Authors:
Dongxu Lv
;
Peijun Wang
;
Wentao Li
and
Peng Chen
Affiliation:
Southwest Jiaotong University, China
Keyword(s):
OpenCL, PFH, Parallel Computing, Point Cloud Data.
Related
Ontology
Subjects/Areas/Topics:
Engineering Applications
;
Informatics in Control, Automation and Robotics
;
Intelligent Control Systems and Optimization
;
Robotics and Automation
;
Signal Processing, Sensors, Systems Modeling and Control
Abstract:
To meet the requirements of railway track point cloud processing, an OpenCL-accelerated Point Feature His-togram method is proposed using heterogeneous computing to improve the low computation efficiency. Ac-cording to the characteristics of parallel computing of OpenCL, the data structure for point cloud storage is re-configured. With the kernel performance analysis by CodeXL, the data reading is improved and the load of ALU is promoted. In the test, it obtains 1.5 to 40 times speedup ratio compared with the original functions at same precision of CPU algorithm, and achieves better real-time performance and good compatibility to PCL.