Substation inspection robot walks along the inspection road to carry out equipment inspection, stops at the scheduled monitoring position, calls the PTZ preset position to align the monitoring instrument with the monitored equipment, and the visible light detection instrument obtains a clear picture with appropriate proportion through optical zoom. . Due to the combined effects of robot navigation positioning accuracy, travel attitude deviation and PTZ preset position accuracy, the position of the same observation target at each inspection has a large deviation in the position of the entire image field of view. Hough circle detection, as a common method of circle detection, can detect the circle in the image very well. However, in order to achieve good detection results, the parameters such as search area and search radius must be a value obtained by many human experiments, so it is not suitable for images in transformer substation with complex background and unknown circle size. In this paper, Hough circle detection and perceptual Hash similarity algorithm are combined. From many results of Hough circle detection, the target meter similar to the circular meter in the reference image is selected, and its location is obtained. Then the position deviations between the circular meter and the one in reference image are calculated, and the second alignment of the meter is realized by turning the platform.