The Research And Realization Of Potato Pest Identification System Based On Intelligent Terminal

With the launch of the Staple Food Strategy of Potato in China,the production and processing of potato will play an irreplaceable role in meeting the needs of large population,ensuring national food security,and alleviating the pressure of resources and environment.Since it grows in the complex natural background,potato is inevitable to be attacked by pests.The traditional way to identify potato pests is mainly performed by growers through eye observation based on the planting experience.This classification method not only consumes manpower and time,but also results in the deviation of identification effect is because of the subjective factors.With the development of precision agriculture technology,there is a growing demand for rapid prevention,diagnosis and treatment of potato pests,and it has become an urgent need to accurately and quickly identify pest species in the growing process of potatoes.Based on this,this paper selects six typical potato pests as research objects,and develops a pest identification system with deep learning algorithm embedded in mobile terminal.The main results and innovations of this paper are as follows:(1)The optimal convolutional neural network is determined through the establishment of a network based on the research of convolutional neural network and the experiment with different structures and parameters.Through the experimental comparison of the three data sets,it is found that the data set after image preprocessing and data enhancement has the best recognition effect in network training,and the training recognition rate reaches 92%.(2)Based on the study of transfer learning,the classification model is implemented by using the lightweight deep convolutional network,MobileNet.Different from conventional convolution that adopts deep convolution technology,MobileNet can effectively reduce the redundant expression of convolution kernel,the model proportion and the calculation amount of mobile terminal equipment.The identification rate of pest test samples is 96%,and the weight file size is 16 MB.(3)Through the research on the Android platform,the weight files trained by the data set are embedded into the mobile terminal for convenient invocation,and the potato pest intelligent identification system based on the mobile terminal is designed and implemented.Multi-thread programming technology is adopted to make thesystem run faster to quickly identify the results.In addition,the system interactive interface provides two functions of real-time shooting detection and album selection detection.The recognition rate of the software is above 89% and the recognition time is 0.3 seconds.The final experimental results show that the intelligent identification system based on mobile terminal designed in this paper has a good effect in potato pest identification.The system has high recognition accuracy,high recognition speed,strong operability and meets the design requirements.