Character Recognition Method for Low-Contrast Images of Numerical Instruments

Abstract

In this paper, a method of preprocessing to extract numbers from a low-contrast image of numerical instruments and recognition based on skeleton structural features is proposed. The approach is based on the grayscale dilation operation which can enhance the difference of targets with the background of the image and eliminate the adhesion between the target and the environment. After that, a skeleton algorithm is used to highlight the shape and topology of the target which makes the target number easier to be identified. Using HALCON to test, the result shows that the identification has a high accuracy rate.