Flammulina Velutipes Classification And Recognition Based On Machine Vision

In the current market,the classification of flammulina velutiper is mainly by factory cultivation first and then by artificial quality grade.However,artificial classification is easy to be affected by employees` subjective emotions,production experience as well as other factors.In addition,employees will suffer visual fatigue after long time engaged in the artificial classification,which will influence the accuracy and efficiency,thus resulting in the economic benefits decline.In recent years,with the rapid development of machine vision,pattern recognition and other technologies,it is of great significance and wide application value to study the classification method of flammulina velutiper based on machine vision to improve its classification accuracy and efficiency.However,there are few researches on the classification method of mushroom based on machine vision at home and abroad.In light of the above problems,the main research contents of this paper therefore are as follows:1.To build up a standard MID.In this paper,a vision system is set up to collect the image of flammulina velutiper.In MID,the heads and roots of each mushroom were collected,with a total amount of 32827,of which the heads were 19595,and the roots 13232.Besides,this paper designed a standard test protocol for the classification task,which was used to comprehensively evaluate the accuracy performance of the flammulina velutiper classification.2.Head classification and root classification of mushroom are two key points according to the standard of mushroom classification.Therefore,we classified the heads and roots of the flammulina velutiper respectively through the study of the flammulina velutiper classification.(1)For root classification,this paper presented a migration learning network model based on VGG-16.Experimental results showed that this method can be effectively classified,and the accuracy was 84.1%.(2)While for head classification,this paper proposed a classification method based on convolution neural network and it studied the classification methods based on AlexNet model,VGG-16 model and Resnet-50 model,as well as the network structure and training method of above three models,which were applied to the mushroom classification under Caffe framework.Experiments indicated that the AlexNet model had achieved good results in MID data set,and the classification accuracy rate was 86.65%.To sum up,through theoretical analysis and practical operations,this paper has theorized the classification and identification of flammulina velutiper based on machine vision technology,and has achieved the expected goal of the research.