摘要
为提高胶囊完整性检测的正确率,提出一种基于机器视觉的检测方法.为避免二值化处理产生的错误分割,采用局部二值化方法,以边缘提取为基础,对投影和形态滤波确定的每个胶囊存在区域进行二值化处理,从而取得了良好的分割效果.设计了一台基于最小错误率的贝叶斯分类器,对胶囊进行分类,以判断药板是否缺损.大量实验表明,检测为合格品的正确识别率为99 95%,识别速率达200粒/s.
In order to improve the correct recognition rate of capsule integrality, a capsule vision recognition system is presented. To avoid false segmentation, based on the edge extraction,the local thresholding method is adopted to deal with the capsule region determined by projecting and morphologic filtering, thus an optimized segmentation result is obtained. And a Bayesian decision system to classify the capsule is designed to judge whether the board of capsules is normal or not. A great number of experiments indicate that the correct recognition rate of this system reaches up to 9995% with a recognition rate of 200 units per second.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2002年第12期1262-1265,共4页
Journal of Xi'an Jiaotong University
基金
国家自然科学基金重点资助项目(60134010).
