Research On Key Technologies For Data Collection And Analysis Of Agricultural Environment

In twenty-first Century,faced with the large population growth and the shrinking of cultivated land,more and more evidences showed that The adverse effects of agricultural input on environment and health need new agricultural approaches to increase production and efficiency of existing farmland.Agricultural research in centuries has greatly improved our understanding of the agricultural system.There are many factors that affect agricultural production,but we still know little about it.Farmers tend to build on key crop management decisions on the basis of personal experience and intuition,because they lack quantitative scientific evidence on how these decisions affect production.In order to meet the growing demand for environmental protection of agricultural products,it is of great importance to optimize crop management decisions.It is an excellent way to provide farmers with data driven evidence to help them make the best decision.Therefore,it is particularly important to collect and analyze agricultural environmental data more accurately and accurately.Using WSNs(Wireless Sensor Networks)and Internet to realize the unification addressing,collection of big data era need huge amounts of information,and establish a wider range of universality to reduce differences design complexity is the Io T(Internet of Things)development of new ideas in recent years.This paper proposes and implements a fusion based on IPv6 gateway protocol of wireless sensor network system,Zig Bee,Wi Fi,bluetooth sensor can communicate with each other under the IPv6 gateway,to meet the wireless sensor network in a variety of needs of different under different application environment,compared to a single communication protocol of WSNs more flexibility and universality.First,we design a gateway architecture based on IPv6,implement different wireless devices,all adopt IPv6 protocol to the format of the communication layer is unified,and Internet address.Second,to achieve the Zig Bee,bluetooth,Wi Fi devices IPv6 transplantation,relative to the adoption of IPv6 wireless sensor,not avoid the gateway for complex application layer protocol conversion,reduce the gateway design complexity and power consumption.We also study the location of sensor nodes in a hybrid wireless sensor network,which are located in the underground(sensor nodes)and on the ground(satellite nodes).We consider two kinds of ranging(received signal strength and arrival time),never transmitted signals between adjacent nodes and sensor nodes and satellites.The problem of joint distribution parameter estimation of received signal strength and arrival time of received signal is proposed.First,we reach the power fading model of the various communication scenes in our network to simulate the propagation distance of the received signal intensity,and therefore,participate in the location coordinates of the nodes.We explained the degradation of all kinds of signals.The effects of fading,reflection,transmission and interference between two signals arriving along different paths.Under the same goal,we derive the statistical model of arrival time based on the parameters of sensor node coordinates.Based on strict statistical analysis,the probability distribution of signal arrival time is derived.Then,the maximum likelihood optimization problem is constructed by using the derived statistical model to estimate the location coordinates of nodes.Results the Scipy optimization package was verified by Python in this sensor location method.We also analyzed sensitivity of soil permittivity and permeability.Finally,we focus on the historical data of the production records of large agricultural products,and the analysis and development tools to predict the relationship between the output of the target data set and the density of insect pests and other factors.In order to predict yield and pest density,we compared three modeling methods: linear regression,random forest and support vector machine.The prediction accuracy of these three methods is the highest using support vector machine,which is used to predict yield and pest density.Prediction of yield is much more successful than prediction of pest density(62.6% average absolute percentage error)(8.2% average absolute error percentage).Our results emphasize that the useful value can be extracted from the historical record of agricultural production,and the machine learning method can provide more accurate and persuasive decisions than the traditional modeling methods.