Search

CN-115830550-B - Method and device for detecting motion state of target

CN115830550BCN 115830550 BCN115830550 BCN 115830550BCN-115830550-B

Abstract

The application provides a method and equipment for detecting a motion state of a target, which comprise the steps of obtaining pose of a collection vehicle at a plurality of collection moments and point clouds of the target at the plurality of collection moments, sampling the point clouds at the plurality of collection moments to obtain a plurality of data sets, obtaining a reference pose of each data set from the pose of the plurality of collection moments according to the collection moment corresponding to the point clouds of each data set, carrying out coordinate conversion on each data set by using the reference pose of each data set, and identifying each converted data set by using a motion state detection model to obtain the motion state of the target at the sampled collection moment. The error among the point clouds of each frame can be eliminated, and the accuracy of identifying the motion state of the target is improved.

Inventors

  • Yu hongda

Assignees

  • 亿咖通(湖北)技术有限公司

Dates

Publication Date
20260508
Application Date
20221208

Claims (9)

  1. 1. A method for detecting a motion state of an object, comprising: Acquiring pose of an acquisition vehicle at a plurality of acquisition moments and point clouds of the target at a plurality of acquisition moments; sampling the point clouds at the plurality of acquisition moments by using a sampling window according to a sampling step length to obtain a plurality of data sets, wherein the sampling step length is smaller than the length of the sampling window, so that the point clouds at the sampled acquisition moments are positioned in the plurality of data sets; performing coordinate transformation on each data set by using the reference pose of each data set; and identifying each converted data set by using a motion state detection model to obtain the motion state of the target at the sampled acquisition time.
  2. 2. The detection method according to claim 1, wherein the motion state detection model is used to identify the data set after each transformation, and the motion state of the target at the sampled acquisition time is obtained, specifically including: identifying each converted data set by using the motion state detection model to obtain at least one identification result of the target at the sampled acquisition time; processing at least one identification result of the target at the sampled acquisition time, and outputting a detection motion state of the target at the sampled acquisition time.
  3. 3. The detection method according to claim 1 or 2, wherein obtaining the point clouds of the target at a plurality of the acquisition moments specifically includes: acquiring original point clouds of a plurality of acquisition moments acquired by the acquisition vehicle and contour data of the target at the plurality of acquisition moments; And processing the original point cloud at the corresponding moment by using the contour data of the target at each acquisition moment to obtain the point cloud of the target at each acquisition moment.
  4. 4. The method of detection according to claim 2, characterized in that the method comprises: clustering is carried out on each converted data set, and noise data outside a clustering area are removed; correspondingly, identifying each converted data set by using a motion state detection model to obtain the motion state of the target at the sampled acquisition time, wherein the motion state detection model specifically comprises the following steps: and identifying each noise-reduced sampling set by using the motion state detection model to obtain at least one identification result of the target at the sampled acquisition time.
  5. 5. The method according to claim 2, wherein processing at least one recognition result of the target at each acquisition time, and outputting a detected motion state of the target at each time, specifically comprises: When only one identification result is obtained at a certain acquisition time, the detection motion state of the target at the acquisition time is the identification result; When a plurality of identification results are obtained at a certain acquisition time, counting the state values of the identification results of the target at the acquisition time, and taking the state value with the largest frequency as the detection motion state of the target at the acquisition time.
  6. 6. A detection apparatus for detecting a motion state of an object, comprising: The acquisition module is used for acquiring pose of the acquisition vehicle at a plurality of acquisition moments and point clouds of the target at a plurality of acquisition moments; The processing module is used for sampling the point clouds at the plurality of acquisition moments by using a sampling window according to a sampling step length to obtain a plurality of data sets, wherein the sampling step length is smaller than the length of the sampling window, the point clouds at the acquisition moments to be sampled are located in the plurality of data sets, the pose of the earliest acquisition moment in the data sets is selected to be used as the reference pose of the data sets, the reference pose of each data set is used for carrying out coordinate conversion on each data set, and the motion state detection model is used for identifying each converted data set to obtain the motion state of the target at the acquisition moment to be sampled.
  7. 7. An electronic device comprising a processor and a memory communicatively coupled to the processor; The memory stores computer-executable instructions; The processor executes computer-executable instructions stored in the memory to implement the method of any one of claims 1 to 6.
  8. 8. A computer readable storage medium having stored therein computer executable instructions which when executed by a processor are adapted to carry out the method of any one of claims 1 to 5.
  9. 9. A computer program product comprising a computer program which, when executed by a processor, implements the method of any one of claims 1 to 5.

