Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - andyroo

Pages: 1 ... 25 26 [27]
Thank you Alexey, disabling the blurry photos worked great. I was thinking that it would be nice to be able to organize them in a "preferred" hierarchy so I could send the blurry ones to the back, so they were only used where no other imagery was available, but I realized I can do this myself by exporting multiple images, then mosaicing them in GIS.

General / Re: aerial photography?
« on: May 07, 2012, 10:05:05 PM »
Hi sigma,

I am getting good results - much better than 5 m resolution - using a canon D10 ($250) bolted to the flightstep of a Cessna 172 with $50 worth of ram mounts (B-231ZU base, short arm, 1" ball).

I used CHDK to boot a modified firmware for the camera and run a 3 second intervalometer script I wrote, with shutter speed of 1/1500s and the neutral density filter forced off.

Using Photoscan Pro I am able to generate good quality DSMs and orthos, just by setting the ground plane with coordinates georeferenced from best available data that I can ID on the photos. I am using some LiDAR to get the z values, and 6" aerial imagery from bing maps to get the xy. The DSMs need some cleaning (I have lots of noise where the water is), but that might be better done in CAD or using meshlab, or something like that.

I am using a few hundred aerials to generate a DSM. I've been including the photos with blur because (1) it's more time consuming to weed them out and (2) it seems like it doesn't affect the DSM results.

But when I generate the ortho, I am finding that some of the important areas get blurry photos (and/or photos from a higher-elevation pass with less detail).

Is there a way to specify which photos are used in the mosaic? If I could check and uncheck photos that would be useful. What I would love though is to be able to assign a priority. Often the lower quality/higher pass/blurry areas fill in important areas where blurry data is better than no data, but the higher resolution high quality data would look much better.

On a related  note, does anyone know about how I can assign a "blurriness index" to images to easily identify which ones are blurry?

Thanks for making PhotoScan Pro. I love it. :)

General / Re: Best way to roughly constrain alignment/geometry?
« on: May 06, 2012, 12:11:46 AM »
I figured out the issue, although I think I could have changed the processing order and fixed it even if I hadn't.

It turns out that I had accidentally removed six images from one of the flight lines, which I think was the first one being processed. This "broke" the imagery into two unreferenced sections. the three other flight lines then "stacked" imagery on each of these broken sections, the flight lines going up-valley tied to the first part of the broken set, and those going down-valley tied to the second part. When I added the six images back in, it properly aligned the images.

Re Alexey's question, there are four groups of images that are all partially overlapping (two passes up-valley and two down-valley). Along-track they have 65%-75% overlap depending on elevation. Across-track is more variable, since it's a hired pilot trying to keep straight and level over the river.

Attached another screenshot to show how it looks with proper alignment. Thanks for your help!

General / Best way to roughly constrain alignment/geometry?
« on: May 04, 2012, 09:51:23 PM »
I am generating surfaces from unreferenced aerial images, then assigning GCPs after geometry creation. On my first model it went very well. With my second one, I get weird results from a bad alignment calculation. Photoscan actually runs the pointcloud from one set of images right through the other (see attached image).

There is about 66 percent overlap between the images...

Since I haven't created geometry yet, I have to specify each GCP individually on each photo (then I was planning to re-run the alignment)

Is there a better way to do this? Or is there a way to better constrain the geometry that can be built? The photos represent 4-up, 4 down passes, so half of them are 180° from each other (and they are four different elevations).

Feature Requests / add superoverlay support for KML output?
« on: May 04, 2012, 10:24:34 AM »
Would it be possible to add superoverlay support for the KML generation? superoverlays offer the best performance of any imagery I've seen on Google Earth, taking advantage of the same tiling scheme that Google uses for the GE imagery. There are good examples here:

Google also mentions that they can be useful for generating models only on demand, and I wonder if this region-based and distance-based tiling of surface and imagery resolution could be incorporated into the output from PhotoScan Pro, either for orthophotos or textured models or both. I especially would use it for orthophotos.

a few years ago I was working on high-resolution (1m) shaded relief and backscatter imagery from multibeam sonar over a large area, and I wanted to make it available in Google Earth. Until I discovered superoverlays, all of the test imagery I generated was either too big to display or too coarse to appreciate the detail, but I was able to generate superoverlays that showed everything in 256x256 pixel tiles, optimized by a cool png-optimization technique, and stored as nested networklinks that put a minimal load on our server.

If this could be done within PhotoScan Pro, it would make it easy for your customers to generate large datasets they could serve or provide to clients that would be plug&play in Google Earth. The tiling schemes you have already allow for tiles to be created, and it seems like it wouldn't take much to add the kml to implement it.

I am generating an orthophoto and DSM from imagery collected with no camera GPS location and very few ground controls. X,Y Ground control is derived from 0.1 m orthoimagery from a few months ago, and Z ground control is from 2-year-old 1 meter Lidar DEM.

In Ground Control Settings:

/Camera accuracy - is this for camera position? Does this setting matter if I haven't specified GPS locations for camera? I have left at default 10 meter for now. When/why would I change this?

/Marker accuracy - I estimated 0.25 meter since my imagery and the orthoimagery have similar (~0.1 m) resolution. Does this value represent my certainty that I picked the right pixel for my markers?

/Projection accuracy - this is at 0.1 pixels. What does it mean? when would I change this?

Finally, if I change these values (and if not), do I want to optimize photo alignment after I add GCPs?

If I understand the manual and the process in PhotoScan Pro correctly, I think that I want to accurately constrain my error, optimize photo alignment, then produce the orthophoto and DSM. I want to see if I can use the imagery to measure volumes of eroded sediment (from a dam removal project), so I am very concerned about accuracy and will be comparing to Lidar that will be flown in a couple weeks. I think that some explanation and clarification could save me hours of trying different settings.

The imagery is of a river, so the three GCPs that I have defined are oriented mostly linearly (see screenshot). I have considered adding a fourth, but I have poorer quality coverage (fewer photos, more noisy surface) further from the river.

I am using ground features like road centerlines and centers of manhole covers and boulders as control, and I am trying to understand the proper settings for measurement accuracy in the ground control settings. Also trying to determine whether I want/need to recalibrate camera positions.

Pages: 1 ... 25 26 [27]