Abstract:In the intelligent reading for indicators,it is necessary to first detect two image features such as indicators contour and pointer position.Thereby,one can locate effective region of indicator and determine the position relationship between pointer and scale line.Based on the geometric structural characteristics of indicator image,a new visual method was proposed for detecting indicator image features.Through image preprocessing,the outline region was extracted.The parameters of circular outline contour were obtained based on the symmetry circle detection algorithm and idea of Hough transform.Then within the outline of indicator and by using outlines center as the rotation center of pointer,the center line of the pointer was obtained with Radon transform.Finally,the pointers orientation was identified by using the elliptic characteristic of binary image.This method can be used to accurately and quickly detect the image features of indicator,which provides necessary technical basis for followup intelligent readings of indicator based on human vision.