摘要
设计了一种基于机器视觉技术的全自动智能灯检机,以实现药液中可见异物的在线检测。首先对采集到的各帧图像通过形态学滤波、帧差等预处理抑制大部分背景噪声。然后提取图像中目标的不变特征,通过各目标不变特征间的匹配,实现相邻帧点迹的关联,得到若干假设的目标运动轨迹。最后根据异物运动轨迹的连续性和平滑特点来确定是否存在可见异物。研制的全自动灯检机样机已经在某药厂的注射液生产线上试验运行。实验表明,该机器的识别准确率和速度均高于熟练的灯检工。
To detect visual particles in liquid in pharmaceutical containers on line, an intelligent automatic inspection machine based on machine vision was designed. First, image pre-processing techniques such as morphological filtering and multi-frame difference were used to restrain most of the background noise. Then some invariant features of the objects in the images were extracted. Through matching these features in adjacent frames the hypothetic motion trajectories of the objects were obtained. Finally, the existence of visual particles is determined according to the continuity and smooth characteristics of the trajectories of foreign bodies. A prototype inspection machine was developed and tested on a pharmaceutical production line. Experiment results show that the recognition accuracy and speed of the machine are higher than the skilled workers.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2008年第3期562-566,共5页
Chinese Journal of Scientific Instrument
基金
山东省自然科学基金(Z2006F05)资助项目
关键词
机器视觉
图像处理
形态学滤波
全自动灯检机
machine vision
image processing
morphological filtering
automatic inspection machine
作者简介
杨福刚，1999年、2005年于山东理工大学分别获得学士学位和硕士学位，现为山东大学博士研究生，主要研究方向为机器视觉检测和图像处理。E—mail：yangfug@163．com。孙同景，1982年于山东大学获得学士学位，现为山东大学教授，博士生导师，主要研究方向为智能控制技术。E—mail：Tj_sun@sdu．edu．cn。
