INTERNATIONAL JOURNAL OF ADVANCE RESEARCH
IN MULTIDISCIPLINARY

Change detection techniques aim to identify differences in remote sensing images taken at different times over the same area, supporting applications such as urban development monitoring, land cover analysis, and environmental management. Traditional radiometric methods often suffer from high false alarm rates caused by factors like shadows, vegetation, or illumination differences. To overcome these limitations, this thesis explores the use of Digital Surface Models (DSM) in change detection. DSM offers structural height information, helping distinguish real changes from radiometric anomalies. However, DSM data also has limitations when height changes are absent despite actual land cover modifications. This study proposes a hybrid approach combining DSM with spectral and RGB data to leverage the strengths of each source. Deep learning techniques, particularly Convolutional Neural Networks (CNN) and Fully Convolutional Networks (FCN), are implemented to enhance change detection performance and automate feature learning. A supervised FCN model is trained using annotated datasets to distinguish between changed and unchanged regions accurately. The proposed method improves reliability and robustness in detecting real-world changes across urban environments, showing significant potential for enhancing remote sensing based urban monitoring systems.