Research On Intelligent Monitoring System And Model For Orchard Environment

Environmental monitoring of orchard and growers activities are closely related toagriculture production. For environmental monitoring of orchard characteristics, themonitoring sensor network of orchard environment was constructed. This paper designs aorchard-based agricultural IoT for environmental monitoring system in the orchard, thesystem consists of network camera for HD intelligent infrared, the Soil Mote, theEnvironmental Mote, the Gateway Mote and control centers, able to monitor the orchardregion’s information on-line monitoring and analysis, and the system has its own dynamic,low power consumption, short delay group, large capacity, high reliability characteristics.Then this article presents a Back Propagation Neural Network (BPNN) approach forpredicting solar radiation by climatological variables in Wanan. Meteorological andgeographical data (mean of daily temperature, mean of daily high temperature, mean of dailylow temperature, mean of daily relative humidity, mean of daily wind speed, mean of dailybarometric, mean of daily UV index, mean of daily leaf temperature, mean of daily leaf wet)is used as input to the network. Mean of daily solar radiation is the output. The BPNN trainsand tests data with multi layer perceptron (MLP) approach which has the lowest meanabsolute percentage error (MAPE).The trained model prediction was in good agreement withthe actual solar radiation figures, hence, producing an R2value of0.99531.The mean absolutepercentage error for the test was5.9674%. The main works in this paper as follows:（1）Aiming at the request of intelligent monitoring for the orchard environment, aorchard-based agricultural IoT for environmental monitoring system was designed in theorchard, it consists of network camera for HD intelligent infrared, the Soil Mote, theEnvironmental Mote, the Gateway Mote and remote control centers, able to monitor the fruitfarm region’s soil moisture/temperature, leaf wetness, air moisture/temperature, windspeed/direction, solar radiation, UV radiation, rainfall and barometric of the on-linemonitoring and analysis, transmit to the control center. The control center processes the dataand displays them using a variety of graphic formats. In a2-year stable trial, this remotemonitoring system shows simple, convenient and intuitive, flexible configuration, low powerconsumption, large network capacity features.（2）Orchard security, tree height, tree stem tiller numbers, fruit nutrition and the pestswas dynamic real-time monitored,we introduced network camera for HD intelligent infrared.It equipped with stepper motor driver of stable operation, accurate positioning precision yutai,and It transferred video information and control information in the system, and also equippedwith built-in zoom lens with high sensitivity, high resolution digital integrated color movement, realizing the functions of the picture clear positioning, the history of video query,all-weather multi range real-time monitoring.（3）It is very important for the Soil Mote and the Environmental Mote and GatewayMote in sensor network monitoring system. This paper designs the hardware circuit Mote andthe Mote of software program, then described all kinds of Motes’ characteristics, thearchitectures the details. The Soil Mote equipped with4soil moisture sensors deployed at thedepths of10,20,40,60cm, and4temperature sensors at the depths of10,20,40,60cm.（4）It presents a BPNN approach for predicting solar radiation and temperature and UVindex by climatological variables in Wanan. The BP algorithm uses the supervised trainingtechnique and provides a generalized description of how the learning process is performed.The training process was performed using the simulator. After several adjustments to thenetwork parameters, the network converged to a threshold of0.001.The trained modelprediction was in good agreement with the actual solar radiation figures, hence, producing anR2value of0.99531.The mean absolute percentage error for the test was5.9674%.Thetemperature model prediction producing an R2value of0.92299and MAPE was5.1835%.TheUV index model prediction producing an R2value of0.99452and MAPE was2.4593%.