Optimizing Urban Planning with Satellite Imagery and Deep Learning-Based Object Detection

Abstract

This paper examines the use of satellite images and object detection deep learning algorithms to align urban planning to make it optimum. The study employs the Faster R-CNN deep learning algorithm by using high-resolution satellite images to identify and label the urban elements, including building, roads, vegetation, and bodies of water. Evaluation was based on precision, recall and Intersection over Union (IoU) scores with buildings and roads displaying high detection performance but vegetation and water bodies proved to be problems. The findings demonstrate that the model is useful in monitoring the trends of urban development, urban sprawling and land usage transformations, hence an important tool in sustainable urbanization. Additionally, the research highlights the significance of proper feature identification in order to improve urban strategic plans, especially in regards to environmental settlement and structure planning. The research ends with a reflection on the strengths and limitations of the model, together with directions in which it may be improved to achieve improved detection performance in cluttered urban scenes.

Keywords: Urban planning, Satellite imagery, Deep learning, Object detection, Faster R-CNN, Urban growth, Land use, Sustainable development, Remote sensing, Feature classification, Urban sprawl, Environmental management