SOIL CLASSIFICATION FOR AGRICULTURE CROPS USING K-NEAREST NEIGHBORS (KNN)

Abstract

Soil is the basis of our earth’s agroecosystems which provide us with fiber, food, fuel etc. Soil classification helps predict soil type and performance for growing agricultural crops that provide us with food. Soil classification is essential for a farmer who can know soil type, and plants the suitable crops depending soil type. The aim of this research is therefore to develop a method that automates soil classification by applying image processing techniques. In the proposed soil classification method, soil classification is performed by using color and texture of a soil image as features and by using the K-Nearest Neighbors (KNN) as a classifier.The proposed soil classification method firstly extracts color features: mean and standard deviation, and texture features: energy and contrast, from soil images in dataset. These features vectors are then saved as a features dataset. In testing phase, the texture and color features from the user input soil image are also extracted as a testing feature vector. The user input soil image is then classified based on this testing feature vector by comparing with all the features vectors in the features dataset using k-nearest neighbors (KNN) classifier. After classifying the user input soil image whether it is clay or clay loam or sandy loam, the system provides the list of crops and vegetations which can easily be grown in the predicted soil type.Soil RGB images dataset applied to our soil classification system contains “sandy loam” and “clay loam” (Red Earths and Yellow Earths) soil images has taken in plantations and farms in Lashio township and collected from Internet. Our own soil image dataset including 200 soil images is applied to the system with the purpose of building the features dataset and testing the system. 150 soil images in the dataset are used for building the features dataset and 50 soil images are, for testing the system, as unknown data. The overall accuracy of the system is over 88% for all 3 soil types: clay, clay loam and sandy loam. The system is implemented in MATLAB programming environment on Microsoft Windows platform.