Photogrammetric calibration of the SwissRanger 3D range imaging sensor

摘要

Many robotic and industrial systems require 3D range-sensing capabilities for mapping, localization, navigation, and obstacle avoidance. Laser-scanning systems that mechanically trace a range-sensing beam over a raster or similar pattern can produce highly accurate models but tend to be bulky and slow when acquiring a significant field of view at useful resolutions. Stereo cameras can provide video-rate range images over significant fields of view but tend to have difficulty with scenes containing low or confusing textures. A new generation of active light, time-of-flight range sensors use a 2D array of sensor elements to produce a 3D range image at video rates. These sensors pose unique calibration challenges, requiring both the usual calibration of lens distortion (intrinsic calibration) and calibration of the time-of-flight range measurement (3D calibration). This paper presents our application of a photogrammetric calibration approach using inexpensive printed optical targets and off-the-shelf software to solve both intrinsic and range calibrations for the MESA Imaging SwissRanger 3100 range imaging sensor. We further identify specific calibration issues stemming from this sensor's correlation of reflectivity with measured range.