Journal of Applied Sciences ›› 2014, Vol. 32 ›› Issue (1): 93-98.doi: 10.3969/j.issn.0255-8297.2014.01.015

• Signal and Information Processing • Previous Articles     Next Articles

Vehicle Identification and Counting Based on Sparse Over-Completeness

LU Peng1, LOU Ya-fei1, LIU Feng-zhe2, LI Yu-song1, HUANG Shi-lei1, TANG Yu-he1   

  1. 1. School of Electrical Engineering, Zhengzhou University, Zhengzhou 450001, China
    2. China Power Investment Henan Technical Information Center, Zhengzhou 450016, China
  • Received:2013-06-03 Revised:2013-07-16 Online:2014-01-31 Published:2013-07-16

Abstract: To detect and count high-speed vehicles accurately and robustly, a sparse over-complete model is established to simulate the visual mechanism. The model uses image units as processing objects. It expresses the internal structure and essential attributes of the target image with a small amount of non-zero response sparse coefficients, and identifies different types of vehicles in different environments. Online counting of vehicles is achieved by setting dynamic lanes based on virtual detection lines. Accuracy of the vehicle counting reaches 98.89% under a variety of environmental conditions. The results show that the proposed algorithm can effectively suppress interference caused by external factors such as wind, light changes and camera shakes. It
has higher robustness and accuracy than traditional algorithms.  

Key words: sparse over-completeness, image unit, virtual detection line, vehicle identification, vehicle counting

CLC Number: