摘要
针对超视距情况下,铁路行车路障监控的实时性问题,提出了一种快速超视距障碍物识别方法.先利用纹理特征进行快速预处理与边缘检测;再利用铁轨的纵向延展特点,用搜索连通域投影最长的方法定位铁轨区域并建立检测窗;并行的用帧差法比较当前帧与检测窗识别出障碍物,最后利用IPM模型计算障碍物大小与距离.实验结果表明,在确保90.3%识别率的前提下,本算法识别障碍物平均耗时43ms.
To deal with real-time obstacle detecting over-the-horizon on rail line, this paper devised a fast over-the-horizon obstacle recognition algorithmt which consists of three parts. Edge detection is used to preproeess according to texture features. Tracks are located by finding the longest connected domain projec- tion based on the longitudinal extension characteristics of two tracks. Finally, obstacle position is located by difference of foreground and background pixels, following by computing its scale and distance. Experiments showed that this algorithm costs average 43 ms per recognition of one frame with a recognition rate at 90. 3%.
出处
《湘潭大学自然科学学报》
CAS
北大核心
2013年第2期103-108,共6页
Natural Science Journal of Xiangtan University
基金
国家"863"计划基金项目(2010AA09Z104)
