A.3.6.4-6 Long-lead-time studies of more demanding vision tasks
A.3.6.4 Recognition of 3-D shape and occlusion of trucks
Contrary to cars, for which the lower part of the body can be well approximated by a (rounded) rectangular box, trucks usually have a simple shape as top part (often a much larger rectangular box than car bodies), and it is above ~ 1 m elevation). A much more irregular appearance can be noted in their lower part with axles, wheels, tanks, chassis etc. Due to the large size, trucks often obscure a large fraction of the scene, especially when nearby. To be able to check the possibility of passing, the length of the truck should be recognized rather precisely from a slightly oblique view.
Another typical dangerous situation is experienced when from a sequence of trucks in the neighboring lane, the second or third one ahead suddenly starts switching lane with no or only a short notice by a blinker signal. This topic has been looked at by [Schmid 1994].
Another lesson learned for oblique views was that the diagonal of a rectangular box can be estimated much better than the equivalent parameters length and width, especially when rounded corners are present like in real vehicles.
Video 13 TruckTracking LaneChanges 1992
video 21 Disocclusion2TrucksHIL-Sim 1995
Schmid M (1993). Real Time Recognition
of Partially Occluded Vehicles by Dynamic Vision. ECC,
Schmid M, Thomanek F (1993). Real Time Detection and Recognition of Vehicles for an Autonomous Guidance and Control System. Pattern Recognition and Image Analysis, Vol. 3, No. 3, pp 377-380
Schmid M (1994). An Approach to Model-based 3-D Recognition of Vehicles in Real Time by Machine Vision. IEEE-Conf. on Intelligent Robots and Systems (IROS'94), Neubiberg
Schmid M (1994). 3-D-Erkennung von Fahrzeugen in Echtzeit aus monokularen Bildfolgen. Dissertation, UniBwM / LRT