Research On Detection And Recognition Of Stored Grain Pests Based On Cloud Platform

At present,the traditional technology of insect pest identification in our country has been exposed to many problems in the process of operation,which is time-consuming,high misjudgment,high cost and low efficiency.In the face of the expending scale of the granary,the traditional method cannot meet the needs of the pest detection and classification.We can identify the pest quickly and deal with these huge data through the distributed processing Cloud platform.The distributed processing cloud platform adopts distributed computing technology,which has multiple processors and storage systems,can be used to calculate the number of programs or concurrent programs with loosely coupled or centralized controlled.Distributed computing will divide the program into a number of parts to be computed in the same network environment,it can reduce the number of data leakage and the cost of computing,storage of massive data as well as on-demand access.A parallel BP neural network algorithm based on distributed cloud platform is used for image recognition of pests in this paper.The working principle of the image recognition system using globally asynchronous and locally synchronous,when the control layer receives the user's request,it calls the algorithm cluster of the transition layer,and the transition layer calls the optimal algorithm at each processing phase to improve the image processing speed according to the detected different pest characteristics.Then deploy the system to the cloud platform,the cloud platform receives user instructions will call the system and the image recognition operations,finally the identification results will be fed back to the user through the network.The performance of this system is compared with that of the traditional single machine identification system through experiments.The experiment shows that the system greatly improves the speed of pest detection and recognition.The main research work is as follows:(1)Firstly,it detailed analysis the current status of and advantages of the several insect pests' detection technology,selective analysis of the design principle,architecture of HDFS system?MapReduce and the processing of reading and writing documents.It introduces the cloud platform service flow and the overall design,using Hadoop to build a cloud platform,then detailed descript the Hadoop environment and the configuration of the cluster mode.(2)Secondly,the paper designs and analysis of insect pest image detection and recognition system based on cloud platform,make a concrete analysis on image preprocessing,segmentation and feature extraction method,each step uses several kinds of algorithms to realize and set up a parameter list at the same time,according to the global asynchronous and local synchronization method,the control layer calls the underlying algorithm module through the algorithm layer.According to the different characteristics of the image data,the cloud platform can call different algorithms to improve the processing efficiency and save the processing cost.(3)Thirdly,designs the classifier based on the parallel BP neural network,and detailed describes the processing of the constructing the network training model.Then take the sample training according to the experimental results,comparing the performance of the recognition system with the traditional system based on parallel neural network algorithm.(4)Finally,introduces cloud platform service flow and the overall design,built cloud platform by Hadoop,and configures the Hadoop pre environment and cluster model.Discusses the implementation of pest identification system based on cloud platform.