VisualEyes - A Saab and VTI simulator study

ViP aims at unifying the extended but distributed Swedish competence in the field of transport related real-time simulation by building and using a common simulator platform. ViP partners are:

• VTI (Swedish National Road and Transport Research), 
• SAAB, 
• Scania, 
• Volvo Cars, 
• Volvo Trucks, 
• Bombardier, 
• Trafikverket (Swedish Road Administration), 
• Dynagraph, 
• HiQ, 
• Pixcode, 
• SmartEye and the 
• Swedish Road Marking Association. 

ViP is funded by its partners and by Vinnova.

The Visual Eyes project's goals are:

• Evaluation of the influence of factors such as wearing glasses and participants age on gaze tracking system performance for a 1-camera and a 3-camera system
• Development of a self-initialized visual attention detection module based on a one-camera system
• Development of a real-time visualization system for gaze direction.

In the first part of the project Smart Eye participated in a study comparing a 1-camera and a 3-camera system. The test runs were performed at the, then GM, simulator at SAAB in Trollhättan, and the subsequent post-processing of data took place at Smart Eye, while the high-end analysis was coalesced at VTI in Linköping. Some 50 test subjects, mainly SAAB employees, participated in the tests. The subjects had to look at well defined gaze targets inside the cockpit as well as outside on the 180 degree film screen, and the accuracy, precision and availibility per gaze target and per test subject could be extracted from the steady stream of Smart Eye data.

Some of the more obvious results of this investigation were:

• The 3-camera Smart Eye Pro system has a greater envelop inside so that more satisfactory gaze measurements can be made, as compared to the 1-camera AntiSleep system.
• The accuracy, precision and availibility degraded for both systems, as the gaze and/or head was rotated further away from the “straight” ahead direction, i.e.  away from the cameras.

One slightly surprising result was:

• The test subjects made one test run with glasses and one test run without. No statistically significant difference could be discerned between these two cases. This might in part be due to an unfortunate choice of mock-up glasses, but is nevertheless worth noticing.

At Smart Eye we are pleased to note another result:

• The 1-camera AntiSleep system competes well with the 3-camera Smart Eye Pro system and can deliver high accuracy, precision and availibility inside the central region, i.e. the region where the driver is looking most of the time. (At present the operational range for the AntiSleep 1-camera system is roughly +/- 30 degrees horizontally and +/- 20 degrees vertically.) We can conclude that the AntiSleep system is well designed and positioned for measuring driver fatigue, inattention and such, at a low cost and swift setup. Of course, if one wants to perform extensive research over a large gaze envelop and headbox, e.g. Driver behavior, then the Smart Eye Pro multicam (2-8 cameras) system is the obvious choice.