Approach for Machine Vision Recognition of Severely Skewed Characters

Abstract

The result of this research is the development of a vision system for recognizing severely skewed characters. A series of image enhancement and reconstruction algorithms expand a skewed image in both the x and y directions. The expanded image is recognized and identified by comparison to pre-stored ideal character templates. The vision system reconstructs and recognizes various alphanumeric characters which are tilted with respect to the camera. The data gathered in this research demonstrate that the system's image reconstruction and recognition performance is acceptable for practical use. The only misgiving is the system's poor time efficiency, caused by the great number of repetitious calculations required in many of the algorithms. However, since the algorithms were chosen for adaptation to parallel processing, the outlook for future research and improvements is promising.