118. EMBEDDED VISION SYSTEM FOR SURROUND UNDERSTANDING OF HIGHWAY DRIVING

Department: Electrical & Computer Engineering
Research Institute Affiliation: California Institute for Telecommunications and Information Technology (Calit2)
Faculty Advisor(s): Mohan M. Trivedi

Primary Student
Name: Sean Lee
Email: yhl014@ucsd.edu
Phone: 858-822-0075
Grad Year: 2016

Abstract
Vision-based driver assistance systems involve a range of data-intensive operations; therefore, on a resource constrained embedded platform, there is a tradeoff in designing a highly robust system while maintaining an acceptable computational speed. In order to balance and optimize both accuracy and speed performances, we develop an efficient region-based vehicle detector for highway driving that utilizes lane localization and under-vehicle shadows in a synergistic manner. Using this detector, a novel two-camera based embedded driver assistance system that models the dynamics of vehicles in the front and rear surround views of the host vehicle is implemented. The system runs real-time on Snapdragon processors and detailed evaluations show high accuracy with true positive detection rates greater than 95% and false alarm rate less than 6%. Furthermore, the system is able to analyze threats imposed by surround vehicles and generates a safe maneuver zone that helps drivers prevent collisions. This application as well as the proposed embedded system was demonstrated at the 2016 Consumer Electronics Show (CES) and it currently operates in our research vehicle testbed under real-world driving conditions. Further applications can be easily extended in the future since the system has a scalable framework to support additional sensors, processors, and functionalities.

Industry Application Area(s)
Intelligent Vehicles, Driver Assistance Systems, Embedded Vision

« Back to Posters or Search Results


Contact:   researchexpo@soe.ucsd.edu   (858) 534-6068