Improved Research And Application Of Crop Disease Recognition Based On Deep Neuro-Fuzzy Network

Nowadays,with the rapid development of information tech nology,global data is growing explosively.How to accurately identify effective information from massive and complex data is an important demand of data mining technology.Among them,image classification is one of the main research directions in data min ing.How to accurately classify images containing fuzzy and uncertain information is an urgent problem to be solved in the development of image recognition technology.In this paper,improved methods of deep neuro-fuzzy network are proposed to identify images containing fuzzy and uncertain information in natural scenes and few-shot task,and the actual effect of these improved methods is tested in crop disease recognition.The main research includes the following three aspects:Firstly,this paper proposed the deep neuro-fuzzy network to solve the problem of image classification with fuzzy and uncertain information.In this paper,fuzzy inference rules are integrated into convolutional layer and pooling layer of convolutional neural network to construct deep neuro-fuzzy network.The experimental results on three open beta datasets show that the precision,recall rate,F ₁-score and accuracy of the deep neuro-fuzzy network are more than 87%.The ablation experiment and comparison experiment were used to test th e effectiveness of deep neuro-fuzzy network in identifying tomato leaf diseases.The ablation experiment results showed that the deep neuro-fuzzy network introduced fuzzy inference rules was superior to the deep neural network.The comparison experiment re sults showed that recognition accuracy of the deep neuro-fuzzy network was 94.19%,2.74%higher than that of Inception V3.Secondly,a meta-baseline method based on deep neuro-fuzzy network is proposed to solve the classification problem of few-shot images containing fuzzy and uncertain information.The deep neuro-fuzzy network is taken as backbone of the meta-baseline to construct this meta-baseline.The effectiveness of the meta-baseline based on the deep neuro-fuzzy network in identifying few-shot crop leaf diseases containing fuzzy uncertain information was verified by parameter optimization,similarity measurement comparison,contrast experiment and generalization.The parameter optimization experiment showed that accuracy of this model was the highest when the learning rate was 0.0001 and the batch size was 128.The results of similarity measurement comparison show that accuracy of this model is the highest when cosine similarity measurement with learnable parameters is adopted.Comparative experimental results show that accuracy of the proposed method on 2-way 5-shot is 83.02%,which is better than this meta-baseline method based on Conv Net4 and Res Net12.In addition,the generalization experiment result show that the accuracy of the model reaches 80.21%in 2-way 5-shot.Finally,deep fuzzy convolutional neural networks are proposed to solve the problem of image classification in natural scenes with fuzzy and uncertain information.The fuzzy neural network replaces fully connected layer of VGG16 to const ruct the deep fuzzy convolutional neural network.Experimental results on three open beta datasets show that the recognition accuracy of this model reaches 87.60%,98.90%and 82.36%respectively.The effectiveness of the deep fuzzy convolutional neural net work in identifying soybean leaf disease was tested by comparison experiment.The results showed that the recognition accuracy of the model reached 98.58%under the shortest training time.It is 11.08%higher than Mobile Net-FNN,4.39%higher than Inception V3-FNN,5.55%higher than VGG19-FNN and 2.17%higher than Xception-FNN.In this paper,improved methods of deep neuro-fuzzy network are proposed to solve the classification problem of image containing fuzzy uncertain information,images in natural scenes and few-shot task,which provides a theoretical reference for the improvement of image classification algorithm.And these improved methods of deep neuro-fuzzy network are applied to crop leaf disease recognition,which is helpful to take timely treatment a nd control measures,promote the integration of artificial intelligence and agriculture and the development of agriculture with high quality.