Classification Of Walnut Kernel,Shell And Distracter By High-Spectrum And Computer Vision

There are vast industrial parks of walnut in China.The reason why people are in favor of walnuts and its reprocessed products is that they are rich in nutrient value and pharmacological effects.As an ingredient of walnut,diaphragma juglandis is a kind of medical material with the efficies of both soothing the nerves and helping someone sleep,which its officinal value is in good abundance.So to speak,the separation of walnut shell,walnut kernel and diaphragma juglandis is an important step in the later processing of walnuts The wild walnuts from veteran trees in Lvliang,Shanxi,are taken as subjects in this research.The combination of hyperspectral imaging technology as well as computer vision technology is applied to the identification and automatic classification of walnut kernel,diaphragm juglandis and walnut shell.And eventually there is a display in the form of interface to achieve human-computer interaction.The main contents and conclusions of the study are as follows1.According to the spectral information obtained from hyperspectral imaging technology,researchers employ different pretreatment methods to build PLS models respectively,and compare the influence of different spectral preprocessing methods on the established models.The results show that the modeling effect of data preprocessed by SNV is optimal2.The full wavelength PLS and PCR models are established to classify the diaphragma juglandis,walnut shell and kernel after SNV pretreatment.Based on the PLS model,the accuracy rate of identification on diaphragma juglandis,walnut shell and kernel is 100%,98.75%and 99.44%respectively.And yet they take separately up 98.31%,98.89%and 97.23%on the basis of PCR model.By contrast,PLS model is superior to PCR model in the course of classification and identification3.The computer vision system is devised and set up for image acquisition.In accordance with the image characteristics of diaphragma juglandis,walnut shell and kernel,the grayscale pretreatment of computer vision images is conducted to reduce the influence of light and noise.Meanwhile,Convolutional neural network(CNN)is modeled by MATLAB R2016a software.The results show that the discrimination rate of diaphragma juglandis,walnut shell and kernel is 100%,90.62%and 97.50%separately,and the comprehensive discrimination rate is 96.73%4.In terms of the CNN neural network model established,the GUI interface is founded to identify the diaphragma juglandis,walnut shell and kernelThis study explored the feasibility of classification of diaphragma juglandis,walnut shell and kernel by means of hyperspectral imaging technology and computer vision technique,realized classification and identification and provided theoretical support for the research of online batching system in the subsequent walnut processing,which,at the same time,could furnish the future agricultural products sorting with design experience and practical significance on account of hyperspectral imaging technology and computer vision technique.