Object Recognition for Orange Construction Barrels using Color Segmentation

The problem of safe navigation and object recognition in autonomous vehicles become increasingly important as more of these vehicles take to the road. To be viable, autonomous vehicles must process road and vehicle information as rapidly and as accurately (or more accurately than) human drivers. Among the challenges for autonomous vehicles is the need to identify and deal with potential hazards such as lines, barrels, potholes, and construction barrels in the roadway. Orange and white-striped barrels, a common indicator of construction, present a particular problem for systems electronically processing information. To identify orange construction barrels, this paper proposes using a general color relationship which would be more versatile and faster than full or partial color segmentation. In addition, the paper offers a methodology that presents a significant cost savings over traditional LADAR systems