Projects

Advanced Driver Distraction Detection System


Source Organization:
University:Carnegie Mellon University
Principal Investigator:Fernando De La Torre
PI Contact Information:ftorre@cs.cmu.edu
Project Manager:Courtney Ehrlichman
Funding Source(s) and Amounts Provided (by each agency or organization):$85,000
Total Dollars:
Agency ID/Contract/Grant Number:
Start and End Dates:January 2016 - January 2017
Project Status:Active
Subject Categories:In-Vehicle Technologies, Vehicle Automation Technologies
Abstract:According to the National Center for Statistics and Analysis about 50% of car crashes are due to driver distraction. Monitoring a driver's activities form the basis of a security system that can potentially reduce the number of accidents by detecting anomalous situations. Because distraction while driving is a leading safety issue, the aim of this proposal is to develop an improved driver monitoring system. To achieve this goal, we propose to build a system that concurrently monitors the driver's focus of attention on the road and performs early detection of some of the driver's activities. Our advanced driver detection system (ADDDS) will have one camera facing the driver and another capturing the driver's field of view. The proposed system is therefore capable of monitoring both the driver's status and the environment surrounding the car. There are two main modules of ADDDS: (1) Detecting where the driver is looking in the road, and (2) detecting the driver's behavior. To detect where the driver is looking, we propose a method for mapping from the driver's gaze to the exterior view that is based on a 3D reconstruction of the interior and exterior of the car. ADDDS avoids the necessity for re-calibration in a controlled environment cause by changes in the driving position, which is the main drawback of other approaches. To detect the driver's behavior, we will propose a novel temporal classifier for detecting a driver's activities such as talking, texting on the phone, operating the radio, eating, or talking to the passengers. Both systems are combined to create a unique framework that takes advantage of a complete reconstruction of the driver/car environment to reliably estimate driver distraction.
Describe Implementation of Research Outcomes (or why not implemented):
Impacts/Benefits of Implementation (actual, not anticipated):
Project URL: