摘要
提出了一种基于饱和度的异色米粒检测方法。该方法由于利用米粒的一维饱和度特征将米粒分成正常白米粒和异色米粒两大类,因而,可通过一个分类阈值同时检测出米粒中的各种异色米粒。由于利用正常白米粒和异色米粒两类饱和度特征优化了分类阈值,而且还通过扫描式区域增长的办法来提高单个米粒的饱和度特征,因而,可在群米粒中检测出单个异色米粒,且检测精度高。实验结果得出该方法对异色米粒检测的检测精度为97.8%。
A method of detection of chromatic rice kernels based on saturation is proposed, which can simultaneously detect various chromatic rice kernels only by a classifying threshold, because rice kernels are classified into two categories as normal （or white） kernels and chromatic kernels according to their 1D saturation features. Due to the optimized classifying threshold based on saturation features of the normal kernels and chromatic kernels, so that the detection accuracy of the chromatic kernels is higher. Because the saturation features of the individual kernels are extracted through region growing with scan ning row by row,the chromatic kernels among groups of rice kernels can be individually detected. In our experiment,the pro posed method was tested on over 700 rice kernels with normal kernels and various chromatic kernels,the detection accuracy of the chromatic kernels was 97.8%.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2007年第1期97-99,共3页
Journal of Optoelectronics·Laser
基金
国家自然科学基金项目(60627002)
天津市应用基础重点项目(043800511
06YFJZJC00400)
关键词
饱和度
正常米粒
异色米粒
大米检测
图像处理
saturation normal rice kernel chromatic rice kemel rice kernel detection image processing
