Research And Implementation Of Forest Pests Intelligent Recognition System Based On Deep Learning

There are many kinds and wide distribution of forest pests,which cause huge economic losses to my country’s forestry every year.Therefore,timely and accurate monitoring of forest pests is the premise of rationally formulating forest pests control programs and reducing economic losses.Among them,accurate identification of forest pests is the primary task of forestry prevention and control.However,the existing forest pests identification methods mainly rely on the staff to identify by experience,which has low efficiency and poor objectivity.Aiming at the above problems,this paper establishes a forest pests identification model based on a Deep Bilinear Transformation Attention MechanismNetwork,and develops an intelligent identification system for forest pests based on We Chat mini program.The main research contents and results are as follows:(1)Research on image recognition model of forest pests based on Deep Learning.In order to realize the automatic and accurate identification of common forest pests,this paper proposes an image recognition method of forest pests based on Deep Learning.Based on the improved ResNet-101 network with residual module,this paper introduces a Deep Bilinear Transformation Module and an Attention Mechanism Module,and establishes a forest pests identification model DBTANet-101 based on a Deep Bilinear Transformation Attention MechanismNetwork.Compared with the six models of VGGNet-19,ResNet-50,ResNet-101,ResNet-50 with improved residual module,ResNet-101 with improved residual module and DBTNet-101,the test results show that the comprehensive performance of the DBTANet-101 model is the best,with an average recognition rate of 91.3% on the common 74 natural forest pest image datasets and 85.1% on the 22 similar forest pest datasets.It shows that the DBTANet-101 model has better identification effect on forest pests.(2)Design and implementation of the client of forest pests intelligent identification system based on We Chat mini program.In order to quickly and easily identify forest pests photographed by mobile phones,this paper develops a We Chat mini program for intelligent identification of forest pests.Use We Chat native UI development framework to design system client interface and functions,and use https protocol to realize front-end and back-end communication.The We Chat mini program client includes an intelligent identification module,a map query module,a knowledge map module and a user center module.The core functional module is the intelligent identification module of forest pests.User shoots or selects an image and uploads it to the server,and the server returns the result after identifying the image.(3)The design and implementation of the server side of the intelligent identification system for forest pests.In order to realize the specific functional requirements of the system,this paper builds the server of the intelligent identification system for forest pests.The server consists of four parts: proxy gateway,application server,image recognition server and application server database.Among them,the proxy gateway is implemented based on Nginx,which mainly realizes protocol conversion and active-standby switching;the application server is built based on the lightweight web framework Gin,which mainly realizes the interaction with the We Chat mini program;the image recognition server is based on the gRPC framework,which mainly realizes the recognition of images by calling specific algorithm models,and communicates with the application server through RPC;the application server database is established based on the relational database Mysql,which mainly realizes the storage of business data information and index optimization.(4)Testing and analysis of intelligent identification system for forest pests.In order to test the functional operation,runtime system resource consumption and request processing performance of the forest pests intelligent identification system,three tests are designed in this paper,the functional integrity test,runtime system resource consumption test and the interface response time cost test.In this paper,various functions of the intelligent identification system for forest pests are tested by simulating user operations;system resources is sampled and tested by using the glances tool;the response time cost of the system interface is tested by recording the breakpoint log of the system interface.After testing,all functions meet the design requirements,system runtime resource consumption meet the design expectations,and the total time required for a single image recognition request is about 3 seconds,which meets the actual needs.This system provides a fast and accurate intelligent identification tool for forest pests for forestry prevention and control.The user shoots or selects an image to upload through the We Chat mini program,the image recognition server recognizes the transmitted image,and finally returns the result to the We Chat mini program for displaying results.