Accurate detection of illegal dumping sites using high resolution aerial photography and Deep Learning

Urban waste impacts human and environmental health. Waste management has become one of the major challenges faced by local governing authorities. Illegal dumping has become an important problem in many cities around the world. Effective and fast detection of illegal dumping sites could be a useful tool for the local authorities to manage urban waste and keep their administrative zones clean. Remote sensing based on satellite imagery or aerial photography is a key technology for dumping management, aiming at locating illegal waste sites and monitoring the required actions after the detection.This study focuses on developing a method for detection and reporting illegal dumping sites from high-resolution airborne images based on deep learning (DL). Due to data unavailability for training a DL model, we use synthetic images. The trained model is evaluated based on a real-world dataset containing images from the city of Houston, USA. The results show that the proposed method solves the problem with high precision and constitutes a useful tool as part of a complete solution targeting dumping management by authorities.