A Novel ORB Feature-Based Real-Time Panoramic Video Stitching Algorithm for Robotic Embedded Devices

Real-time video stitching is a critical technology in panoramic imaging for robotics. Traditional stitching techniques suffer from low computational efficiency, ghosting and artifacts, which limit their application in real-time video processing for robot vision. To address these challenges, we propose a real-time video stitching method based on Oriented FAST and Rotated BRIEF (ORB) feature detection combined with secondary fusion. This approach leverages ORB feature detection and Hamming distance-based matching, followed by homography matrix estimation to establish correspondences between keypoints. Furthermore, a secondary fusion technique is applied to reduce ghosting and visible seams, leading to a more natural video output. To meet real-time requirements, we incorporate resolution down-sampling and frame-skipping strategy. Compared to traditional methods, our approach achieves superior stitching quality while maintaining real-time performance in experimental results, making it suitable for resource-constrained robotic systems.