EcoBrownfieldNet: Brownfield Identification from Satellite Remote Sensing Imagery Towards Sustainable Development—An Artificial Intelligence-Based Framework

Finding suitable land for urban development is an important question towards sustainability. The shortage of such space adversely affects policymakers towards new planning. In this context, locating brownfields from urban areas can be a great solution. An underutilized or abandoned industrial or commercial site that would be difficult to expand or renovate is known as a brownfield. Redevelopment of brownfield sites typically involves turning them into homes, businesses, offices, recreational spaces, or public spaces. Traditionally, brownfield identification from satellite images is done using different remote sensing and GIS tools. Artificial intelligence (AI)-based brownfield identification is not very common as per the reported articles in literature. In this work, we explore the possibility of brownfield identification from satellite images by applying modern deep neural network algorithms. For experiments, we used more than 2000 satellite images from publicly available sources. We proposed a custom-built CNN model named EcoBrownfieldNet and compared its performance with state-of-the-art CNN models. The results show that the possibility of AI-based brownfield identification is very promising.