A novel vision based pointer instrument reading algorithm

Abstract

A novel pointer instrument reading algorithm is presented to replace human reading. The image sequences of pointer instrument are captured. Firstly, this paper proposes an image sequences analysis method based on unsupervised learning to segment the moving objects and background. This paper proposes a new method for extracting motion features based on the gray change rules in image sequence. Then a novel unsupervised learning method is used to classify the image sequences into the motion objects and backgrounds efficiently according to these features. Secondly the according to the characteristic of pointer instrument, the rotation central of pointer is get by Hough transform. And then Then the angles of the center of scales and pointer regions relatives to the rotation center can be calculated separately. At last the readings of pointer instrument are calculated from these angles. This study verified the effectiveness of the proposed algorithm through experiments.