IoT based monitoring system of soil properties

Abstract

Background: In Rwanda and all around the world agriculture plays an important role for living of human beings for survival and economic growth. Different research shows that as population increases the area of cultivation that is used for habitation and need of food is increasing while farming land is reducing.Smart farming is needed to maintain and secure food availability for this generation and future by using IoT. Objective: The overall objective of this thesis is to monitor soil properties by providing real time data of a given land to farmers and decision makers like RAB for proper use and productivity of land and predicting the appropriate croptype in that land measured.Monitoring soil properties helps to monitor degradation of land and know when and how much fertilizer needed in that land based to the croptype you want to plant meanwhile system will keep showing the croptype based to whatever fertilizers available. Methodology:In this thesis,the Descriptive quantitative research approach were used by establishing the relationship between independent variables(NPK,temperature,humidity and PH) and dependent variable(croptype like cereals,legumes,vegetables) where the croptype will depend to the soil properties available in the land.The study provides the way of knowing the current soil properties and appropriate croptype.With this system a farmer can monitor the land nutrients in the given land and can be able to add only needed fertilizers.Again in this system there is model to predict the croptype of a given land based to its soil properties.The data was collected using sensors NPK sensor,PH sensor,temperature sensor,humidity sensor.Data from sensors are sent to microcontroller through direct connection and then sent from microcontroller to the database hosted on the cloud via gprs module and also the realtime data visible on LCD. Data was analysed by using machine learning algorithm known as classification.Among three common types of classification logistic regression,decision tree classifier and knearest neighbor(KNN),we chosed to use decision tree classifier because it provided higher accuracy(99%) than others. Results:In this thesis,results show that the prototype build made up of sensors,microcontroller with internet enabled is able to collect data and send them to database hosted on cloud and those data are processed to make model then we are able to predict the best fit of croptype based to the soil properties measured.in general the expected output was achieved.The main contribution of this work is that, a farmer can know the status soil properties of land by using vi IoT sensors instead of tradition means of using laboratory which is expensive,time consuming and tedious,database hosted on the cloud helps the decision makers to visualize and monitor changing rate of soil properties of a given which can help to mitigate the land use in future and finally we can say using machine learning in prediction of best fitting croptype in any given land is a great contribution.