大输液在线视觉检测系统的研究

针对传统灯检检测精度低，人力成本高的问题，研究了基于机器视觉的大输液中可见异物在线检测系统的关键技术。利用高速旋转装置，获取运动目标的序列图像。根据图像间的交互方差系数MV建立序列图像的背景，利用背景差分提取检测目标。综合质心迭代算法和卡尔曼滤波器实现了运动目标的跟踪，根据运动轨迹方向确定可见异物是否存在，排除气泡的干扰。实验结果表明该系统能够有效地实现大输液中可见异物的实时检测功能。

Research on on-line visual system to detect transfusion solution

With regard to the problems of low detection accuracy and high labor cost in the traditional detection,key technologies of an on-line system to detect foreign substances based on machine vision are studied.The system employs high-speed rotational devices to obtain the sequential images of moving targets.According to Mutual Variance（MV） parameter between images,the background of sequential images is set up.By subtracting the background,inspected objects are extracted.Combining centroid iteration algorithm with Kalman filter realizes tracking motional particles.By means of the direction of targets＇motion tracks,whether visible foreign substances exited are determined and the disturbance of bubbles is eliminated as well.The experiments demonstrate that the system could fulfill the real-time detection of visible foreign substances in transfusion.