Image Text Recognition System Based On Linux Platform

Image character recognition OCR(Optical Character Recognition, Optical Character Recognition) is widely used. Picture identification such as a DNA sequence;PDF text recognition; automatic plate number recognition; handwriting recognition input on the phone. In recent years the more widely, there will be a cell phone in the field of education applications, the mobile phone to take pictures query answer papers,the first is the image character recognition process. I believe more and more applications will continue to produce magic. The image character recognition only important criterion is the recognition rate. Now picture the character recognition is not100% recognition rate, it has also become a major bottleneck image character recognition application development, and thus to explore a good image character recognition algorithm makes sense.Image character recognition relates to image processing, pattern recognition technology, database storage. This picture text recognition the basic process is generally: to get the picture, noise elimination process, binarization processing, image correction, line search, text positioning, thinning, text scanning feature, query and display text and other processing. The main use of the opencv image processing library under linux, the rich library of image processing and good performance; the text of the signature stored in Mysql database. Image recognition computationally intensive, take up storage will be more. Therefore, performance and resources are also a factor to be considered. After character recognition can be performed if the context vocabulary secondary calibration would be a good way to improve the accuracy rate of recognition algorithm because the spirit focuses on accuracy, not included in this step. In the follow-up work and then consider using this method to further enhance accuracy.Opencv provides basic image processing functions, such as image storage, data matrix, etc. traverse, noise elimination process of smoothing, binarization processing.Corrected image is to straighten the image according to the text of the line, this step is relatively critical step for positioning behind so dependent on image correction. For positioning and text anchor text that is determined text area, convenient scanning text features, refinement process is to refine the text of a pixel, without destroying connectivity. Refining is a convenient feature to scan and do the processing. Text stored in the database correspond signatures and UTF8 encoding.Image character recognition includes punctuation recognition, character recognition, letter recognition. In the mainly for the printed text recognition. The identification applications majority. Where character recognition is more complex,this article uses a unique scanning recognition strokes connected domain and generate signatures based on the location of the stroke. Stroke is a continuous writing made. In stroke recognition process used herein, the straight strokes of the notion that all strokes can be simply divided into lines and curves, text simplified to lines and curves and intersections that is precisely the right composition can simplify the text signature scanning.This paper describes the working process of the whole image character recognition systems, and techniques and algorithms used. Use for opencv the use of the database. Each process is required to validate the test. Pictures can change by showing a clear understanding of the process. After extensive tests to verify its accuracy and stability. The image character recognition systems to meet the original design intent.