摘要
为解决卷烟生产过程中条烟存在的小包缺失、小包破损、烟支缺失、烟支错排等问题,基于条烟的X光图像设计了一种新的特征提取算法。通过模仿人眼识别原理,从视觉感知角度对物体边界区域、稠密性、分布状况、方向属性和规律程度进行特征提取,简称VPFE(Visual Perception Feature Extraction)算法。基于VPFE利用不同分类器对条烟生产中的异常情况进行检测试验,结果表明:1VPFE算法在不同分类器下均有较好的检测效果,可以被主流分类器用于学习分类,且检测精度优于其他特征提取算法。烟支缺失和小包缺失情况的检测精度达100%。2VPFE算法还可对破损面积在20%以上的小包以及烟支错排情况进行检测,小包破损检出率达95.8%,烟支错排检出率达96.6%。该算法有效满足了条烟异常情况检测的精度要求,杜绝了条烟缺支和缺包现象,提升了卷烟包装水平。
Aiming at the problems, such as broken packet, packet missing or not properly filled, might happen in the process of cigarette production, a VPFE （Visual Perception Feature Extraction） algorithm based on the X-ray images of cigarette cartons was put forward. VPFE algorithm extracted the features of objects from the angles of visual perception including boundary, density, distribution, direction and regularity. Different classifiers were used to detect the abnormalities in cigarette cartoning. The results showed that： 1） VPFE algorithm offered pretty good detection effects under different classifiers, it could be applied to classification learning by mainstream classifiers and advantaged over other feature extraction algorithms in detection precision, which reached 100% for the detection of cigarette missing and packet missing. 2） VPFE algorithm could also detect the packets with a broken area over 20% or cigarette misalignment with detection rates of 95.8% and 96.6%, respectively. Cigarette or packet missing in cartons is eliminated, and the quality of cigarette packaging is promoted.
出处
《烟草科技》
EI
CAS
CSCD
北大核心
2016年第1期78-83,共6页
Tobacco Science & Technology
关键词
条烟
视觉感知
X光图像
异常检测
分类器
特征提取
Cigarette carton
Visual perception
X-ray image
Abnormality detection
Classifier
Feature extraction