Description

Method and device for detecting motion state of target Technical Field The application relates to the field of automatic driving, in particular to a method and equipment for detecting the motion state of a target. Background Point cloud detection of targets is one of the key technologies in the field of autopilot awareness. Whether the automatic driving function of the vehicle end or the automatic labeling of the cloud end is adopted, the motion state (motion or stop) of the target needs to be estimated, and then the point cloud of the target can be further processed. Judging target based on point cloud the motion state requires a combination of timing. The point clouds at continuous moments are collected for judgment, but errors among the point clouds at continuous moments are larger, such as shaking of the collected vehicle or target position, misorientation of sensors on the collected vehicle, mistracking of the target and the like, which cause errors among the point clouds at each frame, and the errors easily cause misclassification of motion states. Disclosure of Invention The application provides a method and equipment for detecting the motion state of a target, which are used for solving the technical problem that the motion state identification by adopting continuous multi-frame point clouds of the target is easy to be wrongly identified due to the error between the point operations of each frame. An embodiment of the present application provides a method for detecting a motion state of a target, including: Acquiring pose of an acquisition vehicle at a plurality of acquisition moments and point clouds of a target at the plurality of acquisition moments; sampling point clouds at a plurality of acquisition moments to obtain a plurality of data sets; acquiring a reference pose of each data set from the poses of a plurality of acquisition moments; Performing coordinate transformation on each data set by using the reference pose of each data set; And identifying each converted data set by using a motion state detection model to obtain the motion state of the target at the sampled acquisition time. In an embodiment, sampling the point cloud at a plurality of acquisition moments to obtain a plurality of data sets specifically includes: Sampling the point clouds at a plurality of acquisition moments by using a sampling window according to a sampling step length to obtain a plurality of data sets, wherein the sampling step length is smaller than the length of the sampling window, so that the point clouds at the acquisition moments to be sampled are positioned in the plurality of data sets; correspondingly, each converted data set is identified by using a motion state detection model, and the motion state of the target at the sampled acquisition time is obtained, which specifically comprises: Identifying each converted data set by using a motion state detection model to obtain at least one identification result of the target at the sampled acquisition time; processing at least one identification result of the target at the sampled acquisition time, and outputting a detection motion state of the target at the sampled acquisition time. In an embodiment, acquiring the point cloud of the target at a plurality of acquisition moments specifically includes: Acquiring original point clouds of a plurality of acquisition moments acquired by an acquisition vehicle and contour data of a plurality of acquisition moment targets; And processing the original point cloud at the corresponding moment by using the contour data of the target at each acquisition moment to obtain the point cloud of the target at each acquisition moment. In one embodiment, a method includes: clustering is carried out on each converted data set, and noise data outside a clustering area are removed; correspondingly, each converted data set is identified by using a motion state detection model, and the motion state of the target at the sampled acquisition time is obtained, which specifically comprises: And identifying each noise-reduced sampling set by using a motion state detection model to obtain at least one identification result of the target at the sampled acquisition time. In an embodiment, acquiring the reference pose of each dataset from the poses of a plurality of acquisition moments specifically includes: And selecting the pose at the earliest acquisition time in the data set as the reference pose of the data set. In an embodiment, processing at least one recognition result of the target at each acquisition time, and outputting a detected motion state of the target at each time, specifically including: when only one identification result is obtained at a certain acquisition time, the detection motion state of the target at the acquisition time is the identification result; when a plurality of identification results are obtained at a certain acquisition time, counting the state values of the identification results of the target at the acquisit