This research compares traditional change detection (CD) methods, including PCA and k-Means, with advanced approaches using fully convolutional neural networks (FCNNs) for detecting changes in urban and suburban satellite imagery. Aligned with SDG 11 (Sustainable Cities and Communities), it emphasizes the importance of robust CD algorithms for sustainable urbanization and resilient infrastructure. Inefficient detection methods complicate tracking urban expansion, deforestation, and environmental degradation, potentially delaying responses to disasters and undermining resource management. The study evaluates unsupervised models using PCA and k-Means alongside FresUNet-based architectures and Siamese networks tailored for urban planning. Traditional methods often neglect spatial-temporal dynamics critical for accurate detection. In contrast, the Siamese UNet model, incorporating attention mechanisms, excels in identifying subtle changes while minimizing noise, significantly enhancing detection accuracy and aiding disaster risk reduction. Performance evaluation spans diverse data sources, including UAVs, IoT devices, and large-scale Earth observation systems like Copernicus and Landsat. The goal is to identify the most effective algorithm for varied datasets. The Onera Satellite Change Detection (OSCD) dataset, featuring 24 pairs of multispectral Sentinel-2 images from 2015–2018 across Brazil, the U.S., Europe, the Middle East, and Asia, is used for training. This dataset includes 13-band images with pixel-level ground truth for urban changes, offering resolutions of 10 m, 20 m, and 60 m. By leveraging high-resolution data and advanced architectures, this research aims to address critical challenges in environmental monitoring and urban planning