动态背景下基于边缘检测的道路识别-《计算机技术与发展》

文章信息/Info

Author(s):: WANG Bo-xue; WANG Xia-li; ZHAO Xiao-na; WU Qi; School of Information Engineering,Chang’an University,Xi’an 710064,China

Keywords:: road recognition; edge detection; dynamic background; Canny operator

摘要:: 在实际应用中,对道路识别的实时性以及准确性都有很高的要求,而且在很多情况下道路视频图像中的背景都不是静止的,因此,在场景复杂多变的视频图像中完成对道路的识别,是计算机视觉的研究热点,也是研究难点.针对上述问题,对复杂环境下道路识别与道路特征提取进行了深入研究,提出了一种有效的边缘检测方法.该方法通过对传统边缘检测的Canny算法进行改进,解决了识别过程中由于背景变化造成的误差以及检测结果不准确等问题,实现了对视频图像中道路位置的实时更新,并且分别与灰度直方图法和延时摄影法进行了实验对比.实验结果表明,该方法的限定条件相对较少且准确率较高,具有较好的识别效果.

Abstract:: In practical applications,the real-time and accuracy of road recognition have a high demand,and in many cases,the background in video images are not static. Therefore,completion of road recognition in the complex video image is a hot spot and also a research difficulty in computer vision. Aiming at the problem,we conduct an in-depth study on road identification and road feature extraction in complex environments and propose an effective edge detection method. The method is improved based on the Canny algorithm of traditional edge method,which solves the problems of error caused by background changes in the identification and inaccurate detection result,achieving the real-time updating of road location in the video image. Compared with gray histogram and time-lapse photography respectively,it shows that this method has relatively few restrictions and high accuracy with better recognition effect.