Abstract: In order to detect visible foreign substances in oral liquid, which is packed in cylindriod vial, an automatic recognition and sorting system is developed. Using multiple detection station and multiple camera tracking shot mode, it can avoid error judgment effectively. Due to the shape change of plane image caused by the movement of visible foreign substances, the paper presents a kind of detection algorithm aiming at small targets in the liquid. The detection target is obtained via complex difference. A kind of particle filter tracking algorithm based on weighted feature is proposed, overcoming error tracking by single feature, it can distinguish interference and visible foreign substances effectively. Experimental results show that this method can satisfy the speed and accuracy demand of online detection.

Key words: machine vision, visible foreign substances, intelligent detection, complex difference, weighted feature, particle filtering

摘要： 针对西林瓶装口服液中可见异物的检测，开发一种基于机器视觉的智能识别和分拣系统，采用多检测工位和多相机跟踪拍摄的方式，以避免误判断。根据可见异物运动导致的平面图像中形状的变化，提出一种液体中微小目标的检测算法，通过多帧复合差分确定检测目标，利用基于异物加权特征的粒子滤波跟踪算法，克服单一特征导致的误跟踪，准确区分干扰和可见异物。实验结果表明，该方法可以满足在线检测的速度和精度要求。

关键词: 机器视觉, 可见异物, 智能检测, 复合差分, 加权特征, 粒子滤波

CLC Number: 

• TP391