Stereo vision 2021 introduction

Recap of the 2020 team’s work

• procured two cameras and two lenses
  • 2.3MP a2A1920-160ucBAS (Sony IMX392) colour cameras from Basler
  • Horizontal resolution of 1920 pixels at 160 FPS
• tested the stereo vision pipeline using pre-rendered track footage via Blender

2021 stereo vision goals

• Detect real world blue and yellow traffic cones at 15m
• Achieve an estimated displacement error of less than 0.5m

Next step forward

• Familiarise myself with the current stereo vision pipeline
• Gather training data to test the robustness of the current pipeline
• Test the pipeline with ROS/Gazebo
• Test the camera calibration and the effects of motion (moving cameras were not tested previously)
  • Design experiments and methods of dealign with this problem

Once we receive word that we can install the sensors on the race car:

• print mounting tabs for the stereo cameras
• test the thermal requirements (may require external fans?)

Not covered this year

• Hardware optimization for the Jetson AGX Xavier