基于嵌入式机器视觉的多瓶口缺陷检测系统设计

目的 为了解决瓶口质量人工检测效率低且不稳定的问题。方法 采用HD-SDI的高清视频图像技术设计一种基于频率为70 MHz的STM32硬件检测系统，通过对视频读入、256 MB存储、29.97帧的视频输入输出和报警等硬件电路的设计，基于此搭建视频处理主流程、通信程序和图像处理的一系列软件算法来实现瓶喉口的缺陷检测，通过单张相片中塑料瓶口边缘轮廓来对比分析系统的可靠性。结果 实验证明，系统可以在0.26 s内识别1个瓶口，生产线产出速度为2.4 s/个。结论 该系统可以稳定、实时地检测多个瓶体口缺陷效果。

Design of Detection System for Multiple Bottle Mouth Defects Basedon Embedded Machine Vision

The work aims to solve the problem of low efficiency and instability in manual inspection of bottle mouth quality. An STM32 hardware detection system based on 70 MHz frequency was designed with the HD video image technology of HD-SDI. Through the design of the hardware circuits, such as video read-in, 256 MB storage and 29.97 frame video input and output and alarm, a series of software algorithms for video processing, communication program and image processing were built to realize the defect detection of bottle mouth. The reliability of the system was analyzed by comparing the edge profile of plastic bottle mouth in a single photo. Experiments showed that the system could identify one bottle mouth in 0.26 s, and the output speed of the production line was 2.4 s/pcs. The system can stably detect the effect of various bottle mouth defects in real time.