Sensor Fusion, Virtual Focal Plane Array
NEW
Matlab code for calibration of planar laser range finder and camera,
including (minimal) sample data for testing
2.8M Bytes, zip file.
There is a "Readme.txt" file in the "manualCalib" code that
details what else needs to be on the system and how to run the
program.
In an ideal world, the coordinate systems of all sensors would match,
and the data from each sensor would be presented or drawn on the focal
plane array. When this is not the case, it is necessary to find the
transformation between the coordinates of each sensor, and project the
data from each onto a Virtual Focal Plane array.
We have developped techniques for finding the relative calibration of
heterogenous sensors (sensors of different modalities). We use
extensions to standard camera calibration techniques for "in the
laboratory" calibration
(tech report).
Even more relevant to many autonomous vehicle
applications, are the auto-calibration techniques which can find the
relative calibration for a large class of sensors that are mounted on
a moving vehicle, by solving for constraints between the data
captured simultaneously by each sensor.
(tech report).
Sample results are show here, first, a movie of the calibration
between a laser range finder and a video camera slowly improving as
the robot moves and gives joint motion constraints:
auto-calibration movie (0.9Mbytes avi)
In the case below, the laser range finder and the camera have been
relatively calibrated and the laser readings are projected onto the
image. The calibration object can be seen by mousing over the
farthest right option, labelled "Calibration". When calibrating the
laser and camera together, we can extend the constraints to improve
the estimation of the camera intrinsic parameters to be better than
solving for the camera calibration by itself.
Geometric Calibration of Laser Range Finder and Video Camera:
Quantitative Studies of Improvements in Ego-Motion Estimation using
combined data, (coming soon)