Research On Grading Of Wheat Grain Quality And Thousand Grain Weight Based On Image Processing

Wheat is one of the staple foods in people’s life.From the beginning of breeding,to the estimation of yield after maturity,and then to the later processing,wheat appearance detection is indispensable.With the progress of science and technology,image processing technology is more and more applied to all kinds of wheat detection,artificial detection is gradually replaced by machine.Compared with manual detection,although the efficiency of wheat detection using image processing technology has been improved a lot,the recognition rate of wheat grading is still low.Therefore,in order to improve the recognition rate,this study takes hangmai 8805,Jinhe 727 and Henong 861 as the detection objects,proposes a segmentation method for the phenomenon of wheat adhesion,then extracts features,establishes a wheat classification model,and finally estimates the 1000 grain weight of wheat.The main work of this paper is as follows:(1)A segmentation method based on pulse coupled neural network(PCNN)and morphology is proposed.In order to reduce the subjective influence of the artificial placement of wheat grains on the later experiment,the wheat grains are randomly scattered on the flannelette,so the phenomenon of adhesion is inevitable.In order to improve the accuracy of wheat grain classification,it is necessary to remove the adhesion,that is,to segment the image.PCNN has the characteristics of pulse synchronization,and can reduce the difference between similar gray values,and can divide the regions with similar characteristics.Therefore,this paper proposes to use PCNN segmentation model to segment the conglutinated wheat.The segmented wheat image has some problems such as noise,edge discontinuity,holes and so on.The open close operation in morphology can just make up for these defects.Therefore,a segmentation model combining PCNN and morphology is established to segment wheat.Compared with watershed algorithm,it has a better effect.The segmentation accuracy of this method is 99.22%for hangmai 8805,99.07%for Jinhe 727 and 99.03%for Henong 861.(2)A wheat grading model based on support vector machine(SVM)optimized by grey wolf optimizer(GWO)was proposed.Firstly,21 characteristic parameters were extracted from wheat grains.Based on these features,SVM classification model is established.The classification accuracy of the model is 83.61%for hangmai 8805,83.61%for Jinhe 727 and 81.97%for Henong 861.From the experimental results,it can be seen that it is feasible to use the extracted wheat grain characteristics to classify the quality of wheat grain,but the classification accuracy is not high.To solve this problem,a wheat grading model based on GWO optimized SVM is proposed.This model adds GWO algorithm to the traditional SVM classification model.GWO algorithm has the advantages of fast convergence speed and strong search ability compared with other optimization algorithms.Firstly,the parameters(c,σ)in traditional SVM are optimized.According to the gray wolf predation process,the optimal parameters are found,and the wheat classification model is established with the optimal parameters.The experimental results show that the accuracy of wheat classification is significantly improved.The classification accuracy of gwo-svm model was 95.08%for hangmai 8805,93.44%for Jinhe 727 and Henong 861,which was better than the traditional SVM model.(3)In this paper,the linear regression model was used to establish a single linear regression equation for the extracted characteristic parameters of wheat grain area and wheat grain quality.The relationship between the area and quality of hangmai 8805,Jinhe 727 and Henong 861 was analyzed respectively.Then,the 1000 grain weight of hangmai 8805,Jinhe 727 and Henong 861 was estimated according to the relationship.Therefore,it is feasible to estimate 1000 grain weight by using wheat grain area parameters,which provides support for the estimation of wheat yield.