For those of you who are using C2C (Cloud to Cloud) with Scene, many people (in the construction industry) I encounter are not using it correctly, and are getting results that don’t quite match up with external verification (total station control). This post is to let people know the top 3 pitfalls I see people run into when using C2C and what you need to do to get truly get accurate, repeatable results that will match up very tightly to project control in vertical construction/building projects.
1) Always visually QC your scans. What does this mean exactly? It means creating “scan point clouds” during your preprocessing step so that you can actually look at surfaces and see if walls and floors have ligned up.
Do not blindly trust the scan result manager reports; only use it as a guide. It may say you have .0014m (1.4mm) of accuracy and very small errors, but here’s an example of scans that say they have come together to <2mm and yet we see on the floor surface a deviation of 12mm
I encourage all Scene users (who are in the construction industry who actually care about achieving 1/8″-1/4″ global accuracy) to re-visit scan projects that they’ve registered using C2C, and visually look for these gaps. Never consider registration finished unless you visually verify it. There have been countless times where I’ve been able to fix busts for people’s scan projects by simply catching a gap that Scene didn’t report on.
This leads me to the next thing:
2) Understand that probably the MOST important setting to change in C2C is the “maximum search distance”.
Above is the result we get when we make the search distance the default (32.5 feet or ~10 meters which in my opinion is much too HUGE) during C2C (after rough placement using top-based registration of course)
Now, here is the result when we make the search distance much smaller (.06′ or 9mm)
Now we’ve fully closed the gap!
Most people I’ve talked to have gone years with never changing this setting, but once you do, you’ll notice that your overall survey control starts to hit very closely! This is because you vastly reduce error propagation across scans.
Note: the secondary important thing to adjust is the subsampling. I tend to keep this high and have gotten better results than keeping it low. Statistically this will favor scan overlap on the perimeters of scans which will tie down errors; on the other hand, making subsampling too low will result in disproportionately favoring points closer and between scans which can cause faster error propagation across a project, which we’ll want to avoid.
3) Understand that C2C is an “iterative” approach, and the goal is to “chase” down the accuracy. You may need to run C2C more than once. Run it first with a 1′ search distance, then bump it down to .1′, then .06′. Your gaps will typically decrease and the disappear, and that should always be the goal before moving onto the next cluster.
This is part of the typical advanced Faro Scene training program I give to my clients in the construction industry who have been able to consistently register large construction projects (hospitals, universities, high-rises, etc) much more accurately. The techniques continue to evolve and be perfected as we learn better ways of doing things.
If you are interested in a custom-tailored Advanced Scene Training, please look into our the “Minimum Viable Workflow” training program in the products page!