Comparisons of Classification Models on COASTSAT

Abstract

Due to rapid climate change, the task to observe and quantify shoreline position is important to coastal protection and management. CoastSat is a time-series shoreline detection system that using artificial neural network ANN on image classification. This paper further compared 4 image classification models such as decision tree classifier DTC, non-linear support vector machine (SVM), k-nearest neighbors (KNN) and linear SVM with stochastic gradient descent (SGD) on the Qijn images from the satellite, Landsat 8. The experimental results demonstrate accuracies, F1 scores and precision-recall curves on different models. Finally, the work shows that ANN is the best model in CoastSat image classification.