Recently, autonomous surface vehicles (ASVs) have attracted much research attention because of their potential effectiveness in carrying out various maritime missions such as surveillance and environmental monitoring. Situation awareness is a critical ability, and the camera is an essential sensor for ASVs in conducting such missions by automatically detecting and tracking objects in the surrounding environment. We detect and track the objects robustly by extracting the detection uncertainty and applying it to the tracking process. This study addresses automatic ship detection and tracking, considering the uncertainty of deep learning-based object detection using an onboard monocular camera. Gaussian YOLOv3, a deep learning-based object detection algorithm, is applied to detect a ship and estimate the uncertainty of the detection result. The detection results are sampled with uncertainty and propagated based on the camera geometry. Finally, the position of the target ship is estimated by a particle filter using the sampled detection results as measurements.