Optical Flow-Based Obstacle Detection for Mid-Air Collision Avoidance.

Area of Science:

• Aerospace Engineering
• Computer Vision
• Robotics

Background:

• Mid-air collisions pose a significant risk to aviation safety.
• Current collision avoidance systems, including manual tactics and automated technologies, have limitations such as cost and mandatory implementation.
• There is a need for low-cost, effective airborne obstacle detection solutions.

Purpose of the Study:

• To develop and evaluate an optical flow-based algorithm for detecting airborne obstacles to prevent mid-air collisions.
• To provide a cost-effective alternative to existing collision avoidance technologies.

Main Methods:

• Utilized a monocular camera for visual input.
• Employed optical flow vectors to distinguish object motion from camera motion.
• Integrated morphological filters, focus of expansion, and a data clustering algorithm for obstacle detection.
• Developed a simulator to generate realistic flight scenarios for algorithm evaluation.

Main Results:

• The optical flow-based algorithm successfully detected all incoming obstacles within their trajectories during experimental evaluations.
• Achieved an F-score exceeding 75%, demonstrating a robust balance between precision and recall.
• Validated the algorithm's effectiveness in diverse simulated environments with varying object trajectories and altitudes.

Conclusions:

• The developed optical flow-based algorithm presents a promising low-cost solution for airborne obstacle detection.
• This computer vision approach offers a viable method for enhancing aviation safety by mitigating mid-air collision risks.
• The algorithm's independence from extensive training data makes it a flexible and adaptable solution for future aviation safety systems.