This page contains videos related to image aided INS vehicle positioning for highway applications. The objective was reliable where-in-lane positioonsing accuracy (horizontal error <0.5m). This was successfully demonstrated in numerous scenarios using Vision, Lidar, and Radar separately to aid the INS. Demonstrations occurred at two locations: Department of Transportation Turner Fairbanks Research Center and University of California, Riverside.

UCR Videos (Related to [1])

  1. Traffic Light Detection: UCR to CE-CERT (Daytime and replay at 5Hz)
    (No data association, only the feature detection algorithm is tested on real world driving data. The frames are captured at 1Hz while the video playback is sped up to 5Hz.)
  2. Vision/GPS-Aided INS: Parking Lot Data Shown in the Paper (Daytime replay at 5Hz) (with full debugging window)
    (A shorter experiment using only visual feture and code to correct the INS. Doppler is used in the begining to initialize yaw and turned off around the 30th second. The data is captured at 1Hz and the replay is sped up to 5Hz.)
  3. DGPS-Aided INS: Parking Lot Data Extended Experiment(Daytime)
    (includes carrier-phase from 90th to 127th second, the rest of the time use only code and Doppler to aid the INS.)
  4. Vision/Code DGPS-Aided INS: Parking Lot Data Extended Experiment(Daytime)
    (Vision aiding is only available whenever a visual feature is within the field of view and is observable. Visual feature, code, and Doppler are enabled to correct the INS.)
  5. Vision/Code DGPS-Aided INS: Parking Lot Data with 2 Traffic Lights(Evening)
    (Vision is enabled in the beginning to help initialize yaw. Visual feature, code, and Doppler are enabled to correct the INS.)
  6. Simulated Urban Canyon Vision/Code DGPS-Aided INS: Parking Lot Data with 2 Traffic Lights(Light Rain in the Daytime)
    (Code and Doppler measurements along the West-East direction are disabled to simmulate an urban canyon. The uncertainty (sqrt(P)) quickly grows in the east and down directions whenever vision measurements along the north direction is not available. Banned satellite vehicles PRNS are listed under the BANNED PRNS section.)

DOT Turner Fairbanks Research Center Videos
Note the significant tree coverage and blockage by the TFRC building.

  1. Driver GUI and Camera Image: May 24, 2012
    Live driving demo and TFRC. Left is the camera image showing the view from inside the car. Right is the GUI image based on the estimate real-time vehicle state and the computed sign and road-edge map. INS aiding uses differential carrier phase GPS. The 5, 10, and 15 sigma uncertainty ellipses are plotted. The green line that projects from the vehicle location shows the expected vehicle path.
  2. Sign Aiding: GUI and Camera Image: May 24, 2012
    Live driving demo and TFRC. Top row uses only INS and pseudorange GPS aiding. Bottom row uses INS aided by both pseudorange GPS and vision with signs. Left is the camera image showing the view from inside the car. Right is the GUI image based on the estimate real-time vehicle state and the computed sign and road-edge map. The 5, 10, and 15 sigma uncertainty ellipses are plotted. The green line that projects from the vehicle location shows the expected vehicle path.

The videos are encoded using a variety of video codecs. If your PC does not have the required codecs, we suggest that you use VLC media player to play the videos.

[1] A. Vu, A. Ramanandan, A. Chen, J.A. Farrell, M. Barth, ''Real-Time Computer Vision/DGPS-Aided Inertial
Navigation System for Lane-Level Vehicle Navigation,'' December, IEEE Transactions on Intelligent
Transportation Systems, 13, 2, 899-913, 2012