一种基于图像轮廓线的稻种模糊模式识别方法

该文对二值图像轮廓线的提取,提出了一种简捷的算法。在对稻种轮廓线大量试验分析的基础上,确定了以轮廓线中心到轮廓线距离最大的线段为参考,12个方向上的线段长度和轮廓线长度组成特征向量。对模糊模式识别中的欧几里德距离算法提出了两种改进方法,并根据最大隶属度原则,实现稻种的模糊模式识别。试验结果表明:改进后的算法对浙852、Z94-35、广陆矮4号三类种粒的正确识别率分别达到79.89%,89.63%和93.27%。该方法同样适用于果品、机械工件等对象的模糊模式识别。同时该特征参数的确定方法中已经考虑了对象放置方向和放置面的随意性,并将对象按统一方向和统一面识别,因此该方法还可以进一步发展为对象旋转方位角的测定和对象中任意预定义的几何参数的光学测量

Fuzzy Pattern Recognition Method Based on Image Contour Line

A simplified method was put forward on a binary image contour line extraction. Distance between centroid and dots on contour line was calculated, and based on the maximum centroid dot distance, a feature victor including twelve direction segments and lengths of contour line was studied in this paper, and two distance algorithms based on euclid distance were promoted. The experimental results show that the correct recognition ratios on three kinds of paddy seed reach respectively 79.89%, 89.63% and 93.27%. This method can also be applied to fuzzy pattern recognition on fruits, machine parts etc.. As random of object's lay direction and side were taken into account in this paper, and objects are recognized at the same direction and on the same side, so this method can also be promoted to the optic measurements on object rotation angle and any preindicated geometrical parameters.