Research And Development On The Image Recognition System Of The Medical Bottle Cap Defect

Medical aluminum and plastic combined cap is a new type of bottle caps which is combined by hot pressing riveting of aluminum caps and plastic ones. Such caps are mainly used in the antibiotic powder for injection, infusion, freeze-dried and oral liquid preparations’ bottle packages. At present, the unqualified products are detected through manual work. But this method has many problems, such as its inefficiency, high rates of missed detection, and detectors’ visual fatigue, which can affect the testing results and the quality of products. In order to achieve faster and higher precision automatic detection, there is an urgent need to develop a set of detection system, with optical imaging, image acquisition and digital image processing and analyzing, as its platform. This not only has theoretical significance, but also has great economic value, and it can help enterprises reduce production cost and improve the quality of the cap product. So the aluminum-plastic combination cap defect detection system, which is based on the image processing, is studied, and the method of identify defects is also discussed in this dissertation.The cap defects are shown in three aspects, which are the inner side, the outer side and the side face. The inner side defects include lack of six blades and collapse, the outer side defects are parting-line flash and plastic damaged, and the side face defects are small lack of material, nest side and wrinkle.According to the characteristics of the defects and the actual production requirements, firstly, determine the overall design of the system, including the overall structure of the system, hardware selection, then design a segmentation algorithm, image preprocessing, recognition algorithm for a medical cap, and finally make a design for the software systems.When a cap moves to the front of the photoelectric sensor, the cap is following the conveyor movement on the conveyor belt. The photoelectric sensor is triggered, and it transmits the signal to the camera simultaneously. However, the camera doesn’t collect the image, unless it gets the permit of the program. Only the program completes the acquisition of the image could the camera begin to accept the sensor signals. As each image contains two caps, using the time difference, the time of the processed image can be set less than the time of two caps moving after the sensor, which ensures that the detection system is able to detect every cap.As for the image processing, by simulating the production line of the cap and analyzing the collected cap images’ characteristics, a suitable algorithm is studied which can segment image to get complete bottle cap image. According to the characteristics of the detection system, the image moment invariant features are combined with the perimeter, area, and other characteristics of gray scale image to match the image. That can not only ensure the matching precision, but greatly improve the matching speed, which has a strong resistance to gray and geometric distortion, and also has a strong ability of restraining noise.Finally, the cap defect detection system is studied, and the experimental results show that the algorithm can meet online real-time testing requirements.