一种通用的全景视觉水平边线快速检测方法

摘要

The sideline, which is the important mark for robot navigation and location, will be changed to the curve when it is shown in the omni-vision image. So it is hardly to detect the sideline in the image. This paper presents an universal process of the sideline detection for the omni-vision, it detects the sideline in horizontal scene. To reduce the processing time of the hough transform, only use the part points, which are extracted equably, to replace the total point set. For the different capability omni-vision, presents two candidacy point extraction way. The experiment shows this process can rightly detect the sidelines, and has high real-time.%作为机器人导航定位重要标记的边线,在普通全景视觉成像中,往往发生扭曲,难以检测与识别.文章提出了一种通用的全景视觉水平边线快速检测方法,将边线检测的坐标系从图像转换到水平场景中,以较为均匀提取的部分点集代替全部点集,以减少直线Hough变换的运算时间；针对不同性能的全景视觉,提出了两种点集的提取方式.实验表明本方法能够正确检测到全景图像中的边线,实时性好.