Perception task is a key technology for autonomous vehicles. Among the perception tasks, image perception has achieved powerful performance in contrast to perception tasks with other sensors. This is because of a lack of data. When comparing large amounts of image data, other sensors’ data is not sufficient, particularly in LiDAR data. To deal with the lack of data problem, many translation methods have been proposed. Although most of them focus only on image translation and LiDAR translation between simulated and real data, [1] proposed by us is the first LiDAR translation method that can convert sunny data to data of adverse weather conditions while maintaining 3D point cloud information. It is possible to generate fake LiDAR data for weather changes from sunny data. However, there is a big limitation in that noisy points are generated for a huge domain gap, such as the translation between sunny and heavily foggy conditions. Thus, we propose an advanced LiDAR translation for a huge domain gap to handle adverse weather change.
Advanced LiDAR Translation for Huge Domain Gap to Handle Adverse Weather Change
2023-06-01
2276057 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Survey on LiDAR Perception in Adverse Weather Conditions
IEEE | 2023
|Small ferryboats handle huge defense personnel
Engineering Index Backfile | 1942
Development of a self-driving car that can handle the adverse weather
Online Contents | 2017
|Validating Adverse Weather Influence on LiDAR with an Outdoor Rain Simulator
Springer Verlag | 2023
|Development of a self-driving car that can handle the adverse weather
Springer Verlag | 2017
|