Abstract: In this paper, an automatic oral liquid light inspection algorithm is designed to implement the detection of particles in oral liquid. Second-difference and accumulate difference methods are used to remove the background information from the sequence images captured by a high-speed industry camera, and the target images with particles, air bubble and random noise are acquired. Adaptive threshold processing is used to extract moving targets. Real-time detection is completed by determining whether the size of particles, such as hair, fibers and so on, is larger than the pharmaceutical standards. A large number of experimental results show that the proposed algorithm is effective and feasible.

Key words: machine vision, oral liquid, real-time detection, difference and energy accumulation, adaptive threshold processing

摘要： 设计了一种基于机器视觉的口服液异物检测算法，实现口服液中异物的检测。通过高速工业摄像机获取序列图像，利用二次差分与累积差分法对序列图像进行处理，去除图像中的背景信息，得到只含杂质、少量气泡和部分噪声的灰度图像。通过自适应阈值处理提取运动目标。通过比较异物的大小是否超过医药标准来判断药剂中是否含有异物（比如：毛发、短纤维等），来实现对口服液中可见异物的实时检测。通过实验与现场测试，验证了该检测算法的有效性和可行性。

关键词: 机器视觉, 口服液, 实时检测, 差分与累积, 自适应阈值处理