Abstract: Character recognition which applies to library management system can improve the level of digitalization and automation effectively.This paper designs with the purpose of recognizing multiple books by one operation.Firstly the image of multiple books is segmented into images of a single book based on statistic after edge link.Then the character region is extracted by the feature of RGB space，and the region is segmented into single character according to the contour.The last step is training BP neural Network to recognize the segmented characters.The processing flow of this character segmentation and recognition system is built in Matlab.Simulation results indicate that this technique can recognize the bookmarks basically，which has a good practical value and application prospect.