Leveraging Deep Learning and AI-Based Image Classification for Urban Planning in Detroit Amidst Projected Population Growth by U.N.

Abstract:

Over the past five decades, Detroit has witnessed a significant population decline, resulting in notable urban transformations such as the loss of residential structures and an increase in open or green spaces. Traditional satellite image classification methods have been used to monitor these changes but often fall short of accurately capturing the complexities of urban evolution. Recent projections by the United Nations suggest that, for the first time in 50 years, Detroit may experience net positive population growth over the next decade. This anticipated resurgence presents both challenges and opportunities for urban planning.

This study aims to harness deep learning and artificial intelligence (AI) within an ArcGIS environment to enhance image classification and object detection, achieving greater accuracy in mapping Detroit's changing landscape. I have utilized convolutional neural networks (CNNs) and advanced machine learning algorithms to process high-resolution satellite imagery. By integrating these AI techniques into ArcGIS, developed robust training data capable of detecting and classifying urban features such as buildings, vacant lots, and green spaces with improved precision.

The enhanced model effectively captures subtle changes in the urban environment, providing valuable insights for urban planners to accommodate the projected population growth. Accurate identification of available spaces and existing infrastructure can inform strategic decisions on housing development, infrastructure expansion, and the optimization of resources to meet future demands.

Keywords: urban planning, Detroit, population growth

Authors: