Research On The Recognition Of Strawberry Diseases Based On Convolutional Neural Network

In recent years,strawberry industry in our country has developed more rapidly,the main producing areas have brought obvious economic and social benefits,become one of the main producing areas of farmers’ income and labor employment,but strawberry is prone to disease,need to identify accurately in the early stage of strawberry disease,so as to prevent and control.Therefore,it is the basic method to increase the total output of strawberry to monitor and control strawberry diseases reasonably and increase the yield of strawberry.The traditional manual disease detection method relies on the expertise of personnel.Meanwhile,with the rapid expansion of strawberry planting area,the traditional manual detection process is not efficient,poor real-time performance and other factors lead to errors.The method of machine learning should be gradually applied to the automatic detection and classification of strawberry diseases.Although it has achieved good results,there are still some challenges and deficiencies in practical application.For example,the shape and color of strawberry diseases sometimes change under the influence of growing environment,so researchers need to conduct adequate training and testing of strawberry diseases under different conditions.To improve the accuracy of identification.According to the above problems,this paper presents a neural network based on convolution DSDnet(Depthwise Separable Dilated.net)network model,and build the strawberry diseases data sets,precision identification of strawberry diseases.Finally,applying deep learning to agricultural informatization,a multi-terminal access recognition system for strawberry diseases was designed and implemented to achieve accurate recognition of strawberry diseases.The main work of this paper includes the following two aspects:(1)Based on the existing convolutional neural network model,network construction is carried out on the deep separable expansive convolutional module and SA attention module,and finally the strawberry disease recognition model of DSDnet is built.Then,performance test and comparison are conducted with large convolutional neural network,lightweight convolutional neural network,and introduction of different attention.Finally,it is proved that the strawberry disease recognition model based on DSDnet proposed in this paper has the highest recognition accuracy of strawberry disease,reaching 97.96%.(2)Design and implementation of strawberry disease recognition system based on DSDnet,and develop a strawberry disease recognition system based on DSDnet based on flask framework based on analysis of strawberry growers’ demand for strawberry disease recognition.Finally,system testing is conducted to verify the initial demand for strawberry disease recognition and verify the realization of the system.Realization of online strawberry disease recognition function.