Vision-Based Flying Obstacle Detection for Avoiding Midair Collisions: A Systematic Review

Area of Science:

• Robotics and Artificial Intelligence
• Computer Vision
• Aerospace Engineering

Background:

• Midair collisions pose a significant risk to aviation safety, particularly with the proliferation of unmanned aerial systems.
• Effective detection of flying obstacles is crucial for developing robust collision avoidance systems.
• Computer vision offers a promising sensor modality for real-time obstacle detection.

Purpose of the Study:

• To systematically review and analyze the existing literature on computer-vision-based flying obstacle detection.
• To identify trends, challenges, and research gaps in the field of midair collision avoidance for flying vehicles.
• To understand the factors contributing to the growth of research in this domain.

Main Methods:

• A systematic literature search was conducted across major scientific databases (Scopus, IEEE, ACM, MDPI, Web of Science) covering publications up to 2022.
• An initial pool of 647 publications was screened, with 85 selected for in-depth analysis.
• The review focused on articles related to computer-vision-based flying obstacle detection and midair collision avoidance.

Main Results:

• A notable increase in publications on flying obstacle detection and tracking using computer vision was observed.
• Hypothesized drivers for this increase include the availability of commercial drones and advancements in single-board computers and computer vision libraries.
• The majority of reviewed algorithms were evaluated in simulation environments, with only 26% reporting tests on physical flying vehicles.

Conclusions:

• Future research should prioritize enhancing the success rate of threat detection algorithms.
• There is a critical need for more testing of proposed solutions in complex, real-world scenarios using physical platforms.
• Addressing these gaps will be essential for the practical implementation of effective midair collision avoidance systems.