Research On Method Of Farmland Sensor Calibration And Sensory Data Cleaning

Agriculture is the pillar industry of the national economy and development,stability and development of agriculture is a prerequisite for social progress.Development and practical application of information technology in the agricultural sector is the fundamental way to protect the sustained and stable development of agriculture and to ensure long-term quality supply of agricultural products.Development of information agricultural technology is the foundation for the protection of national food security,break the shackles of environment and resources and the fundamental way to accelerate the modernization of agriculture.Our study of the creation of information agriculture technology system,development of agriculture sensor technology and agriculture IOA technology,realization of real-time monitoring and control for the crop and environment during agricultural production and management process have far-reaching significance on improving the degree of modernization of China's agriculture and developing precision agriculture.The current development of China's agriculture needs to face the dual pressures of resource shortages and environmental constraints,conflicts between poor management and high-input agricultural resources,therefore,how to achieve real-time monitoring of agricultural production and improve agricultural production and management level,have become an unavoidable problem.Internet of things in the agricultural sector could significantly smaller investment in human resources and avoid damage to farmland environment.Thereby obtain timely and accurate information farmland environment and crop growth information.It is in this case,the Internet of things in the agricultural sector has been more widely used.Wireless sensor network is an important agricultural Internet of things underlying network technology form,it consists of field information wireless sensor network nodes installed in the monitoring area of farmland.These field information wireless sensor network node connecting to each other in the field,which constitute multi-hop self-organizing transmission network system to collect field information.Field information wireless sensor network node is one of the most important components of agricultural Internet of things underlying network.the field information wireless sensor network nodes directly perceive a variety of physical information of the objects we monitored,signal acquisition and processing of data in the form of information conveyed to the user.Therefore,the measurement accuracy of field information wireless sensor network node has a direct impact on the accuracy of the monitoring data.Therefore,research on sensor calibration and data processing methods to improve data accuracy farmland perception is prerequisite to guarantee reliable implementation of agricultural production and management decisions.In this paper,farmland soil moisture sensor,C02 sensor farmland,crop sensors for the study,in order to improve accuracy of farmland sensory data,take research on method of farmland sensor calibration and sensory data cleaning,the main contents and results of the completion of the following.The hardware equipment of the node based on the farmland perception is introduced.The factors affecting the accuracy of the measurement of the farmland sensor are summarized.Illustrate the accuracy of farmland sensor for sensing the importance of data accuracy.The necessity of data correction for farmland sensor is described.Farmland soil moisture sensor,C02 sensor farmland,crop sensors for the study,applicate the least square method and BP Artificial Neural Network two correction methods to calibrate the three farmland sensor and calibration results were compared and analyzed.The results showed that the method of least squares fitting and BP artificial neural networks can both calibrate farmland sensor,but the linearity of the BP neural network fit is smaller than least squares fit,but its non-linear calibration effect is more significant.In accuracy,BP artificial neural network algorithm is superior to the least square algorithm.Since the BP artificial neural network algorithm is more complex than in the least-squares algorithm,so from the perspective of applications to consider,least-squares algorithm is more suitable for use in embedded systems,and BP artificial neural network algorithm is more suitable for server-side applications.On the base of preamble,considering the environmental characteristics of data collection and transmission of farmland IOA node,inevitably subjected to mechanical shock,external vibration,electromagnetic interference accidentally changed measurement conditions.Sensory data will be lost,such as property values,noise and outliers data issues.Conduct farmland sensory data cleaning based on K-means clustering algorithm and LOF outlier detection algorithms.And the results of farmland sensory data cleaning were compared and analyzed.Excluded the noise data and abnormal data in farmland sensory data and improved the accuracy of the farmland sensory data.Through the research on farmland sensor calibration and sensory data cleaning methods to improve the measurement accuracy of the sensor farmland and to ensure accuracy and stability of farmland sensory data.Making farmland field information sensory data collected by agriculture IOA more in line with actual demand of farming management measures.It will be positive for agricultural production decisions and help managers to make timely and effective management decision-making and response measures according to farmland information sensory data.Analyze and compare difference of algorithms used in farmland sensor calibration and sensory data cleaning.Made exploratory study for further deepen the research on farmland sensor calibration and sensory data cleaning.It ensures farmland information sensory data serve farmland management and decision better.It has practical significance in improving the level of agricultural production informatization,digitalization and precision.