This research integrates Large Language Models (LLMs) with drone surveillance systems to create an intelligent monitoring framework that actively detects anomalies in novel environments.

• Leverages LLMs to process visual data and make real-time decisions during drone surveillance
• Implements a dialogue-based approach for active perception, allowing the system to seek additional information when needed
• Demonstrates enhanced anomaly detection capabilities for previously unseen scenarios
• Creates a more autonomous and intelligent aerial monitoring solution

For aerospace applications, this framework significantly expands drone capabilities beyond simple visual inspection, enabling more sophisticated threat assessment and reducing the need for human supervision in complex surveillance scenarios.

Large Models in Dialogue for Active Perception and Anomaly Detection