Publication Number: FHWA-HRT-11-056 Date: October 2012
Layered Object Recognition System for Pedestrian Sensing
4. KEY INNOVATIONS
The research team made the following key innovations during the execution of this project:
Detected and recognized multiple roadway objects in real-time. In addition to pedestrian detection, the system also detected regions that corresponded to buildings, poles, trees, and motor vehicles.
Used multiple cues to detect, classify, and track people. Stereo images provided depth information. Shape and appearance were used in the classifier, and motion cues were used to perform object tracking.
Used contextual information about regions such as buildings, trees, poles, etc., derived from depth information to reject a majority of the FPs that were detected by the initial pedestrian detection module.
Page Owner: Office of Research, Development, and Technology, Office of Safety, RDT
Topics: research, safety
Keywords: research, safety, Pedestrian Safety, Pedestrian detection, Stereo vision, Disparity map, Histogram of oriented gradients (HOG), Contour-based classifier
TRT Terms: research, Safety and security, Safety, Transportation safety
Scheduled Update: Archive - No Update needed