- There is an opportunity for sensors to become more intelligent.
- The new Z/I RMK-D digital aerial camera may some day be integrated with a laser scanner to extract features at the sensor.
- NASA’s High Resolution Stereo Camera is capable of automatically producing stereo images, including quality checks.
This is something that I have thought was a direction that the 3D imaging industry should be moving in – namely developing the next generation of sensors that can collect and process data more intelligently. As I have noted in the past, scanners are dumb, particularly terrestrial scanners. They collect far too much data that is often not needed to extract the desired information.
Jeff Thurston recently reported on a new digital aerial camera from Z/I, the RMK-D that he thinks may some day include an integrated LiDAR sensor. He goes on to postulate that this could lead to image extraction being done directly on the sensor, such that by the time the aircraft lands that the desired information would be available. I am sure the military would be interested and probably working on this approach.
As possible evidence of this speculation, the September 2009 PE & RS cover image is a digital terrain model for the Mawrth Vallis area of Mars. It is one of the 4 possible landing sites for the next Mars rover mission. This DTM was created from data collected by the High Resolution Stereo Camera – HRSC. This is the first sensor with built-in stereo capabilities that has been included on an interplanetary mission.
The stereo models were created from automated procedures that included standardized quality checks. There were no interactive measurements or data editing procedures used to produce the results. Sounds like what Jeff was thinking about, and an important step in the right direction.
Can the terrestrial laser scanner manufacturers learn from this approach?
It does certainly seem like sensor fusion and on-board processing is a current growth trend. One big benefit of flying with dual sensor technology is that (ideally) on-board processing could be performed to create orthophotos using the DTM from the LIDAR sensor. This would require some thought in re-jigging current ortho workflows (e.g. pull an XYZ for each image pixel from the LAS data without merging or reformatting the terrain) to make the system fast enough to get initial processing done on-board. But the implications are great for areas such as rapid response mapping.
The HRSC image seems to be an example of automatic terrain extraction – something that has been available in commercial softcopy photogrammetry systems for awhile. Some further info is available here: http://aisrp.nasa.gov/projects/5184ed69.html. There is a lot of work being done in this area in both the photogrammetry and computer vision fields – culminating in the commercial release of a few different “next generation” automatic terrain extraction systems.