Study On The Defect Detection And Recognition Of The Capsule Head

In the actual production process of a large number of capsule,traditional capsule defect detection is to use human eye visual detection methods.This method is prone to inaccurate subjective judgment,slow detection speed and two contact pollution,so it can't meet the requirements of modem industrial production.At present,some scholars have studied the defect detection of capsule by using image processing method,and some pharmaceutical factories have already begun to automate the capsule testing,but the detection of capsule head is not found.Aiming at the breakage problem of capsule head,the defect detection and recognition algorithm of capsule head is studied in this paper,and it mainly includes 2 parts of the capsule head positioning and identification.In the part of the localization of capsule head,the modified radial symmetry transform is applied to the localization of capsule head in this paper.The algorithm makes full use of the radial symmetry of the capsule head and realizes the effective localization of capsule head.It achieves a faster detection that can satisfy the real-time requirements.In the part of the recognition of capsule head,we use the method of image processing in this paper.Firstly,the smoothing denoising,image enhancement and image segmentation are used to the gray image of the capsule head,and the image and the edge of the image are obtained.And then the characteristics of capsule head images are extracted by analyzing the differences between the intact capsules head and the defective capsule head.Finally,the method of the support vector machine(SVM)is first applied to the recognition of capsule head in this paper.The effective classification of the capsule head is realized by adjusting the parameters of the kernel in SVM,and running in Matlab.Experimental results show that the algorithm can detect 43200 capsule heads per hour.In NEU Capsule Image Database Version 1.0,The recognition rate for the good capsule head and the bad capsule head are 100%.This fully verifies the validity of the algorithm in this paper.But for the algorithm in this paper,we can also find better features and simplify algorithmto improve the running speed.