一种多信息级联聚类输电线检测方法

A cascade clustering transmission line detection method based on multi-information

针对树障清理空中机器人作业时需规避输电线这一需求,本文提出一种基于级联聚类的输电线检测方法。使用基于海森(Hessian)矩阵的边缘增强算法进行图像预处理;采用霍夫(Hough)变换法提取候选直线;设计一种融合直线空间位置分布特征及长度特征的级联聚类法,有效地删减由复杂背景造成的与输电线特征不相匹配的干扰线,并最终得到对应于输电线边缘的所有直线。实验结果表明,该方法具有较高的准确率及较强的鲁棒性,为树障清理空中机器人的视觉避障研究提供了基础。

To avoid aerial tree-pruning robot touching the power line during operation,this paper proposes a method for power lines detection based on multi-information cascade clustering.Firstly,a method of image edge enhancement algorithm based on Hessian matrix is used to preprocess the image;then,Hough transform is used to extract the candidate straight line;then,a cascade clustering method is designed,which integrates the length and spatial distribution characteristics of lines.This design can effectively abridge the interference lines caused by the background trees,which is discordant with features of power lines.And finally,all lines corresponding to the edges of the power lines can be obtained.Experimental results show that the proposed method can achieve relatively high accuracy and comparatively strong robustness,providing a pre-research basis for exploration of the resolution to visual obstacle avoidance of aerial tree-pruning robot.