I'm not a photogrammetrist, but I've been dabbling with the photogrammetry feature in the Triumph-LS. It's bee a great deal of fun. I'm learning how to capture objects at multiple ranges to produce the best results. Last week I surveyed a steel structure, a replica oil derrick. My first attempt produced ok results. Then I got the idea to include some overhead objects at closer range, and the results went from ok to incredible. I'm hooked now. Photogrammetry is like crack. Good luck shaking it.
I've always wanted to dabble with it, but finding a suitable camera, calibration routine for the camera and software to process it was always a challenge.
I made a video that briefly shows what I did and the results. I know that Trimble offers the V10 that captures panoramic images that can be processed in the office, although I don't know much about it. Given that the technology is accessible and is capable of producing such great results, it will probably be good for us to have healthy discussion about how to get the best results, what the limitations are, what the capabilities are, what differences exist between different processing techniques.
I'm not sure about the Trimble V10, if it automatically processes every pixel in a photo creating a 3D model. In the LS, everything is "on-board" instead of being done on a PC, and only points that are manually selected by the user are processed. The user picks points that are common to multiple photos. This reduces processing overhead and generates only the points the user specifies, but also requires more time on the user's part in manually selecting the points visually.
It's exciting technology. Photogrammetry is nothing new, of course, but the access surveyors have to photogrammetry is only going to become more prevalent with terrestrial close-range options and aerial UAV platforms.
[MEDIA=youtube]Y-LY-_9ywio[/MEDIA]
Cool Stuff Shawn, Bleeding Edge of Technology...
IS this Distance Distance intersection, with RTK GPS? That's what it looks like to me. Only a 3.1 MP cam? What could it do with a 12.1 MP cam?
N
Bob LeMoine, post: 337356, member: 10375 wrote: Cool Stuff Shawn, Bleeding Edge of Technology...
Some days you're the hammer, some days you're the nail. That day I got to be a hammer. 🙂
Nate The Surveyor, post: 337361, member: 291 wrote: IS this Distance Distance intersection, with RTK GPS? That's what it looks like to me. Only a 3.1 MP cam? What could it do with a 12.1 MP cam?
N
No, I think it's more like intersecting angles.
More pixels would provide better resolution for sure. I don't think originally it was intended for the internal camera to be used for photogrammetry. They looked at some mounted camera option, but haven't done much with the idea that I'm aware of. Then they thought to see what the internal camera would do. Turns out, under favorable conditions it can do a lot.
OOPS That's what I meant to say. B/B int. IF it could do distance, you could shoot a cut off stob, sticking up 4', with flag, set the LS on it, then get back and shoot the stob with the camera, including the fence corners etc, and it would essentially give you a backsite, (the stob) Inst pt, (point where cam and LS is) and FS (fence corner, under overhang) Boy, it'd be nice if it had a scanner in it!
Shawn Billings, post: 337271, member: 6521 wrote: I'm not a photogrammetrist, but I've been dabbling with the photogrammetry feature in the Triumph-LS. It's bee a great deal of fun. I'm learning how to capture objects at multiple ranges to produce the best results. Last week I surveyed a steel structure, a replica oil derrick. My first attempt produced ok results. Then I got the idea to include some overhead objects at closer range, and the results went from ok to incredible. I'm hooked now. Photogrammetry is like crack. Good luck shaking it.
I've always wanted to dabble with it, but finding a suitable camera, calibration routine for the camera and software to process it was always a challenge.
I made a video that briefly shows what I did and the results. I know that Trimble offers the V10 that captures panoramic images that can be processed in the office, although I don't know much about it. Given that the technology is accessible and is capable of producing such great results, it will probably be good for us to have healthy discussion about how to get the best results, what the limitations are, what the capabilities are, what differences exist between different processing techniques.
I'm not sure about the Trimble V10, if it automatically processes every pixel in a photo creating a 3D model. In the LS, everything is "on-board" instead of being done on a PC, and only points that are manually selected by the user are processed. The user picks points that are common to multiple photos. This reduces processing overhead and generates only the points the user specifies, but also requires more time on the user's part in manually selecting the points visually.
