摘要： 针对监控视频中的电线杆遮挡问题,提出一种改进的线段分割检测(LSD)算法。将RGB图像进行双边滤波,并将滤波后的图像转换到HSV空间,在HSV彩色空间内利用矢量求导方法计算彩色图像的梯度和方向,利用LSD算法实现图像直线的提取。对直线进行合并得到矩形区域,采用矩形区域在H和S分量的标准差之积和结构相似度确定候选区域,通过合并候选区域完成电线杆检测。实验结果证明,提出算法的准确率可达到89.40%,能够有效并且准确地检测出电线杆遮挡。

关键词: 电线杆检测, 线段分割检测算法, 结构相似度, 直线合并, HSV空间, 遮挡检测

Abstract: For the utility pole occlusion in monitoring videos,this paper proposes an improved Line Segmentation Detection(LSD) algorithm.After bilateral filtering,the RGB image is converted to HSV color space.The vector derivation method is used to compute gradients and directions of color images in HSV color space.And LSD algorithm is used to achieve the extraction of image lines.The rectangular regions are obtained by merging the lines.The candidate region is gained by the product of standard deviations of H weight and S weight as well as the structure similarity in rectangle regions.Utility pole detection is realized by combining candidate regions.Experimental results show that,the accuracy rate of the proposed algorithm is up to 89.40%,and it can effectively and accurately detect utility pole occlusion.

Key words: utility pole detection, Line Segmentation Detection(LSD) algorithm, structure similarity, line merging, HSV space, occlusion detection

中图分类号: 

• TP391