and / or millimeter wave radar.

3. The radar data pre-labeling method based on clustering according to claim 1, characterized in that: The reading of radar data specifically includes: Load the radar data frame and keep only the X, Y, and Z dimension information of the radar data.

4. The radar data pre-labeling method based on clustering according to claim 1, characterized in that: The clustering process of the radar data specifically includes: The radar data read are clustered, and a bounding box of the clustered target area is drawn.

5. The radar data pre-labeling method based on clustering according to claim 1, characterized in that: The target label is edited based on the clustering result, and the normal target after clustering is labeled, which specifically includes: After clustering, the clustering targets that meet the expectations are labeled, and the clustering targets that do not meet the expectations are eliminated.

6. The radar data pre-labeling method based on clustering according to claim 5, characterized in that: The desired clustering targets include vehicles and pedestrians.

7. The radar data pre-labeling method based on clustering according to claim 1, characterized in that: After the clustering process and before the target label is edited, the radar data pre-labeling method based on clustering further includes: Eliminate abnormal targets.

8. The radar data pre-labeling method based on clustering according to claim 7, characterized in that: The elimination of abnormal targets specifically includes: Targets that are not of interest after clustering and abnormal clustering results are eliminated.

Images