It's exciting technology. Photogrammetry is nothing new, of course, but the access surveyors have to photogrammetry is only going to become more prevalent with terrestrial close-range options and aerial UAV platforms.
[MEDIA=youtube]Y-LY-_9ywio[/MEDIA]
Interesting video.
Javad's Triumph LS website states, "Offset Survey with Photogrammetry
Offsets can be calculated using the internal camera of the TRIUMPH-LS or with an external camera!"
Have you used "an external camera"? What images can be used/processed?
George Matica, post: 337633, member: 6663 wrote: Javad's Triumph LS website states, "Offset Survey with Photogrammetry
Offsets can be calculated using the internal camera of the TRIUMPH-LS or with an external camera!"Have you used "an external camera"? What images can be used/processed?
I don't know much about the external camera, except for this:
It has to be calibrated. It has to be mounted to the receiver. It has to have a known offset from the ARP of the receiver. I've not used the external camera, nor have I seen it, except in ads.
Part of the success of the system is having the known position of the camera. I believe this is also true for the Trimble V10. It removes a variable that must be solved.
From the September 18, 2013 GPS World article... http://gpsworld.com/new-javad-triumph-ls-receiver-features-864-channels/&apos ;">New JAVAD TRIUMPH-LS Receiver Features 864 Channels
"The TRIUMPH-LS provides visual stake-out, six parallel RTK engines, more than 3,000 coordinate conversions, advanced coordinate geometry features, and rich attribute tagging on a high-resolution 800 x 480 pixel display. When used in photogrammetry, offsets can be calculated using the internal camera for 10-centimeter accuracy, or an external camera for 5-centimenter accuracy. TRIUMPH-LS is the first JAVAD GNSS receiver to offer photogrammetry for land survey."
It would be nice to read the docs but we haven't been able to locate/download a manual for J-Field or specifics regarding "Offset Survey with Photogrammetry" from Javad' website.
Thanks.
George,
It would appear to me that refinements have provided the internal camera with performance exceeding what was previously expected from an external camera.
I am sure that the manual for specifics regarding "Offset Survey with Photogrammetry" will be coming soon. The J-Field is available in .pdf and printed formats. The video produced by Shawn Billings would be a good start regarding techniques.
I am sure more will be made available soon.
George,
I don't have a full handle on accuracy. It's dependent on baseline length, so those values would have to be qualified with a distance. I did two examples today at my home. I posted details about it on the Javad forum, so I won't go into a lot of detail here, but this was what I did:
Test 1
Close range photogrammetry with Five Quarters. I placed four quarters in the shape of a square (roughly) on my level concrete driveway with the fifth at the intersection of the diagonals (roughly). I then took four photos around them from close range: 15-20' away. I then taped between the five points. This gave me 10 baselines between the quarters from 1.3' to 2.9' in length. I inversed between the points derived from the photogrammetry and compared them to my tape measurements. The differences were 0.000' to 0.006' with an average residual of 0.003'. No BS. Three thousandths of a foot. This was using the meager 3MP onboard camera. Elevations ranged from 9.051'-9.074'. I don't know if that much variation could be in my driveway. Either way, it demonstrates that all five points were within +/-0.012'. This could have some applications for outdoor industrial work.
Test 2
Intermediate photogrammetry with prism pole (limited geometry). I wanted to see what I could expect when I had to maintain a longer distance from my object and limit the field of view from the object to about 90å¡. I took four photos around a prism pole at a distance of 45-50 feet from the pole. I selected points on the pole, beyond the pole and left and right of the pole. I would have preferred to have included a near point, but I neglected to do so. The results were about 0.12' horizontal and 0.18' vertical from photogrammetry to direct RTK. Had I been able to get a bit wider or have additional targets at varying depths, I believe I could have done better, but that was not the experiment.
In either case, the internal camera appears to be able to produce much better than decimeter level results. I guess it's better to find the manufacturer spec'd it to be less accurate than it actually is than vice versa.
Based on my experience in UAV Photogrammetry, I would think that 3D model can be generated using your photos.
Let me know when you can stick a camera upside down into a manhole to get good pipe sizes, inverts, etc