Recognition of dot-matrix character-degraded based on affine registration

Dot-matrix characters are widely used for steel slab labeling. Different from the universal stroke characters, the shape and quality of dot-matrix characters are prone to being degraded by various factors, which makes it difficult to recognize them by conventional methods. In this paper, a new dot-matrix characters-degraded recognition method is proposed. Firstly, this method adopts the affine transform model to guide the transformation of degraded characters with their references. Then, an objective function is designed to iteratively optimize the desired correspondence. Finally, according to the alignment performance, a scoring mechanism is proposed to evaluate the recognition results. Various experiments were conducted and the results verify the suitability of the proposed recognition method for dot-matrix characters with degradation like distortion, tilt and dot-missing.