| Intelligent agriculture is a current research hotspot.With the popularization of Internet of Things,modern agriculture urgently needs to realize remote monitoring of the environment of crop growth and real-time assessment of crop growth status by means of advanced technology.At present,the technology of Agricultural Internet of Things is relatively mature,but there are still some difficulties in real-time evaluation of lodging status and water requirement of crop.Based on situations above,machine learning algorithm is introduced into rice growth status assessment to realize automatic irrigation and early warning system of rice lodging based on Internet of Things cloud service.The task is of great value to the development of modern agriculture and the improvement of crop management technology.The main work of this paper includes:1.Scheme design.In order to meet the needs of remote monitoring and assessment of crop growth status,a rice growth status monitoring system,with machine learning algorithm,based on Internet of Things was constructed.The scheme combines ZigBee with 4G to realize remote data communication,and stores and parses growth state data through cloud server.2.Water requirement prediction algorithm.XGBoost algorithm is improved in this paper,and the model is constructed using open source data sets.At the same time,the improved algorithm is compared with neural network model,Bayesian model,random forest model and Penman-Montes formula recommended by FAO.The results show that the improved XGBoost model improves the prediction accuracy greatly.3.Algorithm of rice lodging state early warning.In this paper,the optical flow algorithm,whitch is commonly used in dynamic target tracking,is applied to extract the lodging feature of rice straw,and the lodging situation will be warned by detecting the swing feature.In order to solve the problem that the frame rate of Farneback optical flow algorithm on PC platform is too low to satisfy the real-time performance,the algorithm is encapsulated as the IP core of FPGA to realize parallel processing,which makes the processing frame rate 6-10 times higher than that on PC platform.4.Design of embedded software and hardware.The central controller,the node controller board and the underlying software are developed.The centralized controller carries the optical flow algorithm and is responsible for forwarding the communication data between the server and the node.Node controller is responsible for collecting sensor data and controlling irrigation actuator.5.Server design.In order to meet the needs of users to view data of rice growing environment online and to control remote agricultural equipments online,a background management system oriented to Web application was developed,which can provide data,instructions and real-time images.6.System testing.The function of the system is tested,including sensor data acquisition,wireless communication,server,etc.The correctness of the system function is verified.The paper also makes a stress test for system server to verify its ability to handle concurrent requests.Experimental data show that the improved XGBoost algorithm proposed in this paper can improve the prediction accuracy of rice water requirement.At the same time,the early warning algorithm for lodging status of rice is of great value to the prevention and control of crop disasters.The test results show that the monitoring system of the Internet of Things(IOT)for rice growth state with intelligent algorithm runs well and can improve the automation level of rice production. |
