1. Stereo Camera-based Free Space Estimation for Docking in Urban Waters
- Author
-
Trym A. Nygård, Nicholas Dalhaug, Rudolf Mester, Edmund Brekke, and Annette Stahl
- Subjects
maritime autonomy ,free space ,safe navigation ,stereo camera ,docking ,Electronic computers. Computer science ,QA75.5-76.95 - Abstract
Operating in urban waters with an autonomous vessel can be challenging. The autonomous vessel must be able to react quickly and detect obstacles to avoid collisions and risky maneuvers. Exteroceptive sensors such as LiDAR and RADAR have typically been used with great success in the maritime domain, but the measurements are often too sparse to represent smaller obstacles during docking and other maritime operations. However, other sensor modalities, such as stereo cameras, can provide both appearance and dense depth information. In this paper, we present a stereo camera-based free space estimation method for the maritime domain. The mapping of navigable areas is crucial for path planning and collision avoidance systems. To robustly estimate the free space, we use vertically oriented rectangular segments known as stixels. We utilized both stereo correspondences and a recent image segmentation network trained on a large, generalized dataset to create the stixels. To validate our approach, we analyzed the estimated free space, evaluating both the accuracy and consistency in the estimated depth over time. We demonstrate the approach using a real dataset recorded with a stereo camera mounted on an autonomous ferry and compare its accuracy against measurements from a LiDAR.
- Published
- 2024
- Full Text
- View/download PDF