期刊名称：International Journal of Signal Processing, Image Processing and Pattern Recognition
印刷版ISSN：2005-4254
出版年度：2016
卷号：9
期号：8
页码：263-274
出版社：SERSC
摘要：Zebra-crossing Recognition is one of the essential parts of the visual based intelligent vehicle navigation or intelligent driving assistant system. In order to address the real-time and robustness, Zebra-crossing Recognition Method which is based on spatial-temporal correlation has been proposed. Firstly, calibrate a camera mounted on the vehicle by a practical method. Then, according to the prior knowledge such from GPS etc, a judgment whether it's in the Crossing area is made. Next, utilize the bipolar property of Zebra-crossing to extract features. Finally, the recognition results are obtained according to the model constraints. In this paper, proposed methods can improve real time identification of the zebra line by using spatial correlation, reduce the cost of recognition and lower errors during identification. The method overcomes some disadvantages of traditional identification approaches based upon video recognition, for instance higher cost and errors
