Change Detection for Multispectral Remote Sensing Images Using Deep Learning

Abstract

Urbanization is a global process. Rapid industrialization and urbanization in developing nations represent a serious threat to the environment. The method of “change detection” examines how a region’s features have changed during a span of two or more times. Deep learning methods have outperformed more conventional change detection methods in tests. Artificial neural networks are the foundation of the specialized machine learning method known as deep learning. Comparing the deep learning model to other machine learning techniques reveals significant capability and versatility. Two visuals taken at various timestamps are given as input to a Siamese neural network for change detection. For each of the input images, higher-level feature vectors are produced using a sequence of convolutional layers. This method aids in the extraction of significant features that may be utilized to evaluate and quantify the differences between the images that were viewed. The Euclidean norm is then used to compare the feature vectors to determine the degree of change. Deep Learning Change Detection (DLCD) outperforms traditional change detection due to three factors: better information representation, enhanced change detection techniques, and performance improvements. To encourage better decision-making, it is essential to identify agricultural changes, which we can do with DLCD.