Semantic segmentation of landcover for cropland mapping and area estimation using Machine Learning techniques

ABSTRACT The paper has focussed on the global landcover for identification of cropland areas. Population growth and rapid industrialization are somehow disturbing agricultural lands eventually food production needed human survival. Appropriate land monitoring requires proper management resources. proposed a method mapping by semantic segmentation to identify boundaries estimate areas using machine learning techniques. process initially applied various filters features responsible detecting through edge detection process. images masked or annotated produce ground truth label croplands, rivers, buildings, backgrounds. selected transferred model methodology Random Forest, which compared two other techniques, Support Vector Machine Multilayer perceptron, Our dataset is composed satellite collected from QGIS application. derived conclusion that forest given best result segmenting image into different regions with 99% training accuracy 90% test accuracy. results cross-validated computing Mean IoU kappa coefficient shows 93% 69% score value respectively found maximum among all. also calculated area covered under segmented regions. Overall, Forest produced promising mapping.