Intelligent Vehicles: Creating smarter vehicles: So you don't have to be

Intelligent Media Lab University of Vermont Burlington, VT 05405

Research Done by:   Michael Finnefrock                                    337 Votey Building                                    University of Vermont                                    Burlington, VT 05405     Reason of Research:  MS research in Electrical Engineering in the field of robotics and computer vision.      Abstract:  Over the past decades most advances in vehicle safety has been on the behalf of the driver and the passengers within the vehicle. However, the increased number of safety devices on a vehicle will not be able to effectively reduce the number of fatalities caused by traffic accidents [a]. In order to make vehicles safe, advances must be made, which not only protect the vehicle’s passengers, but also protect the surrounding environment. Until advances in autonomous vehicle control become widely accepted and trusted by the average driver, other devices must be used. These systems will not have control of the vehicle’s trajectory. Therefore, they must provide safety by giving early warning to the driver of dangerous obstacles. The goal is to provide the drivers with information that they can not usually obtain by themselves [c]. I propose to use a single infrared camera to provide the driver with pertinent information such as object/collision detection. Stereo-vision systems are considered to be effective candidates [a]. However, there are two major drawbacks to these systems. The first being the extra overhead involved in computing correlations between left and right images. The second drawback is the computations involved in extracting image patterns. A color-camera based system is a good way of providing the driver with visual feedback. This allows the driver to judge and respond accordingly. Similar visual feedback can be supplied to the driver via infrared cameras. Using far infrared cameras allows for object detection in low lighting situations, where normal color - and even black and white - cameras are unable to produce clear images. Night driving is responsible for more than 50% of fatal accidents, yet less than a 50% of all driving occurs at night. An on-board vision system should be able to assist a driver during expressway and high speed driving as well as in urban conditions where the ability to detect pedestrians is essential. Current pedestrian detection is aimed at sensing and localizing the objects with a human shape [d]. Using infrared imaging one can localize ‘hot-spots’ (HS) using relatively simple algorithms and relatively low computations. A HS can be characterized by both its temperature, or by comparing it to the ambient temperature. Once a HS has been found its pixel area can be localized and further examined for further qualities. This second process will determine what the object is that has been discovered. The object will be examined for trajectory and collision calculations can be made. Based on this outcome the driver will be alerted to the position, trajectory, and potential danger of the object via HUD. A few companies have already placed infrared devices on production vehicles. However, these systems were limited and short lived. In order for these options to become popular they must have two qualities. There must be a buying fascination for the customer as well as effectiveness to prevent the accidents [a]. I believe infrared imaging and Heads Up Display (HUD) units have enough glamour to attract potential buyers. All that is missing is functionality. With this system in place there is will be a significant increase in vehicular safety. References: [a] Furukawa, Y. “Status and Future Direction of Intelligent Drive Assist Technology.” IEE Intelligent Transportation Systems. 2000 [b]Cherfaoui, Veronique and Burie, Jean-Cristophe. “Dealing with uncertainty in perception system for the characterization of driving situation.” IEEE ITS Conference Proceedings. 2000. [c] Yamaguchi, Toru, Hiroshi Nitta, Tomohiro Takagi and Hideki Hashimoto. “Human Centered Sensory Intelligence Architecture for ITS. [d] A. Broggi, M. Bertozzi, A. Fascioli, M. Sechi. “Shape-based Pedestrian Detection.” IEEE Intelligent Vehicles Symposium. 2000.