Platform for Detecting Illegal Landfills Using Computer Vision and Satellite Imagery From Web Map Service

This study addresses the environmental problems of landfills through an innovative solution that combines artificial vision, satellite image analysis and agent-based architecture. The detection of these sources of contamination is particularly complex due to the visual heterogeneity of the objects present in the aerial views. To this end, a proprietary dataset has been developed aimed at identifying irregular landfills from aerial and satellite images. The research focuses on the application and comparison of the latest generation algorithms YOLOv8, YOLOv11 and YOLOv12, deployed in a multi-agent system developed with PANGEA, which allows a distributed and efficient management of the detection process. In addition, a dynamic platform has been implemented that allows the user to select geographical areas within Spain and verify the existence of landfills in these areas. In the event of detection, the system cross-references the results with official public information to determine whether the landfill is legal or illegal. Experiments show that YOLOv8 achieves outstanding performance with an accuracy of 97.25%, recall of 89.59%, F1-score of 93.27%, mAP@50 of 94.74% and mAP@50-95 of 79.06%, evidencing its potential for automated landfill detection in a real operating environment.