基于平行四边形对角线理论优化的直线拟合角点检测

Corner Detection of Line Fitting Based on Optimization of Parallelogram Diagonal Theory

在汉字图像识别的过程中,角点的检测有助于分析汉字的特征,反映出汉字图像的各个部分的重要信息。在以往的角点检测算法中通过曲率来判断角点的算法有不错的效果,但存在钝角点或者是圆弧处的角点检测不够精准。在自适应直线拟合角点检测的基础上通过对圆弧处的角点进行平行四边形对角线曲率估计来筛选出最终的角点集合。通过对比已有的角点检测算法,在兼顾时间效率的基础上,所提出的方法对于汉字的角点检测精确率更高,应用效果更好。

In the process of Chinese character image recognition,corner detection helps to analyze the characteristics of Chinese characters and reflect the important information of each part of the Chinese character image.In the previous corner detection algorithms,the algorithm for judging corners by curvature has good results,but the detection of corners at obtuse corners or arcs is not accurate enough.On the basis of self-adaptive straight line fitting corner detection,the final corner set is filtered out by estimating the diagonal curvature of the parallelogram at the corners of the arc.By comparing the existing corner detection algorithms,on the basis of taking into account the time efficiency,the proposed method has higher accuracy and better application effect for the corner detection of Chinese characters.