Multispectral Classification of Remote Sensing Imagery for Archaeological Land Use Analysis with Machine Learning Techniques

Human history can be traced through the impact of their actions upon the environment. The use of remote sensing technology offers the archeologist the opportunity to detect these impacts which are often invisible to the naked eye. The extraction and classification of regions within a multispectral remotely sensed scene from a particular geographical region allows the generation of training data that could be used to train a machine learning model that automatically could perform the identification of regions in new and larger scenes. This can be achieved using a developed multispectral image classification approach based on supervised machine learning models, which is referred to as the Archaeological Learning Classification (ALC) model. This paper presents some prospective results of the proposed methodology for supervised segmentation and classification of sensed archaeological sites for land use analysis. The results obtained with this study uses real multispectral scenes obtained with remote sensing techniques to probe the efficiency of the classification technique.