Machine Learning-Based Multi-temporal Image Classification Using Object-Based Image Analysis and Supervised Classification

Abstract

During last decade, there has been tremendous research related to the image-based technique in remote sensing; object-based classification is one of the popular techniques due to its capacity of promising results. This paper presents a novel approach where a hybrid method of object-based image analysis and supervised classification is used. The data used in this study is high-resolution multispectral 4-band images from 2017 to 2019 provided by the PlanetScope satellite of region Chandigarh, India. First, the data has been pre-processed through passing it in a pipeline of steps followed by a multi-resolution segmentation algorithm and classifying the image into seven classes based on the spectral signature using algorithms like maximum likelihood (ML), support vector machine (SVM), Mahalanobis distance (MD). Comparing the three algorithms, it was observed that SVM and ML have given the highest overall accuracy of 95.21% and kappa coefficient = 0.9159. Also, the overall accuracy 91.91% and kappa coefficient = 0.8860 were achieved.