Research On Pear Quality Detection And Grading System Based On Machine Vision

The cultivated area and yield of pear in China are the first in the world,and almost all of them are cultivated in rural areas of China.At present,pears are classified mainly by manual and mechanical methods in China.Manual grading is inefficient and easily affected by subjective factors;mechanical grading generally has a single function and is destructive to pears.Therefore,the rapid and efficient nondestructive testing of pears is of great significance to the revitalization of rural areas and the increase of farmers' income.In recent years,the rapid development of machine vision technology provides a new research direction for the rapid and efficient nondestructive testing of pear quality.In this paper,Yali pears and Huangguan pears are selected as the research objects.Firstly,three cameras are used to collect images of pears from the right,left and top of pears;Then,the pear image is processed by color space conversion,graying and denoising,and image segmentation to get the image which is convenient for feature extraction.The preprocessing operation includes color space conversion,graying and denoising;After feature extraction,pear diameter,color,number of defects and shape were calibrated by grading algorithm to determine the grade of thesis.In the above research,bilateral filtering is used in denoising;In the process of image segmentation,the weighted average method is used to segment the gray image;In the edge detection stage,Canny operator is used to detect the edge of the lightness component image to get the edge image,and finally the defect free part and defect part of the pear are obtained;In the feature extraction stage,the geometric feature,color feature and defect feature of pear were extracted respectively;In the grading stage,the images collected by single camera are classified into first-class pear,second-class pear,third-class pear and out of class pear according to fruit diameter,color,number of defects and fruit shape,and then the image grading results obtained by three cameras are comprehensively analyzed.The final grading result of pear is the worst grade of middle pear image in three directions.In the defect classification part,firstly,the defects are identified by the SVMDT defect classifier,and then classified according to the defect area standard.The pear quality detection and grading system shows users the characteristic information of pear image in three directions,the results of single direction grading and comprehensive grading.The user can select the variety of the current grading pear,and set the pear diameter,color,defect area standard,and pear shape standard.The accuracy of the system is about 85%,and the shortcomings still need to be improved.