Research And Implementation Of The Automatic Identification Of Alien Plant System Based On Deep Learning

China has always been a country that deeply harmed by invasive alien plants.Invasive alien plants continue to multiply and spread,seriously threatening our country's agro-forestry ecology and aquatic system,and have great harm to economic development.Therefore,the automatic classification and identification of invasive alien plants is of great significance for early prediction?warning and later eradication control.However,existing foreign invasive plant identification methods mainly rely on manual identification statistics,which are inefficient and cannot guarantee accuracy or real-time performance.In response to the above-mentioned problems,this paper established an automatic identification model of invasive alien plants based on deep learning,and developed an Android-based intelligent identification system APP for invasive alien plants.Mobile phone users use the system APP to take pictures of invasive plants and upload them to the server.The server calls the automatic recognition model of invasive plants to feed back the recognition results to the user in real time.At the same time,the user can view the detailed information of the invasive plants in the system.The main research contents and results are as follows:(1)Research on image recognition algorithm of invasive alien plants based on improved DBTNet.Under the MxNet depth learning framework,the external invasion plant identification model DBT-PlantID based on improved DBTNet is established,and compared to GoogleNet,DenseNet,ResNet,DBTNet or other models.The common 35 families and 135 species of invasive plants were selected as the research objects.The test results showed that the average accuracy and recall rates of the DBT-Plant ID model were the highest,93.8% and 94.2%,respectively.Indicate that the DBT-Plant ID model is suitable for the identification of invasive alien plants.(2)The design and implementation of Android client for automatic identification system of alien invasive plants.In the Android Studio development environment,set the overall UI framework of the app by linear layout,at the same time okgo network transmission framework is used to realize the connection between the client and the server.The client includes user login module,details query module,automatic identification module and remote expert identification module.Users can query and search the detailed information of invasive plants in the system,and take images of invasive plants with mobile phones for identification by the system;when users have doubts about the identification results,they can also request experts for remote identification to determine the identification results.(3)Cloud server construction and model deployment.Use Tencent cloud server and Jetty application server to build the cloud server side of the automatic identification system for invasive alien plants.The Jetty server is responsible for receiving and processing the request sent by the client,and realizes the connection between the client and the server.A server-side framework based on Spring Boot is built for server-side program development.Use Mongo DB non-relational database to store the details of invasive plants,identification results,identification records and other information.Finally,the system business logic code and recognition model are deployed on the Tencent Cloud server.