Capsule Inspection System Base On Machine Vision

In recent years, machine vision technology is more and more used in drugs’ production and inspection. However, the quality inspection of vacant capsules still rely on manual way. Not only the quality of inspection can’t be guaranteed and the efficiency is low, but also increases the costs of production and difficulties of management. In order to solve this problem, this paper developed a capsule inspection system based on machine vision. It can replace the manual inspection, and has very high practical value.In this paper, after an analysis on current research, the existing shortage of all kinds of research is summarized. On this basis, hardware and software of the inspection system and the related image processing algorithms are designed.In terms of the hardware system, the overall structure and the main operation mode of the system is given. In order to get a complete collection of the capsule surface image, a rolling type transmission structure and a reasonable trigger method of cameras is designed. Then, parameters of visual part’s hardware are analyzed and selected. At the same time, the way of lighting and image acquisition is reasonably designed to confirm the requirements of the system. A corresponding elimination device is designed according to the structure of transmission device. Finally, design of the serial port is given to control the external hardware.Software system is the basis for realization of each system function, related to the stability, real-time of inspection. Base on the program framework and the design of database, C# is used to complete the various interfaces of the program and realize all functions of the system, combined with serial port communication and task parallel processing technology. A corresponding elimination logic to the system’s operation mode is also added to the program.The image processing algorithms are the core of the capsule inspection. According to the characteristics in the images of each type of capsule, threshold segmentation, erosion, contour extraction and region filling algorithm is proposed to extract the image of capsule and it has good robustness. After segmenting the image of capsule body with rotation and first-order difference, the smallest enclosing rectangle is used to measure the size of capsule. Then the local adaptive threshold algorithm is improved for defect detection on capsule’s cylinder and good effect is obtained. Defects on the ends of capsule has different effects in the image with a good one under a special way of lighting, so image sharpening, local adaptive threshold, contour extraction is used for pre-processing. Finally, k-neighborhood chain code for inflection point detection algorithm is proposed to count the number of them in the image to realize defect detection.Finally, an actual test for real machine is completed in the field of production. The system developed in this paper can make accurate identification of the capsules’ defect. It can basically meet the needs of the enterprise, has a good market value.