Machine vision detection of pointer features in images of analog meter displays

In automatic and accurate reading recognition of analog meters based on machine vision, one of important issues is the detection of pointer features, which includes the meter center and pointer image processing. The current automatic-recognition approaches to reading analog meters often consist in locating the meter center based on the dial region or its border. The located center is not coincident with the rotation center of pointer which leads to inevitable reading errors. In the paper, the centripetalism of annular scale lines is used to calculate the position of the pointer rotation center. First, it uses the region growing method to locate the dial region and uses the eccentricity measure to extract annular scale lines. Second, the parameters of these scale lines are estimated with the Hough transform method. Then, the common intersection of a group of lines, i.e., the meter rotation center, is determined with the maximum probability criterion. Finally, the pointer centerline and direction are detected through the calculated center and the Hough transform results. The simulated and experimental results demonstrate that the proposed method can accurately locate the pointer rotation center and obtain pointer centerline. Moreover, it is applicable to the meter image captured under a slant camera view or with uneven light illumination