Research On Character Extraction And Recognition Of Blurred Image In Procuratorate

With the development of information technology in the world,China’s information construction has entered a period of rapid development,information construction has been vigorously promoted in various industries.The image materials provided by the work of the Procuratorate are mostly electronic photos of paper information,which is difficult to provide effective electronic information intuitively.In order to extract text and other effective information from the image,we need to use advanced modern technology to facilitate the work of the procuratorate.Therefore,it is very necessary to apply the "intelligent procuratorial" technology to the information extraction of procuratorial image materials,which is of great significance to realize the information construction of procuratorial organs and improve the work quality and efficiency of procuratorial organs.In this paper,the character region extraction of different features of inspection service image is studied,and the extracted characters are recognized,The specific work includes the following aspects:(1)Aiming at the problem that the complex background in the fuzzy prosecution image affects the accuracy of character region extraction,a character extraction algorithm based on cellular immunity is proposed.The algorithm calculates two adaptive thresholds based on the cellular immune mechanism,one is used to segment the image background to remove the interference of background information on the extraction of character regions;the other is used to extract the character regions.The proposed algorithm draws on the principle of cell-specific immunity and solves the problem that traditional threshold segmentation algorithms cannot eliminate background interference.(2)Aiming at the problem of the interference of the red seal in the fuzzy inspection service image,a character extraction algorithm based on quantum mechanics and color clustering is proposed.The algorithm compares image pixels to electrons outside the nucleus,and uses the movement rules of electrons outside the nucleus to guide the clustering of image pixels.Finally realize the use of image color features to extract the character region.(3)In order to improve the recognition rate of characters in the inspection service image,the characters are segmented on the basis of character region extraction,and recognized by convolution neural network(CNN).In this paper,we improve the lenet-5 network and add a normalization layer after each convolution layer in the traditional lenet-5 network model.The normalized layer is used to converge the convoluted data to prevent the gradient explosion and disappearance,so as to improve the network training speed and the accuracy of network recognition.