面向配网带电作业机器人的导线识别定位方法

Power Line Identification and Localization for Live Working Robot of Distribution Line

由于配网带电作业机器人作业过程中相机受阳光干扰严重,导线等细长弱纹理作业目标图像特征点匹配困难,导线识别定位困难。该文提出了基于图像预匹配的点云生成方法,使用Improved-DCEnet无监督网络进行光照无关处理抵消阳光干扰。在图像匹配阶段,使用VGG16-Unet卷积神经网络来进行导线的图像分割避免复杂背景干扰,通过图像几何矩提取拟合导线中心线,使用导线中心线与双目光心分别确定双目的导线物像面,双目导线物像面交线即为导线的真实位置,将复杂背景下细长弱纹理物体识别定位转换为一个几何问题。通过实验室环境下大量实验标定后,该方法可以实现相机坐标系下xy方向上1mm、z方向上3mm综合5mm定位精度。最后该方法应用在真实场景中完成了导线的识别定位,实现了机械臂自主完成实际接火作业,提高了配网带电作业机器人的导线识别稳定性。

Since the working of the camera is seriously disturbed by the sunlight during the operation of the live working robot on the distribution network,it is difficult to match the image feature points of the slender and weak texture operation targets such as the wires,and it is difficult to identify and locate the wires.In this paper,a new method of point cloud generation method based on image pre-matching is proposed,which uses the Improved-DCE net for illumination-independent processing to offset the sunlight interference.In the image matching stage,the VGG16-Unet convolutional neural network is used to segment the power line image to avoid complex background interferences.The power line center line of the wire is extracted and fitted by the image geometric moment,and the power line center line of the wire and the binocular center are used to determine the object image plane of the binocular power line object image plane.The intersection of the object image plane of the binocular power line object image plane determines the real position of the power line wire and transforms the identification and positioning of the slender and weak texture objects in a complex environment into a geometric problem.After a large number of experimental calibrations in the laboratory environment,the method is verified to achieve a comprehensive 5mm positioning accuracy of with 1mm in the xy-direction and 3mm in the z-direction.Finally,the method is applied in the real scene to complete the identification and positioning of the power line and realize the robot arm’s to complete the actual power line connection autonomously.It is also applied to improves the power line identification stability of the live working robot in the distribution network.