DM642-based English Text Recognition System

With the development of computation technology, more and more electronicproduct and intelligent devices are used frequently in daily life and begin to be ainevitable part of our life, they greatly facilitate our daily work. In terms of humanintelligence, text recognition is a very important area. The prospect that machine issmart enough to learn from handwrite or print text and thus is able to communicatewith reality is more than mere stirring. Actually, it is partly technical realizable, butthere is still a lot of problem in this area.In this paper, a image system is built based on DM642，which is a type of DSPchip designed by Texas Instruments for image processing, and then necessaryimage-processing codes is finished to fulfill the task of image pre-processing,bounding box acquisition,document skew estimation and other steps so as to allowthe system “read” the text information from the text image. The proper selection ofhardware and improvement in recognition algorithm award the system a higherrecognition precision and faster speed. Generally, the paper refers to:(1)Select proper hardware according to system requirement in recognition speedand precision, and then build a system based on acquired documents on CMOShigh-definition camera and DM642, which is divided into three parts of imageacquiring system, image processing system, image feedback system.(2)Select proper image processing algorithm for text image based on long timestudy on text image processing algorithm proposed by other researchers. Firstly,connected components of each character are acquired through block adjacent graph,then document skew is estimated through coarse estimation and refine estimation andcorrected by rotating pixel blocks other than pixels.(3)Perform simulation on MATLAB platform to evaluate the effectiveness andperformance of the proposed image-processing algorithm, and then developcorresponding codes for DM642chip, and test it on built system. Test results showsthat the system works well in text image acquisition and recognition in a speed that’s50%faster than similar text image processing system built based on ARM series.(4)Check all design and algorithm, and give a summary on its advantages anddisadvantages, and make a prospect for further work.