Research And Implementation Of Intelligent Identification And Monitoring System For Alien Invasive Species

The continuous reproduction and spread of alien invasive species seriously threatens China’s agroforestry ecology and aquatic system,and has great harm to human and animal health,agricultural production and economic development.There are many kinds of alien invasive species,and there are inter-class homogeneity and intra-class heterogeneity,which brings challenges to the identification and prevention of alien invasion technicians.Therefore,the intelligent identification of alien invasive species is of great significance for early prediction and early warning and later prevention and control.However,the manual field investigation workload is large,the probability of data error is high,and the existing foreign invasion biometric software is difficult to complete the identification task under the condition of field network bandwidth limitation.In view of the above problems,this paper establishes an intelligent recognition model of alien invasive species based on deep learning,and develops the server,Android client and web client of the intelligent identification and monitoring system of alien ingress species.The main research contents and results are as follows:(1)Optimization of image recognition algorithm of alien invasive species.In order to solve the problem of low recognition accuracy caused by the wide variety of alien invasive species,intra-class heterogeneity and inter-class homogeneity,this paper proposes an improved method for image recognition of alien invasive species based on deep learning.Based on the MobileNet network,the attention mechanism based on channel domain and the deep connection attention network are introduced,and through the ablation experimental results,it can be seen that the accuracy of the model in adding the attention mechanism of the channel domain and the deeply connected attention network is improved by 3.8 and 5.9 percentage points,respectively.(2)Research on lightweight alien invasive species identification models.Aiming at the demand for offline recognition even in the field under unstable network conditions,this paper proposes a lightweight improvement method for the model.In this paper,the model optimized by the algorithm is BN channel pruning and knowledge distillation,so that the model can not only achieve lightweight,but also make up for the reduction of accuracy caused by channel pruning.Through the experimental results,it can be seen that the recognition accuracy of the model is improved by 6.0 percentage points compared with the original model,and the number of parameters is reduced by about 50%.The alien invasive species model MobileNet-SpeciesLW was established,and the test results showed that the MobileNet-SpeciesLW model obtained the highest average accuracy,average recall and F1 values for the identification of 120 alien invasive species,reaching 85.8%,84.6%and 85.2%,respectively.And the number of parameters is only 2.2M.(3)Design and implementation of client and server of intelligent identification and monitoring system for alien invasive species.In view of the needs of intelligent and lightweight prevention and control of alien invasive species,this paper builds the client and server of the intelligent identification and monitoring system of alien invasive species.Use Vue framework to implement the web client.The Android client is implemented in the Android Studio environment.Build an application server through Tomcat to achieve request interaction with clients.The development framework Django is used to establish an image server to complete the recognition of deep learning models and the return of recognition results.And use the MySQL relational database to store all business data.(4)Testing of intelligent identification and monitoring system for alien invasive species.In order to detect the operation of the intelligent identification and monitoring system of alien invasive species,the intelligent identification and monitoring system of alien invasive species was tested by functional integrity test,interface response timeconsuming test of key functions of the system,and compatibility test.The test results show that in the functional integrity test,all functions of the system meet the test cases.In the time-consuming response test of key interface of the system,the response time of the key functions of the system meets the design requirements.In compatibility testing,the system can be compatible with mainstream browsers and mainstream Android phones,meeting design expectations.Based on the test results,the intelligent identification and monitoring system of alien invasive species can realize the identification,investigation,identification and monitoring functions of alien invasive species in a variety of browsers and Android mobile phones.