An image-based Accurate Alignment for Substation Inspection Robot

Abstract

The precision of SLAM navigation positioning algorithm based on laser ranging, especially in complex scenes, is not satisfied to meet the accuracy requirements for automatic meter reading and isolating switch status reading. A method of PTZ accurate alignment for substation inspection robot is based on monocular vision camera is proposed. The desired position of robot and orientation of PTZ (Pan Tilt Zoom) camera are defined by the pre-capture reference image. The Scale Invariant Feature Transform (SIFT) algorithm is applied to extract the visual feedback information by matching the current view and the reference image. A Random Sample Consensus (RANSAC) algorithm is used to solve the affine transform between the current and reference image. The Jacobian matrix, representing the mapping relationship, is created by processing control signal of the PTZ control platform and feature space for the image. Besides this, PTZ accurate alignment is realized combined with the image multi-scale transformation. The experiment results of a realistic outdoor substation environment demonstrate the effectiveness of the processed method.