Computer Vision for Advanced Driver Assistance and Intelligent Transportation Systems.

摘要

With recent technological advancements in computing and sensing capabilities, the past decade saw an increasing number of computer vision applications being incorporated into intelligent transportation systems. These applications include in-vehicle systems as well as on-road monitoring systems, both of which aim to improve drivers' safety and ultimately all participants on the road. On one hand, automobile manufacturers' primary focus has been advanced driver assistance systems ranging from lane departure warning systems to adaptive cruise control to Tesla's autopilot. On the other hand, government sectors place their emphasis on roadside monitoring systems, including speeding detection, traffic flow measurement, and accident detection.;The unlimited potential of these systems is matched only by the research challenges involved in designing, building, and optimizing relevant computer vision techniques. For instance, one not only has to consider the accuracy of a vehicle detection module, but the time constraint it has to operate under as well as the hardware limitations of the device it has to operate on.;In this defense, we discuss our three major research focuses, namely a real-time mobile lane departure warning system, an overtaking vehicle prediction system, and a first-of-its-kind stereo system capable of detecting features indicative of drunk drivers. Experimental and field study results confirm that our proposed systems operate not only with high accuracy but also with great efficiency making them suitable for real-world applications.;It is of utmost importance to note that the aim of our proposed systems is not to replace human drivers or police officers. Rather, they are intended to reinforce the symbiosis between machines and human operators. We conclude at the end of this defense by outlining future research directions and the potential impact our work will have on the future of the intelligent transportation systems community and the society as a whole.