With the passing of Dr. Ashjaee, I've been looking at the products and technology that he has created again.?ÿ I've always been intrigued by them.?ÿ ?ÿ
?ÿ
I've been watching some of the great videos that Shawn Billings has made, and was wondering if any of the users here could speak to the photogrammetry aspect of the software/hardware.??ÿ?ÿ
Do you use that capability much??ÿ Do you use it for residential surveys??ÿ ?ÿ It almost seems like you could limit the use of optical/robots with some of the capabilities.?ÿ ?ÿ ?ÿ
Always been a tecnho equipment nerd, so I'm just curious how you users are integrating the equipment.?ÿ?ÿ
?ÿ
I'm not sure if I ever met Dr. Ashjaee, but I know Dr. Reilly at NMSU always spoke highly of him and I've been to a few ACSM National conventions, so I might have gotten a chance to say hello at least.?ÿ ?ÿHe will be missed.?ÿ ?ÿ
?ÿ
PS,?ÿ Shawn should do voice over work for audible.com !?ÿ?ÿ
Andy, thank you for the kind words.
?ÿ
The photogrammetry module works as advertised but its application is somewhat limited. One of the things I appreciated about it was that I learned a lot about photogrammetry using it. One of the things I learned was just how critical geometry is. Photogrammetry is based on a lot of bearing-bearing intersections - in 3D. It's really not good enough to simply look at photogrammetry from a 2D perspective, even if all you are interested in is the planimetric outline of, for example, a building. Each point (such as a building corner) needs to be located three dimensionally. To do this you really need three dimensional geometry. Also, you need a minimum of 4 photos to calculate the intersections with some measure of redundancy. In breaking up a building for photogrammetry, I would look at each wall as a "scene" (my word - probably not technically correct). Each wall would get 5 photos minimum. Due to the relatively low resolution of the camera, my working distance is going to be less than 100 feet. Varying the distance helps with that third dimension, particularly if you have an object that is common to all photos but not in the same range as your building wall. For example, in the video for the oil derrick, the key element that made the height calculation work was the horizontal bolt in the top of the power pole that was between my longer camera locations and the derrick. The photogrammetry module requires the user to manually select the objects in the photo, carefully putting a crosshair on the exact point in each photo, careful to make sure that the name of each point selected is consistent across the photos. So it's not like Pix4D that will automatically create a surface from the images. This is because of the processing overhead on a portable device like the LS. As I mentioned before, each of those items must be carefully selected in 3D. So for a building corner, you need a specific target at a specific height, not just somewhere along the corner's edge.?ÿ
There are places I think that photogrammetry could be very useful. One thing I found in one experiment was that I could actually get better precision from the combination of multiple RTK points and photogrammetry than I would get from a single RTK position. This experiment with five quarters on my concrete driveway discusses it:
?ÿ
https://support.javad.com/index.php?threads/camera-offset-survey.802/page-3#post-10802
?ÿ
I've sold a the module to a couple of customers, but I try to manage expectations. If nothing else the experience with the module has convinced me of the incredible power of photogrammetry in general and in the limitations of photogrammetry. It's not quite as simple as a couple of photos and process (which I thought might be the case when I first started using it), but it has the ability to produce some incredible precisions when you can apply good procedures to a scene (varying angle of camera position around scene, varying distances of camera position from scene, selecting processing points at varying depths and varying heights).?ÿ