Study On Online Uyghur Handwritten Character Segmentation Based On Local Trajectory Information

Although Uyghur script recognition technology has achieved appreciable research results in print character recognition,but the research on handwritten Uyghur word recognition is still in the primary research and exploration stage.The recognition of handwritten words is a very complex research topic,and there are many contents that need to be studied.This paper mainly carries out a series of research work in the direction of character segmentation and has obtained some research results.This article is mainly from the following aspects to carry out:(1)Research on the segmentation of offline handwritten words,based on the commonly used vertical projection segmentation method,this paper proposes an over-segmentation method based on the baseline position of the vertical projection.Firstly,the horizontal projection of the binary image is used to find the baseline.Then the vertical part of the baseline is projected to find the location of the slicing point.Although the simulation results of this algorithm show that the segmentation effect is not ideal,it still has some reference significance for subsequent researchers,reminding them not to try this segmentation idea in the research in this field.(2)According to the writing characteristics of Uyghur,this paper proposes an over-segmentation algorithm based on local trajectory information.In this algorithm,the method of combining horizontal segmentation and vertical segmentation effectively reduces the occurrence of neglected situations.The experimental result of 69.97% of the average correct segmentation rate shows that this method can effectively solve the impact of the correct segmentation point on the searching for the segmentation pointwhen it is folded.At the same time,the correct segmentation rate for the overall tilt of the word is improved,and the method also successfully avoids the occurrence of over-segmentation.