Project image

Aerial Detection Prototype

A computer vision system that uses satellite and drone imagery to detect buildings, land usage, and illegal waste disposal sites.

The Challenge

Local municipalities face the difficult task of monitoring large territories for issues like illegal construction and unauthorized waste disposal. Traditional field inspections are slow, costly, and can't provide continuous oversight, creating a need for a more efficient, automated solution.

The Technical Solution

The prototype was developed as a computer vision system to automatically analyze satellite and aerial imagery, flagging areas of interest for municipal services.

  • High-Precision Object Detection: The core of the system is a YOLO model fine-tuned specifically on local aerial data. It was trained to accurately detect and map structured objects like buildings and classify different types of land use (e.g., cultivated vs. uncultivated agricultural land).
  • Advanced Pattern Recognition for Waste Sites: To identify unstructured and visually complex targets like illegal dumpsites, the prototype explored the capabilities of a large multimodal model, Google Gemini. This approach leverages advanced pattern recognition to classify potential waste sites, a task often challenging for standard object detectors.

Results and Impact

The project successfully demonstrated that an AI-powered system could significantly enhance spatial management for public sector entities.

  • Proactive "First Alert" System: The prototype enables a shift from reactive field visits to proactive monitoring, allowing officials to identify potential issues directly from aerial data.
  • Efficient Resource Allocation: By automatically highlighting suspicious locations, the system helps municipal inspectors prioritize their work and focus field visits on confirmed or high-probability cases.
  • Proven Model Reliability: During testing, the fine-tuned YOLO model demonstrated high precision in detecting buildings and classifying land, establishing a strong technical foundation for a full-scale deployment.

This project was developed as part of an EDIH Adria 'Test Before Invest' initiative to explore AI applications in public sector spatial management. For more details, you can read the official success story.