Automatic Matching Of Seals And Character Recognition

All along,the seal is has the legal effect of personal identity symbol represents the collective reputation,and is widely used in politics,economy,culture in the military field,however,with the rapid development of information technology and computer science and technology,forge seal appear more frequently in the society,headed by laser engraving seals alter chapter is disorder of public security,and to the people and the country has brought a huge loss by automatically identify authenticity stamp can reduce fraud and the crime rate,effective national collective and personal interests.Based on this,this paper studies the automatic matching of seals and the automatic matching of the automatic identification of seal contents.The automatic matching of seals can effectively make use of the feature information of seal images for matching,and determine the authenticity of seals by the degree of empirical matching.As the auxiliary expansion of this paper,the identification of seal content can start from the content of the seal text to achieve the purpose of identifying the fake seal.Firstly,by learning seal matching technology at home and abroad on the basis of this paper proposes a algorithm based on SIFT feature points matching in the early stage of the study,in order to improve the efficiency of matching,to determine the normalization model of seal image to identify the model aims to through the background removing image noise reduction gray level of image binarization bilinear interpolation normalization processing such as image size to realize image pre-processing model when building experiment proved that after the pretreatment of image matching has stronger robustness and stability.Secondly,the SIFT feature extraction of seal matching problem is studied,a series of methods and processes in the interpretation of the pattern matching,focuses on the process of image matching based on SIFT operator,contains six important steps,and,in turn,are:the structure of the scale space and the structure of the differential gaussian pyramid image space local extreme value point detection and extraction of key extreme value point of accurate positioning for the main direction and key point size computing feature descriptor BBF search algorithm combined with K-d tree matching of feature points matching combination of SIFT operator process eventually get seal diagram and the match time.Thirdly,the HS optical flow algorithm as an auxiliary matching is studied,the theoretical background and basic principle of the optical flow method are given,and the calculation method of the optical flow is discussed.Finally,the speed field of the seal image to be matched and the archived image is obtained by using HS optical flow method.Finally,by studying the circular seal and oval seal problem due to the particularity of the two types of seals,the text is difficult to directly extracted from seal and recognition so the paper introduced a new kind of image an algorithm,implements the stamp of the image circle moment transformation,effectively extract the seal characters of secondary units and successfully make the words straight line and then expounds the basic thought and calculation process of BP neural network,a preliminary explore the inline text based on BP neural network to realize identify problems after character segmentation As the development and exploration of the paper,the success rate of character recognition is effectively controlled at more than 80%,which lays an important link for OCR optical character recognition.