Intelligent Book Recognition And Detection System Based On Machine Vision

With the rapid development of the information society,people's desire for knowledge is increasing while improving their own quality.Library has become a priority for people to seek knowledge.With the increase of the number of books and the flow of readers in the library,the open shelf reading mode is widely used in the library.It is a common phenomenon that books are placed in wrong order after use.It is a difficult problem that the library must face to rearrange the misordered books.The traditional book arrangement relies on manual work,which is timeconsuming,wasteful and error prone.With the development and popularization of machine vision technology,the extensive application of intelligent technology and the continuous improvement of software and hardware performance,it is possible to automatically identify and sort out the misordered books based on machine vision.The purpose of this paper is to design and implement the machine vision based intelligent Book Identification and detection system,so as to improve the automation of the misordered book sorting.This thesis around the machine vision in the wrong order book to identify the key issues in the study,in view of the book edge as a straight line prior conditions,using edge operator with spine straight line detection algorithm of image are extracted,and the target detection algorithm based on Faster-RCNN,fixing the call number region extraction,to improve the accuracy of extracting the call number.The character segmentation technology in the extracted call number region was studied.In order to improve the accuracy of the system in character segmentation,the call number was segmented based on the connected domain segmentation algorithm.Then the segmented characters are processed for character recognition based on convolutional neural network.The identified call number information interacts with the pre-stored book order information in the system database,and the detection results are fed back.Based on the above technologies,this paper uses the C++ language,Caffe deep learning framework and combines the database technology to complete the design of a book error detection system.The system is divided into three basic functional modules: pre-processing module,recognition module and post-processing module.The preliminary test of the system shows that the overall scheme and the development system proposed in this paper can be used to detect the basic information and the order information of the books in the universal call number style,and provide a feasible scheme for the automatic arrangement of the library.