A robust pointer meter reading recognition method for substation inspection robot

Abstract

A robust method of pointer meter reading recognition was proposed for the inspection robot in intelligent substation. The new method consists of two stages. Stage 1: all scale lines in a template image were marked artificially and all centroids of scale lines were figured out. A least-square method with random sample consensus (RANSAC) was used to fit the meter circle using the centroids of all scale lines, and then the fitting circle and the centroids of min-value scale line and max-value scale line were saved in the database. Stage 2: The speeded up robust features (SURF) method was utilized to match with the template image for detecting the meter region from the real-time image captured by the inspection robot. Consequently, the pointer of the meter was extracted from the processed image. An image thinning method was utilized to thin the pointer image. Finally, the Hough transform was used to detect the pointer, and the result of the analog meter can be recognized. The least-square method with RANSAC and the image thinning method can eliminate the noise of the image and improve the robustness of the proposed method.