Title:
METHOD FOR PROCESSING LASER POINT CLOUD AND RELATED DEVICE
Document Type and Number:
WIPO Patent Application WO/2021/238306
Kind Code:
A1
Abstract:
A method for processing a laser point cloud and a related device, applied to the laser sensing field in the field of autonomous driving, and specifically applied to an intelligent body (such as an intelligent vehicle and an intelligent connected vehicle) for intelligent driving. The method comprises: clustering a laser point cloud (e.g., clustering the laser point cloud in an OGM by means of a DFS) to obtain N initial clustering clusters of rough classification; performing semantic segmentation on the laser point cloud by means of a neural network (e.g., PointSeg and DeepSeg) to obtain a category tag of each laser point; querying the category tag corresponding to each laser point for each initial clustering cluster, and reprocessing (e.g., re-segmenting) each initial clustering cluster according to the queried category tag, to obtain a target clustering cluster corresponding to a target object. By performing semantic segmentation on the laser point cloud, and in combination of the conventional clustering algorithm, the problems of the laser point cloud in laser sensing such as over-segmentation and under-segmentation are solved, thereby improving the detection performance of critical obstacles.
Inventors:
LI ZHIGANG (CN)
PENG FENGCHAO (CN)
LIU BINGBING (CN)
YANG ZHEN (CN)
ZHANG WEI (CN)
PENG FENGCHAO (CN)
LIU BINGBING (CN)
YANG ZHEN (CN)
ZHANG WEI (CN)
Application Number:
PCT/CN2021/076816
Publication Date:
December 02, 2021
Filing Date:
February 19, 2021
Export Citation:
Assignee:
HUAWEI TECH CO LTD (CN)
International Classes:
G06K9/00
Domestic Patent References:
| WO2019168869A1 | 2019-09-06 |
Foreign References:
| CN110264468A | 2019-09-20 | |||
| CN110110802A | 2019-08-09 | |||
| CN109840454A | 2019-06-04 | |||
| CN111126182A | 2020-05-08 | |||
| CN110136182A | 2019-08-16 |
Attorney, Agent or Firm:
SHENPAT INTELLECTUAL PROPERTY AGENCY (CN)
Download PDF: