Real-Time Video Stitching Technology for Ship Camera Arrays

Video stitching is a key technology in remote control and autonomous shipping. Aiming at the low efficiency of video stitching, the uncertain installation positions of different ship camera arrays, and the large parallax, a real-time video stitching technology for ship camera arrays is proposed. The stitching process is divided into three modules: stitching parameter initialization, online stitching, and recalibration. The initialization module obtains stitching parameters based on feature-matching algorithm to solve the problem of difficult calibration of ship camera arrays. The online stitching module stitches images based on the best stitching line and multi-band fusion to optimize the stitching seams. The recalibration module regularly corrects the stitching parameters to adapt to scene changes. Finally, the stitching algorithm is accelerated based on the multi-threaded characteristics of GPU. Experimental results show that the proposed method can maintain good stitching quality while reducing the time required for online stitching of 6-channel 1080P video streams to 98 ms, meeting the needs of real-time video stitching. The research results can provide support for video stitching technology in remote control and autonomous shipping.