Paper Info
Title
Dynamic object culling and map construction algorithm based on real-time point cloud data
Abstract
With the development of the field of unmanned driving, surveying and mapping has become a core issue in the field of mobile robots. The constructed environment map can be used to perform specific tasks, such as road cleaning, on-site rescue, and material transportation. At present, most mapping algorithms work well in static environments, but the existence of dynamic objects will seriously affect the mapping accuracy of the system, resulting in large deviations and affecting the use of subsequent maps. This paper studies dynamic crowded scenes and proposes a real-time mapping system for general dynamic object recognition and culling based on traditional point cloud processing to obtain reliable static environment maps. Firstly, the point cloud data obtained by solid-state lidar is preprocessed and clustered, and multi-target matching and tracking are carried out according to the feature information obtained by the clustering. Then, single points are voted by using postframe matching to distinguish static and dynamic objects. Point cloud culling, construction and update of static object point cloud map based on octree structure. Experiments on actually collecting road scenes in mines show that the algorithm can successfully eliminate dynamic objects in point cloud data and update static environment maps efficiently.
Year
DOI
Venue
2022
10.1109/CACRE54574.2022.9834159
2022 7th International Conference on Automation, Control and Robotics Engineering (CACRE)
Keywords
DocType
ISBN
Dynamic objects,real-time mapping,multi-target matching,map construction,efficient update
Conference
978-1-6654-6669-1
Citations 
PageRank 
References 
0
0.34
4
Authors
4
Name
Order
Citations
PageRank
Chao Yang100.34
Liangcai Ren200.68
Xin Wu300.34
Xinyi Zhou400.34