Research And Implementation Of Capsule Defects Detection System Based On Machine Vision

Drugs and health products produced in the form of capsules play an important role in people's lives,and the quality of capsule is closely related to people's health.The traditional method of capsule detection is manual sorting,which has these problems of uncontrollable detection precision,large consumption of human resources and low detection efficiency.Automatic detection system of capsule defects based on machine vision can realize high precision and efficiency industrial detection.According to the actual production requirement,a detection system of capsule defects based on machine vision is designed in this paper,and the main research contents are as follows:1)A detection system of capsule defects has been designed.Firstly,the structure of the system can be designed,which includes the design of transmission device and data acquisition device.The acquisition device is analyzed in detail,including the selection of light source,camera,photoelectric switch,acquisition card and the design of image acquisition scheme;secondly,the software of the system is designed,a human-computer interaction software system is developed,which includes the design of training and recognition module,the design of interface and multithread processing mechanism.2)The capsule segmentation is realized.Firstly,most capsules produced by the manufacturer are transparent and oval,both ends of the capsule are dark and the middle is bright under illumination,and the distinction between transparent capsule and conveyor belt is small;secondly,the slot size can't be adjusted with the fixed transfer mode,in this paper,capsule images are acquired in unfixed transmission mode,now the traditional segmentation method can't extract the capsule accurately,then a segmentation method of image projection and constraint gradient search is proposed,which includes image projection,decomposition and reconstruction of projective signal,constrained gradient search and capsule extraction,this method realizes the accurate extraction of transparent and opaque capsules.3)Capsule shape and color detection algorithm is designed.Firstly,capsule shape defect includes large capsule,small capsule and irregular shape,some irregular shape capsules are not easy to distinguish,in this paper the contour area,length,width,ratio of length and width,rectangular degree and the two order invariant feature are extracted for training and recognition;secondly,for these transparent capsules,these distributions of color features in different color space are analyzed in detail,and the method of variance threshold color feature extraction is proposed,which can effectively improve the reliability of the color feature,finally the shape and color of capsules are detected accurately.4)The problem of black spots and bubbles detection algorithms of transparent capsules is solved.Firstly,some black spots are small,and they may be not detected or the detection result is error.In this paper a method of edge detection of S space image and partition threshold determination is proposed to solve this problem;secondly,the edge of the bubble is the important feature of the bubble capsule,a method of contour analysis and least squares circle fitting is designed,which includes median fitting,threshold segmentation,contour extraction,contour selection and least squares circle fitting,and finally these edges of the bubble are accurately detected.