Abstract: This paper presents a liquid intelligent inspection system based on machine vision and support vector machines ensemble.The inspection system is designed，which can capture image sequences of the rotating liquid in the bottle.The motion regions are identified by motion analysis.And then the features of these motion regions are extracted.The multi-kernel function support vector machine ensemble is used to judge if the motion regions are impurities.The experiments demonstrate the inspection precision of the liquid inspection system is above 97%.

Key words: intelligent inspection, machine vision, support vector machines ensemble, motion analysis

摘要： 提出了一种基于机器视觉和支持向量机集成的液体质量智能检测系统。该系统采用了专门装置采集瓶内旋转液体的序列图像。对采集到的序列图像进行运动分析，确定图像序列中的运动区域。然后提取运动区域的特征，提出多核函数支持向量机集成方法来确定运动区域是否为液体中的杂质。通过实验证明，该液体质量检测系统的检测准确率可达97%。

关键词: 智能检测, 机器视觉, 支持向量机集成, 运动分析