Knowledge-Based Airborne Hybrid System for Sparse Minefield Detection

Landmines come in a variety of shapes and sizes. They can be square, round, cylindrical, or bar-shaped. The casing can be metal, plastic, or wood. Traditional techniques to detect landmines are both dangerous and time consuming. Airborne-based landmine detection system often uses high resolution optical sensors for the target detection. They use various image processing techniques that fall in two broad categories – signature-based and anomaly detections. Since landmines appeared in airborne IR images often don’t show stable signatures, signature-based approach is less robust. Anomaly detection usually generates many false alarms, making the system practically less useful. To overcome these challenges, in Phase I, we propose to develop a suite of anomaly detectors and false alarm mitigators for the target detections. We combine these detections in a minefield composite image to form a sparse minefield. Degree of randomness algorithms will be developed to detect the minefield. To further improve the system performance, we propose to design a knowledge base using mixture ontology. This knowledge base is capable of reasoning and can automatically select parameters and their settings based on the operating conditions and target types. External data sources can also be incorporated into this knowledge base.