A Degraded Character of Printed Number Recognition Algorithm

Intelligent recognition based on machine vision can't be separated from the image processing and analysis, but the process of images acquired, transmitted and saved is vulnerable to interference of various factors, inevitably there will be degradation and distortion of the image. Specially numeric characters printed on SMD resistor is small, magnified image degradation is serious, even character issues such as adhesions, which brought great difficulties to image recognition. For this class problem, paper proposed a fast efficient of character segmentation and based on gravity and strokes line of classification decision tree recognition method, the method through on pretreated of image for by column scan analysis character of high, wide and features, can be effective to eliminate invalid strokes and segment out more normative single character (including adhesion character of segmentation), then through positioning character upper half part and bottom half part of mass points, analysis strokes anyway line to extraction features, last through decision tree for recognition, experiment proved the method is shortcut, efficient, strong anti-interference ability, robust, and can effectively recognize some faults and deformation of the non-standard characters.