Targetless registration with HDS7000 and "The Coco Method"

Faro FOCUS S350, S150, S70, M70, X30, X130, X130HDR, X330, X330HDR. Leica RTC360, C10, P15, P20, P30, P40, P50, BLK360, HDS6*00, HDS7000,HDS8800, HDS8400. Riegl VZ6000, VZ4000, VZ2000, VZ1000 & VZ400i. Topcon GTL-1000, GLS2200, GLS1500. Trimble X7, X9, X12, SX12, SX10, TX8, TX6, S7 & S9. Z+F IMAGER 5016, 5010X, 5010C, 5006H, 5006EX & PROFILER 9012
Post Reply
User avatar
d.raneri
I have made 40-50 posts
I have made 40-50 posts
Posts: 45
Joined: Thu May 02, 2013 11:47 pm
10
Full Name: Domenic Raneri
Company Details: NSW Police Force
Company Position Title: Crime Scene Officer
Country: Australia
Linkedin Profile: No

Targetless registration with HDS7000 and "The Coco Method"

Post by d.raneri »

After seeing Jonathon's presentation in San Deigo at the IAFSM conference, I was very keen to see if we could experiment with a similar solution for our outdoor scans. We use a Lieca HDS7000 which contains no equipment for providing orientation or location to the scanner; we rely on either surface matching or targets. At our outdoor and crash scenes we find the black and white Leica targets quite impractical due to their need to be 20-25m from the scanner for Cyclone acquisition and the time taken to place, rotate and protect so many targets at a chaotic crime scene. As such we typically rely on surface matching and do not deploy the 3D scanner to most of our crashes.

Having seen Jonathon's fantastic and elegant solution to combining a total station with a laser scanner, which I have dubbed "The Coco Method" (check it out here), we started looking at ways to rig up something that would work with our equipment.

Our approach is a little different to Jonathon's, as we dispatch a single Crime Scene Officer to capture most crash scenes; having two operators (one for the scanner and one for the total station) is not very practical for us. So instead of backsighting each scan, we simply backsight the first scan and then use rotational surface matching for the following scans along with their co-ordinates from the total station. This allows a single operator to move through the scene much faster and to register using a much smaller overlapping scanned area than would otherwise be required for surface matching.

Image
The first scan is backsighted and the scanner positions captured with the total station. The second scan does not have the correct bearing (note the small green scanned surface).

Image
Using rotational registration and the surveyed scanner positions, the bearing can be determined.

We needed to test the idea out prior to having our armourers make up a customised rig, so I present to you my finest proof of concept contraption yet.. guaranteed to make surveyors far and wide shudder in horror.. The FrankenScanner!! :D
Image
Yes that's right, it's a tribrach, a multitrack 360 degree prism, another tribrach, and a super heavy HDS7000 precariously perched on top of it all!

We (carefully) took this bad boy out for some testing and found that the concept worked very well, though the rig as expected was not very stable. As such we decided to run each scan, swap the scanner with the tracked prism, and then zap the prism. Over the weekend we were able to deploy this system to a job and found it to work extremely well. We were able to capture a 200m+ scene very quickly and register the scans rapidly as well. A little tweaking of the survey data allowed Maptek I-Site Forensic to read it and we could then place the scans in their positions before rotating and registering them using smaller areas which would not have been at all suitable otherwise. We achieved RMS values of 0.002m by focusing the registration algorithm on regions such as poles or signs located around the scene.

Image
Scene overview.

Image
The three small areas focused for registration between these two scans (distances along the green lines).

The next step will be to compare these scans with some control survey targets to determine accuracy and to examine the effect of scan noise and range on these smaller areas as they relate to the registration accuracy.

And of course thanks to Jonathon Coco who I believe started this whole idea!
User avatar
jcoco3
Global Moderator
Global Moderator
Posts: 1724
Joined: Sun Mar 04, 2012 5:43 pm
12
Full Name: Jonathan Coco
Company Details: Consultant
Company Position Title: Owner
Country: USA
Linkedin Profile: No
Has thanked: 70 times
Been thanked: 157 times

Re: Targetless registration with HDS7000 and "The Coco Metho

Post by jcoco3 »

:D This is great Domenic, and from the looks of your rig I can tell that you are as crazy as I am when it comes to finding solutions that take the pain out of scanning large areas. Very interesting idea to use one scan as the orientation for all scans. We did something similar once just shooting the rough initial position for some outdoor scans then forcing their orientation to the indoor cluster using Scene's C2C. There were limitation of course, but it worked well enough. I still wish their was a better way to mix Targeted and C2C in Scene, if only for this one application. I have still been working on "things" and will have some neat stuff to share in future. Looking forward to seeing some more on your setup too.

Oh by the way, you might want to take a really good look at that active prism you are using. A bright young surveyor I know found some issues with ours, and of course gave Trimble some great feedback. Center it on a tribrach, lock on to it with a total station taking continuous distance readings, all while slowly rotating the prism on the tribrach's center. You might find that the distance readings are different depending on the rotation(or not), but it is worth confirming. We are currently testing the newer version of the active prism, and so far it works better although it has an entirely different set of issues.

