Research On General Technology Of Self-adaptive Extraction Of Bill Seal Content

In view of the fact that there is no problem of the extraction method of the general high-precision bill seal and the extraction method of the general high-precision bill seal content,the paper proposes a series general method of self-adaptive extraction of bill seal content.These method includes extraction technology of bill seal based on different degrees of clarity and a self-adaptive extraction technique of a bill seal contents that does not rely on a seal shape.That is the bill content extraction method based on playback technology of optimal point.The extraction technology of bill seal based on different degrees of clarity includes the method of threshold selection bill seal extraction method.This method can make the best choice for different clarity of the bill seals and ensure to gain the high accuracy seals.The bill content extraction method based on playback technology of optimal point mainly utilizes the window filtering technique of convolution layer in the convolution neural network.Seal content is located and divided by finding concentrated area of the pixel points in the seal.Therefore,this method need not rely on the shape of seal to extract the seal content so that it can be used in various shape of seals.The self-adaption of algorithm is mainly following two aspects.In the one hand,On the process of extracting the seal from the checks,this algorithm realizes the adaptive selection of seal extraction method by a threshold selection method.On the other hand,in the process of segmentation and positioning of seal content based on the playback technology of optimal point,the paper realizes self-adaption in the aspects of selecting a series of parameter,such as the size of the filter window,the width of the seal frame,and the distance between the best points.It means that the reasonable parameter can be selected on the basis of the different seal image when the seal content is extracted.Meanwhile,in the part about correction and sorting for the output text,this paper proposes a method for correction and sorting of seal characters which is universal in cambered seal characters and does not have to convert to polar coordinates.Experiments show that the universal method of self-adaptive extraction of bill seal content is feasible and effective,and it has good universality.When using OCR to identify the seal contents extracted from this paper,the correct recognition rate is as high as 85%.