Scanning a Motorcycle


#1

I recently picked up a Sensor with the intent of using it for my business. I build custom motorcycles and have a background in Industrial Design, so I often will design new parts in CAD before either making them by hand or having them prototyped.

Having a scanner that could give me accurate dimensions would save a TON of time, so this thread will be about figuring out if that level of accuracy and detail is possible with the Structure Sensor.

I’ll be starting off using the Structure Sensor and Skanect Pro on my Surface Pro 3 with i7-4650U 1.70/2.30 GHz. It’s not an ideal machine for this I’ve found out so I have to really take my time when I scan.

I’ve only scanned a handful of things so far so I’m still learning. Here’s the first bike scan I did.

Specs:
Bike surface as-is (no scan powder, but a little dirty)
Scan feedback: Med
Scan time: 20min
Reconstructed at V. High

My goal is to use this scan to plan out a custom exhaust so I concentrated on the lower right side of the bike.

I haven’t yet pulled it into Rhino so I’m not sure how accurate it is. I tried scanning a set of triple clamps earlier (the clamps that hold the front forks on the bike) and the accuracy was pretty poor. The 55 mm ID clamp holes came out pretty ovaled and wonky.

My stage 2 and stage 3 will be covering the bike in alcohol/baby powder and then trying out the 4 eyes case with magnifier lenses from Thor Labs. I’m hoping that will add some detail. In it’s current state it’s not really usable as a foundation to design parts from. At best it’s good for designing things like fairings and exhausts, aka parts that are kind of floating around the bike.

Ideally I’d be able to use the scans to pull mounting hole dimensions and redesign parts like footpegs or engine covers.


#2

The scan looks pretty good for the sensor to be honest. It looks like a ducati V2 engine, did you take out the covers for the timing belt? Not to shatter your dreams, but I don’t think you will get the accuracy you need for motorcycle parts, if you are strap for cash you might want to look into the shinning3d scanner (it will still cost around 3K to 6K depending on the model). Another option is to try and implement a white paper that suggest using a polarized lens with with different orientations and fix the point cloud… but it is very research oriented and I have yet to see anyone replicate it, heavy on the custom programming side too.

However you can still use the cloud for design as it does seem to provide a starting point. Where are you located?


#3

Yup, it’s a Ducati 1000DS motor in a Sport Classic with the belt covers removed.

Can you explain the white paper thing you’re talking about? I’m not familiar. Is that like putting a standard object in the scene for reference?

In the end I may throw down for something like the Einscan because I feel like I’d use it on a daily basis. I thought it was best to see the limits of the consumer grade stuff first since the price gap is so extreme.

Is there really nothing between the $500 and $3000 marks? I see there’s an Einscan SE desktop unit for $1300ish. Has anyone turned that into a handheld?


#4

I have a 2006 800SS and do all the work myself, just looking at the scan reminded me on changing the belts.

The white paper, is actually a research paper that describes how to map the values you obtain from a canon DSR with a polarized lens. You need to compute all the values of the image and correlate them to fix the point cloud. I have not tried this, but the kinect point cloud and the one from the structure sensor is the same. But there are a lot of factors on this, and I am not sure how well it actually works and you would spend countless hours investigating this.

As accuracy goes up, the prices goes up almost exponentially. Also you can’t just turn it into handheld scanner. But the best way would be to take the scans, with overlap between each scan, and then align them and merge them. It is a manual process, but it is a solid working one.