An OpenCL Parallelized Traffic Sign Recognition

De-kai KANG, Xing CAI, Xu-sen GUO, Jie-xin ZHENG, Xiao-mei ZHOU

Abstract


Traffic sign detection and recognition (TSDR) plays crucial roles in advanced driving assistant system(ADAS). In this work, we propose an OpenCL parallelized TSDR method to address the time-consuming challenge. The method employs AdaBoost algorithm for traffic sign detection and Fisherface algorithm for the traffic sign recognition. The Haar feature extraction and Adaboost algorithm are accelerated by sliding windows paralleling and stage classifier group scheduling strategies. The results of experimental work reveal that our approach offers about 12x speed-up for 1920x1080 resolution, which effectively compress the computation time.

Keywords


TSDR, AdaBoost, OpenCL, ADAS


DOI
10.12783/dtcse/mmsta2017/19710

Full Text:

PDF

Refbacks

  • There are currently no refbacks.