The Research And Realization Of Jujube Test Classification Technology Based On Machine Vision

In recent years,the jujube industry has developed rapidly because of the high nutritional value and good economic benefit of jujube.In some regions,the jujube industry has become a pillar industry in the rural economy.However,in the process of commercialization of jujube,most of the jujube products still use artificial grading or primary machinery grading,which results in uneven quality of the date,low price,low product added value,export volume is always small.In view of the above problems,this paper used the machine vision inspection grading technology to ensure the classification quality of jujube,which improve the market competitiveness and economic benefit of jujube,and has a strong social and economic significance.This paper taken Lingwu jujube as the research object,machine vision technology was used to study the phenotypic characteristics and surface defects of jujube.The main contents of this paper are as follows:(1)Samples of jujube were collected and their geometrical parameters were analyzed.The grading standard of Lingwu jujube was formulated based on GB/T 22345-2008 "Quality Grade of Fresh Jujube" and combined the characteristics of Lingwu jujube.(2)The overall scheme of jujube grading equipment based on machine vision was put forward.Through the research and analysis of each mechanism of the grading machine,a three-dimensional solid model of jujube nondestructive testing grading machine was established and the whole machine was optimized.The grading equipment mainly consists of four parts: monomer feeding and conveying system,image acquisition and detection system,air blowing grading system and transmission system.(3)The shape detection method of equivalent ellipse for jujube was proposed based on the physical characteristics of jujube.Using preprocessing methods such as graying and median filtering,edge detection,background segmentation,and finally equivalent ellipsoidization.The ellipsoid volume was calculated from the major and minor diameters of the equivalent ellipse.A linearregression equation was established according to the relationship between volume and mass to obtain the quality of the jujube.The result showed that the correlation coefficient between the measured jujube volume and quality was0.9328.(4)A method of jujube surface defect detection based on support vector machine was proposed.According to the characteristics that each component of the HSI color model is uncorrelated and easy to calculate,the mean and variance of hue(H)component in the HSI color model are selected as feature parameters.The best kernel function and regular constant were selected by orthogonal experiment,and a jujube surface defect detection model based on support vector machine was established.The test verified that the recognition rate of the test sample defect reached 94.6%.(5)With the modular design concept,the system functions were divided into modules,according to the actual production needs.The design of the control interface for jujube automatic detection classifier was performed through Halcon12.0 image processing software and VS2010 programming software.