Research Of Capsule Appearance Detection System Based On Machine Vision

With the aging population and increasing levels of material,health products have become essential goods to improve the quality of life,especially the demand of capsule health products have been increasing.Manual sorting clearly cannot reach the requirements of industrial production and pharmaceutical standards,in this case a higher degree of flexibility and automated machine vision inspection mode can be more competitive and industrial prospects.The non-contact and non-destructive testing methods are adopted in the machine vision to ensure that the capsule is not contaminated by artificial detection,which improves the productivity and quality of the capsule.Based on the actual investigation of a capsule manufacturing workshop in Beijing,and a lot of references on the capsule detection of domestic and foreign literature,designed a capsule appearance specifications detection system based on Machine Vision by theories and practice.It has contained the several common appearance problems of capsules,such as capsules of varying lengths,stain,bubble,unique shape,color,etc.In this paper,the main research contents and innovation points include:(1)Paper has developed the software platform of system.The software system is mainly based on the VS2010 platform and MFC architecture,is a friendly interface,and simply processed images by the use of image processing algorithm toolkit OpenCV.The software system mainly includes two parts: training system and the identification system.(2)According to the requirement of the capsule inspection,hardware selection of the machine vision system has been analyzed in this paper.According to the related literature and machine vision system requirements of the system,it has determined the selection and position of camera,lens,image acquisition card,photoelectric switch.And it has designed processing algorithm of capsule detection training system,mainly including measurement threshold,feature extraction and export and other algorithms.(3)This paper acquires the identification standard of capsule size and color.The paper clusters and analyzes the characteristics of the capsule sample data by clustering algorithm,and obtains a measure of the size of the capsule size scale.(4)The paper designs capsule appearance defect detection algorithm.The type of capsule appearance detectionfor following four types:(1)size,(2)color,(3)bubble,(4)stain.Detecting capsule stains after the capsule full expansion,paper designs the detection of the spot by the light of a plurality of end faces of the camera.This paper presents StationaryWaveletTransformto detect mutation point of the image signal,and the position of the mutation pointdetermines the location of spot.According to the design of the bubble capsule B-spline least square circle fitting method based on the control points are selected to obtain the least squares fitting and round information.Finally,this paper summarizes content and results of the research,and indicates the shortcomings of the research work,as well as the research direction of the future work.