Analysis of a prototype LDW system designed by University of Minnesota's Center for Transportation Studies.
Nationwide, United States
Development of an Innovative Prototype Lane Departure Warning System
Summary Information
The University of Minnesota's Center for Transportation Studies (CTS) developed a prototype lane departure warning (LDW) system in order to determine components and settings that would effectively reduce the number of false alarms provided to the driver when utilizing LDW in their vehicles. Reducing the number of false alarms received by drivers would likely encourage drivers to keep the system enabled, providing greater safety benefits than are seen in practice currently. Many current systems deployed in the vehicle fleet rely on the presence of road lane markers to determine whether the vehicle is staying inside its lane, however, the presence of lane markers is not guaranteed because of normal wear and tear on the nation's roadways. The CTS LDW system utilizes one of its two detection techniques at any given moment, depending on the roadway conditions.
The CTS system utilizes an integration of the Lucas-Kanade (L-K) optical flow method, as well as the Hough transform-based lane detection method, in order to reliably track vehicle position within the lane from frame to frame, even when lane markings become temporarily unavailable. The front-view images captured by an in-vehicle camera were converted to their corresponding top-view images via a homography. Based on these top-view images, the L-K optical flow is then used to track points (i.e., the Harris 'corners') from frame to frame to find the vehicle's heading angle and thus its lateral position by evaluating the relationship between these tracked features. The lane detection method looks for the lane markers painted on each side of the road and if available, the system calculates the distance to it and compares that to a past value to determine where the vehicle is in the lane and in which direction it is headed. Only one of the two systems is utilized at any given time, determined based on the environmental factors detected from the front-view camera. Twenty separate tests were conducted for approximately forty-five seconds each on roadways at speeds ranging from 45 to 70 mph (depending on the roadway speed limit).
Findings
- Valid warnings were provided to the driver 92 percent (48 of 52) of the time when the vehicle crossed the lane boundary.
- False warnings, where the system issued a warning when it should not have, were only sounded 1.18 percent of the time (47 out of 4000).
- The Optical Flow tracking method, used when lane markings are not detected, issued false warnings at a rate of 18 percent, but was only utilized for five percent of the test time.
- The Lane Detection method only issued false warnings at a rate of 0.32 percent.
It is important to note that this is a prototype system that was mounted inside the vehicle for testing purposes and thus does not account for all "real-world" situations. Because it is mounted in the interior of the vehicle, the system is not functional in periods of rain (because of the wiper movement). Problems with light road coloration and reflectivity issues with the windshield were also noted.
