Fig 1.
Schematic of NOD-AVM (dark regions (① ~ ④): detection areas of four NOD-AVMs based on AVM cameras).
Fig 2.
Flowchart of NOD-AVM model for near obstacle detection and localization.
Fig 3.
Illustration of inverse perspective mapping (IPM) for off-ground and on-ground objects.
Fig 4.
Illustration of NOD-AVM formation process: (A) raw images; (B) undistorted images; (C) IPM images; (D) Projection of corresponding points onto ground plane; (E) final binary image(red polygon: detection area, red rectangle: detection obstacles).
Fig 5.
Coordinate system for obstacles state estimation via NOD-AVM: (A) ICS-ACS relationship; (B) ACS-VCS transformation.
Fig 6.
Calibration of the right wide-angle camera for IPM: (A) prototyped intelligent vehicle; (B) raw image; (C) IPM image.
Fig 7.
Results of obstacle detection on campus.
(A) undistorted images; (B) IPM images; (C) detection results.
Fig 8.
Detection result of a small obstacle (cone).
(A) undistorted images; (B) detection results.
Table 1.
Detection results of NOD-AVM on campus.
Table 2.
Performance comparison with other state-of-the-arts detection methods.
Fig 9.
Detection results using NOD-AVM on urban road.
(A) undistorted images; (B) IPM difference map; (C) detection results.
Fig 10.
Detection result of isolation fence: (A) undistorted images, (B) IPM difference map, (C) detection results.
Table 3.
Detection results on urban roads.