变电站巡检机器人盲区压板识别方法及应用

An Identification Method of the Plate in the Blind Zone of Substation Inspection Robots and Its Application

为解决变电站主控室内轮式机器人拍摄继电保护屏底部压板时存在视觉盲区的问题,提出了一种基于除法模型矫正的广角副摄像机拍摄方法。该方法首先通过在轮式机器人侧面安装广角副摄像机实现继电保护屏底部视觉盲区的压板图像信息采集,然后采用除法模型对所采集的图像信息进行矫正,最后采用卷积神经网络算法对压板进行识别,实现了继电保护屏压板识别的全覆盖,提高了压板识别的准确率。在某220 kV变电站进行了压板核对实验,结果表明,采用该方法改进的轮式巡检机器人解决了视觉盲区压板的核对问题,压板核对准确率达到100%,相比原有轮式机器人巡检覆盖面更广、结果更准确,具有良好的推广价值。

In order to solve the problem of visual blind zones when the wheeled inspection robot shoots the bottom plate of relay protection screen in the main control room of the substation,a wide-angle secondary camera shooting method based on division model correction is proposed.Firstly,a wide-angle secondary camera is installed on one side of the robot to collect the image of the visual blind zone at the bottom of the relay protection screen.Then a division model is used to correct the collected image.Finally,a convolutional neural network algorithm is used to identify the plate.The proposed method realizes identification of all plates on the relay protection screen and improves the accuracy of identification.A plate check experiment was carried out in a 220 kV substation.The results show that the wheeled inspection robot improved by this method solves the problem of shooting plates in the blind zone,achieving 100%accuracy in term of plate state recognition.Compared with the original inspection method,both coverage and accuracy are improved,demonstrating the potential of further application of the proposed method.