A Review on Aerial Image Classification Techniques for Urban Planning

Abstract

Aerial image classification is important in urban planning by providing detailed insights into land usage, environmental impact, and land cover analysis. This review paper explores the different deep-learning models applied to aerial images, focusing on their performance, and architecture and its application. Several models, including convolutional neural networks (CNNs), relation-enhanced multiscale networks, and attention-based architectures, are evaluated using high-resolution datasets such as the ISPRS Potsdam and UC Merced. The study compares the accuracy, feature extraction techniques, and classification results across these models, comprehensively analyzing their applicability to urban land cover and land use mapping. Furthermore, this paper examines the challenges associated with aerial image classification, including handling occlusions, spatial resolutions, and model generalization. The findings highlight the effectiveness of deep learning techniques in addressing complex urban planning tasks and offer recommendations for future research. This review serves as a comprehensive guide for researchers and practitioners seeking to enhance aerial image classification methods in urban planning.