The image formation process for airborne lidar systems utilizes physical sensor models
in order to create three-dimensional imagery
from range, scanning, position and attitude sensor measurements.
To estimate surface locations from data collected with Geiger-mode lidar systems,
typically range measurements are combined from many laser pulses,
from many camera pixels, over multiple viewing geometries and multiple collection times.
Sharp image formation requires well-calibrated camera and pointing models.
The calibration method presented here utilizes aerial data collections
spanning the full sensor field of regard to iteratively refine the system model parameters
while minimizing relative imagery misplacement.
Step one of the calibration adjusts parameters that only affect image quality
and evaluates image sharpness using, for example, a Sobel sharpness filter.
Step two adjusts parameters affecting image placement.
This calibration method utilizes the vendor-supplied transform code
inside the parameter optimization algorithm
and is therefore independent from the lidar configuration and the physical sensor model.
Utilizing the vendor-supplied transform ensures that the resulting calibration parameters
are interpreted correctly.
To reduce computational burden in step two, the input data from thousands of laser pulses
is condensed into a single effective pulse that is passed through the transform step.
This method has been used to calibrate four different Geiger-mode lidar systems
and has enabled sharp imagery over the full field of regard
prior to incorporating ground control points or performing registration with adjacent imagery.
We present representative results for the Cuchillo lidar system
and discuss execution time requirements and performance limitations.
|