Work

ステレオビジョンによる道路領域の推定

Year: 2017
Type: dissertation
Abstract: Navigation of autonomous vehicles is a very difficult challenge as the environmental recognition from dynamic environments is inherently difficult.In addition, actual applications need to be executed with a limited computing capacity.In particular, road detection is an important issue in advanced driving support systems (ADAS) especially for urban roads.If the road is not wellorganized unlike a highway, it is necessary to consider various factors, such as pedestrians, parked vehicles and sidewalks for road detection.In this thesis, we propose a method for estimating a road region using stereo vision, which combines 2D image processing and 3D reconstruction from images.First, we calculate the superpixels for the input in-vehicle image.Then, we implement a robust stereo matching on the road plane with stereo images transformed based on such plane.Next, plane detection is applied within a limited range from the superpixels and the disparity map.Finally, we estimate the road region using these results.The final results using the proposed method were evaluated using the KITTI-ROAD data set, which is a road detection benchmark for urban environments.
Language: Japanese
Open Access status: green
OpenAlex ID: W7146383279
Cites: 36
Cited by:
Related to: 10