Synthetic velocity measurement algorithm of monocular vision based on square-root cubature Kalman filter

Area of Science:

• Computer Vision
• Robotics
• Navigation Systems

Background:

• Vision-based velocity measurement is crucial for Unmanned Aerial Vehicle (UAV) navigation.
• Traditional Lucas-Kanade (LK) optical flow (OF) algorithms struggle with accuracy and adaptability under uneven illumination.
• Existing methods lack robustness in dynamic lighting environments.

Purpose of the Study:

• To develop a robust monocular vision integrated velocity measurement system for UAVs.
• To enhance the accuracy and adaptability of velocity estimation under challenging illumination conditions.
• To overcome the limitations of traditional optical flow algorithms in variable lighting.

Main Methods:

• Integration of the Lucas-Kanade (LK) optical flow algorithm with an optimized oriented FAST and rotated BRIEF (ORB) feature descriptor.
• Utilization of a square-root cubature Kalman filter (SRCKF) for sensor fusion.
• Processing visual information from a monocular camera for velocity computation.

Main Results:

• The proposed system effectively fuses LK OF and optimized ORB information.
• Improved velocity measurement accuracy was achieved in environments with variable illumination.
• The system demonstrated reduced sensitivity to illumination intensity variations compared to traditional LK OF.

Conclusions:

• The monocular vision integrated velocity measurement system based on SRCKF offers accurate velocity estimation for UAVs.
• The combined approach enhances robustness and adaptability in diverse lighting conditions.
• The method provides a reliable solution for UAV navigation in challenging visual environments.