Detection, tracking, and re-identification (ReID) of objects in maritime environments in UAVs video stream presents significant challenges, particularly in search and rescue operations. In UAV based multi object tracking the ReID is hindered by the small object characteristics, sudden movements of the UAV’s gimbal and limited appearance diversity. To address this, we proposed a integrated method which includes detection and tracking of maritime object classes: boats, swimmers, and floaters—using the challenging SeaDronesSee dataset. Our approach leverages spatio-temporal features by re-engineering the YOLOv7 network with Video Swin Transformer model to capture 1) object-related spatial features and 2) to enhances detection by learning spatio-temporal dependencies. Central to our method is the Metadata-Assisted Re-ID (MARe-ID) for object tracking, which harnesses critical metadata from UAV, like GPS, UAV altitude, and camera orientations etc. to enhance tracking accuracy. Our experiments demonstrate the state-of-the-art performance of our method in maritime object detection, tracking and Re-ID, with significant improvements observed on the SeaDronesSee dataset.