Research On Capsule Defect Inspection And Sorting Technology Based On Machine Vision

With the development of science and technology, machine vision has not only been more mature and better used in certain areas of industrial production, but also played a significant role in the improvement of the production efficiency and product quality. At the same time, because the quality of medicines related to everyone’s health, the detection of drug quality monitoring requirements also gradually improved. At present, most of the major domestic pharmaceutical manufacturers are still using the artificial visual observation method to identify defective capsules. Manual inspection is not only inefficient, high cost and because of some of man’s own inherent factors, making the detection accuracy cannot be reliable guarantee. Based on the above backgrounds, this paper aims to build a set of capsule surface defect recognition machine vision systems to take place of the traditional manual inspection.In this paper, after a full analysis on the related technology research and its limitations home and abroad, we have conducted extensive research on the capsule digital image processing techniques and defect recognition technology.We put forward a whole set of the capsule surface defects recognition programs and methods. To detect the capsule stable, fast and accurate, we have specially designed capsule conveyor. We also designed a unique capsule recess such that capsule with the drive of the conveyor and flip approximately90°each time, and totally flip three times, we can get four different surfaces of the capsule images, with completely covering the capsule, to achieve complete detection effect.Based on machine vision, we have established a system on identifying capsule surface defects and eliminating defect capsules. We also designed illumination light source, and have enhanced defect characteristics needed in this paper. Besides, after electing the appropriate acquisition devices, capsule positioning detection device, we complete triggering and image acquisition of the capsule stalely and efficiently. Also, we design unique removing means with high pressure gas.Based on the minimum rectangular area, we designed a capsule image segmentation algorithm. After exploring the classic region growing segmentation, watershed segmentation, firstly, we proposed a segmentation method based on Hough transform, tilts capsules correction and horizontal projection, then according to the characteristics of the capsule image using enhanced central combination of region. Then using two smallest rectangular region contained manner, make the lower left corner as the fulcrum, with a certain step size within a certain range, we adjust the rectangular inclination, under the premise of fully contained. The rectangular area to the smallest area of the realization is the segmentation of the capsule image.We have written a human-computer interaction visualization software based on MFC. It contain capsule selection parameters import, BP neural network training, real-time monitoring and statistical of the detection results and other functions, user-friendly operation and management of the machine.Finally, the content and results of the research projects is summarized. It not only elaborates the characteristics of study and innovations, but also points out that the lack of work and exposes the directional proposals for future work.