一种基于全卷积神经网络的横担姿态测量方法

Attitude measurement for cross arm based on fully convolutional network

为满足配电线路维护机器人更换避雷器的作业需求,提出了一种基于全卷积神经网络的横担姿态测量方法.通过分析三维几何特征建立了横担姿态模型;采用基于全卷积神经网络的图像分割方法获得横担区域,并以此作为掩膜进行边缘检测,去除环境干扰;采用基于投票法和霍夫空间约束的直线检测提取横担主体区域轮廓直线,并给出了求解横担姿态向量的算法.实验结果表明:所提出方法能较为精确地测得横担姿态,为机器人自主更换避雷器奠定了良好基础.

Aiming atmeasuring the cross arm attitude for live-line maintenance robot in lightning arrester replacement operation,an attitude measurement method for the cross arm based on the fully convolutional network was proposed.The cross arm attitude model was built by analyzing its three-dimensional geometric features.The image segmentation method based on the fully convolutional network was applied to obtain the cross arm region,which was used as a mask for edge detection and removes environmental interference factors effectively.The line detection based on voting method and Hough space constraint was used to extract main contour lines of the cross arm,and the algorithm to find the cross arm attitude vector was also given.The experimental results show that the proposed method can accurately measure the cross arm attitude and lay a good foundation for the robot to replace the lightning arrester independently.