Camera offset survey

[Jim Campi]

Gentlemen,

Is there preliminary documentation on the use of camera offset survey?

 

[Matt Johnson]

As with most screens there is documentation available if you press the Help hardware button.

 

[Jim Campi]

As usual, i forgot anout this.

Thanks Matt.

 

[Shawn Billings]

I've been skeptical about this feature since learning about it almost two years ago. In the late spring I briefly looked at it and it still needed some improvements, so I set it aside. Then we went to work testing the new RTK V6+ engines, then Beast Mode 5Hz RTK. Now that my testing is all but finished with those, I've turned my attention to Camera Offset Survey this week. My first day of testing showed a lot of promise. I've never done photogrammetry before, so it's a learning process for me, as well as testing the software. My second day of testing showed unbelievable results, even for my friend Mikhail, the lead engineer on this feature. Then I took it to a real job that we had done a couple of weeks ago, and did it again as if from scratch and compared the results. The original survey was a mix of RTK and total station, as almost all of my work is. My test, however, was all RTK and Photogrammetry - 100% Triumph-LS. An article will soon be posted to Javad's website about it, but I can say that the accuracies I am seeing are very encouraging. I located 30 points by photogrammetry that were also located by total station from RTK control points. The average residual (horizontal coordinate difference) between the two was 0.11'. The standard deviation of the residuals was: 0.08', with the highest extreme residual of 0.29'. Vertically the results have been even more impressive, likely a function of the geometry of my Camera Points. There were four points in the two surveys which had elevations determined, the vertical differences were 0.10', 0.02', 0.05', 0.04'. I believe I'll be able to produce more accurate results with a bit more practice, better targets (I used rolls of survey ribbon for photo control in this test) and a little more consideration for time of day (some of these shots were looking into the setting sun).

I have not previously recommended the purchase of this feature to users because I was not sure of its capability. Now I can highly recommend it for those who do a great deal of topographic work. Particularly for solo operators, this may allow you to keep the total station in the truck and complete the job with one complete tool. Be sure to catch the article in a few days at http://www.javad.com/jgnss/index.html

 

[Jim Campi]

Thanks Shawn. I really appreciate you following up on this post. I will certainly be watching for your article.

Do you have an estimate of the time involved collecting the 30 photogrametry points?

 

[Shawn Billings]

I actually collected over 100 points. I just happened to have 30 points that I could compare. Capturing photos took about 1.5 hours. Processing photos took about 4 hours. With practice I think this will be reduced quite a bit.

Yesterday I photo surveyed an oil derrick, it's about 100' tall. I wanted to see if this would be good for surveying small towers. I need to get some total station data to verify results, but it looks very promising.

It's like I'm not even doing rtk anymore. The rtk just supports the photogrammetry. Everywhere I look when I'm driving down the road I'm seeing points in space to capture with photogrammetry. It's like a drug.

 

[Jim Campi]

Hi Shawn,

I have been considering this feature for some tine and had it activated yesterday. I'm going to do some preliminary testing this weekend and coming week.

Do you have any quick thoughts to get started? I am looking forward to your article for a detailed review.

Thanks again for your time...

 

[Matt Johnson]

Jim, here is a rough draft of the camera manual that Mikhail wrote.

 

[Jim Campi]

Thanks Matt!

 

[Zoltan Varga]

Hello,
I also tested the camera offset feature two weeks ago. I measured only 5 points, and they were only 10-20 meters from the TLS. The result was promising, but I am still sceptic. I am not convinced that this measuring method is more productive than survey with total station.

 

[Zoltan Varga]

Here is the result.

 

[Zoltan Varga]

Here is the result.

 

[Zoltan Varga]

and some screenshots

 

[Jim Campi]

Hi Zoltan,

I appreciate the data. I haven't tested yet, but will over the next few days.

The bulk of surveying I do is support for my design projects. So, it's mostly topos and as-built drawings.

Prior to the increase in base transmission rate there were many sites that require supplementation with an optical instrument to mitigate overhead or lateral instructions.

I'm still testing however; I just resurveyed a site in Napa (alone) with the LS that wasn't efficient or possible with prior to the update. So I used a crew and total station.

There were 2 points that the LS wasn't able to collect in an area adjacent to a concrete wall with a heavy vegetative canopy. I was able to make the survey work without those points. In some cases that won't be possible. My hope is that photogrammetry will allow collection of these difficult points without the requirement for a robotic total station and the cost/maintenance/hassle of taking a second instrument to every site.

 

[Jim Frame]

My hope is that photogrammetry will allow collection of these difficult points without the requirement for a robotic total station

My thought is that most -- if not all -- points that are susceptible to T-LS photogrammetry within normally acceptable accuracy limits would be obtainable with equal or greater accuracy using a reflectorless instrument. It wouldn't have to be robotic. And the major advantage of reflectorless is that you'd have the results in the field, without having to wonder if you got enough photos to capture the points. So I'm still skeptical of the value of the photogrammetry option. If it were included as part of J-Field I'd probably try it out, but I'm not about to spend money on it at this point.

 

[Jim Campi]

My thought is that most -- if not all -- points that are susceptible to T-LS photogrammetry within normally acceptable accuracy limits would be obtainable with equal or greater accuracy using a reflectorless instrument. It wouldn't have to be robotic. And the major advantage of reflectorless is that you'd have the results in the field, without having to wonder if you got enough photos to capture the points. So I'm still skeptical of the value of the photogrammetry option. If it were included as part of J-Field I'd probably try it out, but I'm not about to spend money on it at this point.

Hi Jim,

I completely understand your position, especially for dedicated surveyors. If I happen to be in a remote location and don't have my optical instrument, then this feature can pay for itself quickly.

 

[Shawn Billings]

It's got limitations but it's extremely portable. It can be processed in the field, it just doesn't need to be. I'm still learning how to get the best accuracy with it. Every time I use it I learn something new and get better results.

 

[Shawn Billings]

This is a brief article I wrote about my third attempt at photogrammetry, locating structures around a gas station convenience store. We did the job a few weeks ago, and I went back and did it again with photogrammetry.

http://www.javad.com/jgnss/javad/news/pr20150916.html

 

[Jim Campi]

Thanks Shawn,

Are your control points referenced to an autonomous base position?

Just out of currosity, how low was your calibration number?

 

[Shawn Billings]

Everything is based on fixed RTK - photo coordinates and control points. I would locate the control points with RTK then collect photos. The base is irrelevant provided that all of the RTK rover positions are related to the same base position. In this case the base was my office base station, broadcasting corrections over the internet.

I believe my calibration px RMS was around 0.73.