Thanks for the shout out too...owe you another beer now :lol:
Andrej
V.I.P Member
V.I.P Member
Posts: 264
Joined: Wed Nov 07, 2012 2:42 am
11
Full Name: Andrej
Company Details: Meixner Vermessung and Freelancer
Company Position Title: Project Manager
Country: Germany
Linkedin Profile: Yes

Re: Targetless registration with HDS7000 and "The Coco Metho

Post by Andrej »

Hi

I would mount the 360 on top of the scanner if you don't need any information above. Also you can use the leica 360 which you screw under the tilt and turn targets. However I would use a sphere as you don't need to turn it to the scanner.

I normally work alone and just done a job last week with the HDS7000 + Nodal Ninja + GPS + Total Station survey 30 high resolution scans in 5 hours. I use self made half spheres which I can place on a mark on the ground like a Nail or pen mark. This mark I then survey conventionally.

All the best
You do not have the required permissions to view the files attached to this post.
User avatar
d.raneri
I have made 40-50 posts
I have made 40-50 posts
Posts: 45
Joined: Thu May 02, 2013 11:47 pm
10
Full Name: Domenic Raneri
Company Details: NSW Police Force
Company Position Title: Crime Scene Officer
Country: Australia
Linkedin Profile: No

Re: Targetless registration with HDS7000 and "The Coco Metho

Post by d.raneri »

My thoughts are that by using a single scan for orientation and then surveying the scanner positions that we will heavily reduce any compound error over a large distance through surface matching. And considering the speed benifits this approach offers us it seems more than acceptable thus far!

Thanks for the heads up, I will have our guys take a look at the prism and see if there are any problems with it. And after the "happy hours" in SD I'm pretty sure we owe you the beers! We brought back a bottle of Tio's and have been lighting up the mules ever since ;)
User avatar
d.raneri
I have made 40-50 posts
I have made 40-50 posts
Posts: 45
Joined: Thu May 02, 2013 11:47 pm
10
Full Name: Domenic Raneri
Company Details: NSW Police Force
Company Position Title: Crime Scene Officer
Country: Australia
Linkedin Profile: No

Re: Targetless registration with HDS7000 and "The Coco Metho

Post by d.raneri »

Hi Andrej,

Yes a prism on top of the scanner would be best, however we don't have the equipment on hand to do so precisely. That sounds like a lot of gear to be hauling around on your own! I'm guessing you are moving about with a vehicle? Looks like a pretty big survey! We wouldn't be able to drive through a crash scene and so being able to carry all of the gear on foot is a requirement for us.

What sort of distance from the scanner do you need to be for acquisition of a half sphere? I have 3D printed some spheres for trajectory rods, and though I only had a few minutes to test this out I wasn't able to model them accurately in Cyclone.

I'm interested in your Nodal Ninja workflow, are you producing your panoramas with PTGui? Do you have any panorama levelling issues? We always need to include vertical control points and where they aren't available the panorama curves slightly and the alignment accuracy is reduced. You also seem to have a very consistent colour cast between nodes. Is this from consistent lighting throughout the day or have you worked around this?
Andrej
V.I.P Member
V.I.P Member
Posts: 264
Joined: Wed Nov 07, 2012 2:42 am
11
Full Name: Andrej
Company Details: Meixner Vermessung and Freelancer
Company Position Title: Project Manager
Country: Germany
Linkedin Profile: Yes

Re: Targetless registration with HDS7000 and "The Coco Metho

Post by Andrej »

Hi Domenic

In this case and most other jobs I only use the Car to get there and work by foot as I believe you are quicker in this. You are constantly moving and have to think 3 steps ahead otherwise you don't get it done in this time. To be fair I didn't move or placed targets so I have 3+ in each scan, I had at least one per scan as this job didn't required higher accuracy.

For modelling or registration with Spheres a TOF scanner is batter as their point size doesn't get so big. Then I was using the VZ400 years back, sometimes having only 10 points on the sphere still gave me a reasonable result in my registration. With the 7000 I will ignore spheres which are to fare away as the modelled centre is quite out. My half spheres a 217mm diameter and very solid. With a TOF scanner I would do up to 40m however with a Phase Based Scanner not more then 20m.

My friend and I have created a CAD sphere model with a prism inside however we don't have excess to a 3D printer which prints larger then 100mm diameter and the costing we got was $600 for one prototype. So we dumped the idea quite a while back.

Yes I am using PTGui. I have got a program which sorts all photos automatically to the right scan station and then have set up a template in PTGui for automated batch processing. I still need to do some manual post processing as the sky and clouds give me lots of trouble control points. Using HDR would give us batter results however it's taking to much time in the field with camera we have. The most issues we have regarding colour is the photo which is direct to the sun as it get sometimes really dark. Most of my work is in the field at day time or under relatively good light conditions. I am not understanding panorama levelling issues and vertical control to create better panoramas. As said most issues I am having is not good control points. One automated work flow only requires that the first station is done correctly and this will be applied to all other stations as long the photos there taken in the same order and same angle.

Andrej
User avatar
3DForensics
Honorary Member
Honorary Member
Posts: 1979
Joined: Mon Aug 03, 2009 1:52 am
14
Full Name: Eugene Liscio
Company Details: AI2-3D Forensics
Company Position Title: Owner
Skype Name: eliscio
Location: Toronto, Canada
Has thanked: 13 times
Been thanked: 70 times
Contact:

Re: Targetless registration with HDS7000 and "The Coco Metho

Post by 3DForensics »

Great posts guys. We want to use this workflow at our IAFSM conference in Orlando this November. I think it has a lot of merit. Thanks for sharing this work Dom!


Sent from my iPad using Tapatalk
Post Reply

Return to “Terrestrial Laser Scanners [TLS]”