一种基于几何约束的跌落式熔断器图像识别方法

Image recognition for drop-out fuse based on geometric constraint

针对配电线路维护机器人跌落式熔断器的熔丝管识别问题,提出了一种基于几何约束的直线拟合法。采用概率Hough变换,结合直线平行和距离约束等条件,提高了熔丝管管体的识别率。针对操作环识别困难的问题,利用熔丝管管体和操作环的几何位置关系,提出了基于位置和几何尺寸约束的椭圆拟合法识别操作环。为消除复杂背景中其他物体对熔丝管识别的干扰,提出了基于熔丝管管体和操作环互相约束的熔丝管识别方法。实验结果表明,所提方法可以识别不同场景中的熔丝管,识别率高且鲁棒性好。

Aiming at the recognition problem of fuse tube of the drop-out fuse for the distribution line maintenance robot,a straight-line fitting approach based on geometric constraint is proposed here. The accuracy of the fuse tube body recognition is improved by using the probabilistic Hough transform combined with the lines parallelism and distance constraints. In order to recognize the operation ring correctly,a novel ellipse fitting method based on position and geometry size constraint is proposed by means of the geometrical position relation between the fuse tube body and the operating ring. To remove the interference caused by other objects in the complex background ,a recognition method for the fuse tube is presented based on the mutually constrain condition of the fuse tube body and the operating ring. Experimental results show that the proposed method can identify fuse tubes in different scenes with high accuracy and good robustness.