不透明瓶质量视觉检测系统研究

针对不透明玻璃瓶传统灯检方式检测精度低，人力成本高的问题，研制了一套基于机器视觉的自动检测系统。研究了内壁检测系统的机械与机电控制系统结构，开发了动态成像系统；提出了基于120°交叉投影中值定位方法，根据内壁污迹、杂质灰度值的分布特征采取专家决策算法进行识别。选择128 mL不透明白色回收瓶为对象进行测试，检测的正确率在99.125%以上，远高于人工灯光验瓶。

Research on quality detection system for opaque glass bottles based on machine vision

With regard to the problems of low detection accuracy and high labor cost in the traditional detection of opaque bottles, automatic inspection system based on machine vision is studied. This paper researches the mechanical and electrical control system structure of the detection system for the inner wall, and develops the dynamic imaging system. It puts forward the middle-value positioning method based on the 120° cross-projection to make an identification by adopting an expert-deciding algorithm according to the inner wall stains, impurities gray value distribution characteristics. The 128 mL opaque white bottle is selected for testing the object. The correct detection rate at 99.125% is much higher than the artificial lighting inspection.