Mining images of high spatial resolution in agricultural environments

Satellites are widely used for remote sensing applications. High resolution images are used for different geographical applications. Using geographical objects or spatial objects for analysis became prevalent in the contemporary era. Many supervised classification techniques came into existence to have efficient classification of high-resolution imagery. There are many factors that may affect classification of geographic images. They include the presence of mixed objects, feature selection, size of training set and segmentation scale. When these factors are considered for a systematic mining of images with high resolution, it results in improved performance. Especially in agricultural environments, it is essential to have such study to ascertain which supervised learning mechanism can best deal with the factors aforementioned. An algorithm named Feature Subset Selection (FSS) is defined to enhance classification accuracy. Different classification techniques such as Support Vector Machine (SVM), Random Forest (RF), Naive Bayes, k-Nearest Neighbour (KNN), Adaboost.M1 and Decision Table (DT) are used for the empirical study with spatial data mining. Useful analysis of the techniques is made and thus this paper provides valuable insights on mining images of high spatial resolution in agricultural environments.