Sensor Fusion for IoT-based Intelligent Agriculture System

Sensors in agriculture are in use from weather monitoring to autonomous watering. Using low-cost sensors allows designers to create a prototype for a hardware environment to implement data acquisition and mining process. Thus, the relation between sensors can be understood and a test environment for sensor fusion can be created. In this paper, different input devices are synchronized by using a microcontroller system and each data, obtained from the sensors, are sent wirelessly by an (Internet of Things) IoT device to the cloud, by recording and monitoring from the graphical user interface on the web as a real-time environment to apply data mining algorithms thereafter. This study uses the regression trees to obtain the sensor data relations from 8 different data related to light, temperature, humidity, rain, soil moisture, atmospheric pressure, air quality, and dew point. Each sensor data has a different effect on the agricultural monitoring, however, reducing the number of sensors can reduce the cost of a system, by giving still accurate observations via sensor substitution proposed. Therefore, by using the regression trees, the classification of sensor data is inspected in this study. A test prototype of the hardware together with the software design is created for data monitoring and sensor fusion in different combinations. In the end, after fusion tests for all possible cases, outstanding results for each sensor substitution is presented. Temperature and dew point can be obtained using other sensors by fusing the train data on the regression tree by 92% and 84% accuracy respectively with a 5% numerical error margin in the leaf nodes on the regression tree.