Character Recognition Technology And Its Application In Machine Vision Measurement And Controlling

The concept of Optical Character Recognition (OCR) has been proposed in 1929, after decades of development, the technology is relatively mature, present mainly used in office automation applications, license plate recognition, identification notes and so on. With the popularity of automation and high-tech applications such as internet of things, there is huge potential market for character recognition technology in industrial applications, the main focus of this study is the character recognition technology in industrial applications.This topic comes from the research cooperation project "Research and development of embedded machine vision controller" (contract number:07398). After collecting images,this paper reduces noise,adjusts round ring or inclined characters into horizontal characters based on the ROI; for the inkjet and printing character,we introduce the concept of expansion wave to split into right branchs, and then extract the individual characters of each line of information; For the adhesion characters, this paper, can effectively separate adhesion character through a practical and fast segmentation method ;in the process of character feature extraction, integrate structure and statistical characteristics, complement a variety of features, proposed location of the projector features accroding to the characteristics of the object identified,have a good effect for an alphanumeric identifier; classify the characters through the holes and crossing feature,divided into several small character set and then separately weighted,identify numbers and letters using the traditional template matching,for some characters, use the BP neural network classification.After the completion of some algorithm,design the man-machine interface and the font to constitute an.integrated system.This topic study in accordance with the engineering ideas,after identification of program simutlte on Matlab platform, and be transferred to the host computer and underlying DSP platform.this paper have Repeated tests on algorithm as well as running time,the character recognition rate is higher than 99%,the running time is in 10ms for the images of 640*480,meeting the requirments of actual site,sufficiently validated the usefulness of the subject.