Vision-Based In-Flight Collision Avoidance Control Based on Background Subtraction Using Embedded System.

Area of Science:

• Robotics
• Computer Vision
• Aerospace Engineering

Background:

• Advancements in unmanned aerial vehicles (UAVs) and vision systems enable autonomous flight.
• Mid-air collision avoidance is critical for the mission readiness of autonomous UAVs.

Purpose of the Study:

• To propose and demonstrate a vision-based in-flight collision avoidance system for UAVs.
• To develop a robust system for detecting, tracking, and avoiding fast-moving objects using an embedded computing platform.

Main Methods:

• Dynamic background subtraction for moving object detection.
• Morphological operations, binarization, and Euclidean clustering for noise reduction and object segmentation.
• Kalman filtering for object tracking and stereo-camera-based distance estimation for 3D trajectory analysis.
• A novel motion-compensating background subtraction framework with 2D transformation approximation.

Main Results:

• Successful detection and tracking of moving objects.
• Implementation of decision-making techniques for evasive maneuvers.
• Demonstration on a test quadrotor UAV, providing insights into parameter tuning.

Conclusions:

• The proposed system offers a viable alternative to high-vision systems for collision avoidance.
• The developed framework enhances the safety and autonomy of UAV operations.
• The study validates the effectiveness of the vision-based approach for real-time collision avoidance.