摘要
针对高速医药生产线上安瓿溶液中可见异物的在线检测与跟踪的速度和精度要求,提出概率阈值分割与卡尔曼滤波器预测相结合的检测与跟踪方法:应用初次差分结合概率加权阈值分割去除静态干扰,进行二次差分检测目标;利用初步得到的目标信息通过卡尔曼滤波器预测,在卡尔曼滤波器的预测点附近进行搜索并识别,减少卡尔曼滤波器的学习时间,并同时缩小搜索范围,达到提高检测效率的目的。实验结果表明,该方法可以满足检测精度要求并且满足在线检测速度要求。
To satisfy the speed and precision requirements of detecting and tracking foreign substances in ampoules,a detection and tracking method that combines probability weighted threshold and Kalman filter is introduced.First,difference and probability weighted threshold are used to get rid of static noise,then second-difference is applied to detect foreign substances;at last,Kalman filter is used to predicate next probable position from old information.The search work is expanded around the predicated position so as to reduce detection time and improve detection efficiency.Lots of experiments verify that the proposed method can satisfy the detection accuracy and speed requirements.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2011年第3期488-494,共7页
Chinese Journal of Scientific Instrument
基金
国家863计划项目(2007AA04Z244)
国家自然科学基金(60835004)
湖南省研究生科研创新基金(CX2009B0703)资助项目
关键词
概率阈值分割
卡尔曼滤波器
异物检测
probability weighted threshold
Kalman filter
foreign substance detection
