Research On Capsule Defect Detection And Recognition System Based On Machine Vision

In recent years,capsule vision detection and recognition technology based on machine vision has become increasingly mature.The corresponding products have appeared one after another at home and abroad,and the machine detection has been gradually replaced by manual detection,which has improved production efficiency.Although the foreign capsule detection and identification equipment has a good detection effect and recognition rate,the price is high.Therefore,research and design of a more cost-effective capsule defect detection and recognition system has practical significance,and the research of high-performance capsule defect detection and recognition system has become a key content of smart manufacturing.In this paper,a high-performance machine vision-based capsule defect detection and recognition system is designed.Image processing technology and pattern recognition technology are used to realize capsule defect recognition and classification.Firstly,the overall structure of the capsule defect detection platform based on machine vision is introduced,including hardware system design and software system design.The hardware platform is built including industrial cameras,LED light sources,industrial computers,microcontrollers,etc.,and various devices are described.Based on the OpenCV open source image vision library,the software platform designs and improves the image processing algorithm to form a capsule defect detection and recognition system including image processing and machine learning.Secondly,the area segmentation of the capsule body image and the edge detection algorithm of the defect area are studied.For the region segmentation of the capsule body image,the tilt angle of the capsule is first obtained by the minimum circumscribing rectangle method to calculate the rotation angle,and then the tilt correction of the capsule is realized by affine transformation,and then the capsule is divided into capsule caps by the Scharr operator.The capsule body and the intermediate bonding portion.For the edge detection of defect areas,the improved Canny edge detection algorithm is applied to the capsule defect detection and recognition system for the first time.Compared with the traditional Canny edge detection algorithm,the edge information detected by the algorithm is clearer and more complete,which is the edge of the capsule defect area.The accurate extraction provides a guarantee.Finally,for the common feature types and extraction methods of capsule defects,a set of vectors with typical feature values is proposed as the input of the defect recognition classifier.Based on the standard BP algorithm,the improved algorithm of introducing momentum factor and the improved algorithm based on Levenberg-Marquardt are analyzed and compared.Finally,a BP neural network algorithm based on Levenberg-Marquardt improvement is adopted.The classifier was designed using an improved algorithm,and the designed classifier was trained and tested using a capsule defect sample set.The experimental results show that the method is reasonable and feasible,and the convergence speed is fast,the prediction accuracy is high,and the test accuracy is also high.Therefore,the capsule defect detection and recognition technology studied in this paper is feasible,effective and highly practical.