How do I scan this object

[grukx]

How do I scan this accurately(mini2 on PC) ? I have tried a “million” times. Originally black, but after having used scanning spray too many times I just gave up and spray painted white it has become a bit of an obsession to get this right for some reason it is impossible for the scanner to see any features and I always end up with shifts. I think I have tried all tricks I have found except adding features to the object itself. I really don’t understand what is considered “features” by revoscan.

It is quite thin, 5mm at the thickest and I am trying to get a full scan. Scanning flat with markers works, but aligning both sides is a bit hit and miss (using cloud compare). Scanning standing up, on short/long side mostly end up in tracking loss, despite adding items such as small tracking cubes. I am considering 3d printing a tracking tower to see if that helps. I have tried adding paper, bits and pieces of varying height next to the object etc. with only marginal success.

I cannot understand why the scanner is incapable in seeing any “features”.

Too be honest I am quite disappointed in the mini 2 scanner so far, unless you are scanning trivial objects it seems to be quite finicky and temperamental, as well as sensitive to lighting conditions and object shininess/color/color-contrast. So far my attempts with photogrammetry and Reality Capture has proven to be far more successful. Is it just me that is having this user experience? I would have thought that the software would be quite a bit more advanced at this stage especially with regard to object tracking.

Geir

[PUTV]

Hi Geir

Features are points that are not symmetrical , and at least in proper distance to each other depending on the FOV .
Your object is just not the right one for the software recognition , no features and symmetrical form .

You have 2 options , adding additional objects for stabilization , or use magic matt and market mode … on the other hand this object has almost no volume , I suggest you scan well one side only and use thickness in your 3D software to finish it … because scanning both sides and merging will not works here .

If you don’t have magic matt just stick some markers on the ground where you scanning like black garbage bag will do it and use marker mode function to scan it .

[grukx]

Hi PUTV,

I will probably end up doing it this way, as I have tried all of the other suggestions.

One thing I really would appreciate is some writeup from Revopoint detailing how the feature detection algorithm works, with examples of good and bad features, as well as how FOV (scanner distance) affects the feature detection. Trying to figure out what works and what doesn’t work by blind trial and error is not very efficient or user friendly.

Another option, and possibly better, would be to add some visual indication in UI of the detected features. Just a small bounding box around what is considered the features used for tracking at a given moment.

Geir

[PUTV]

For a new users definitely a little struggle understanding how the device operate , but that is with all 3D structured scanners on the market that relays on features … How closer you getting to an object how more dense the features need to be from each other .

Imagine a mountains and topography , with all the lines on top, that’s. how the hardware sees its object, not by taking photos like photogrammetry do but by scanning the distorted lines , and what do distort the frames ? objects features
the software keep tracking once it detect the anomaly from one feature to another . The feature distance should be from 1 cm/ 10mm apart for MINI series to 8 cm /80mm apart for MIRACO far mode and Range series .

[bau-3d.ch]

features are hard edges and angles acting as destinctive geometrical patterns. i have 3d printed white dodecahedrons to place my item onto for scanning to add a easy to detect geometry the scanner could grab onto. also i have black ones with markers on them, to have a 360 degreees marker base for the scan.

print two of those at a fitting size and use them as a base below the sunglases, place the glasses standing on a edge using some black putty so the scan is 1. hold securely in place 2. devided from the base. flip over and make a second scan to merge

[grukx]

Thanks for trying to explain this to me PUTV, very much appreciated. Apologies for asking some more “dumb” questions to get to the bottom of this.

• A feature is any distinct point cloud deviation in the topology inside a certain radius? Do we know what this radius is?
• 10mm is a minimum feature distance
• Do X number of features have to be in FOV at once? If so what is X? I guess this will be something like 3 to provide ability to move scanner without loosing track between frames?
• Is the “shape” of a feature used to distinguish it from other features in FOV? Or is it the constellation (relative placement) of visible features that is tracked? Or both?

Regards,
Geir

[PUTV]

@grukx

A feature is any distinct point cloud deviation in the topology inside a certain radius? Do we know what this radius is

The radius is not important but the distance is , the distance between at least 3 features depends of the FOV you are using , each scanner uses different FOV
So in your case using MINI will be the shortest distance to maximum 3 cm /30mm between features , remember at least 3 features are needed to be visible at time once while scanning .

@grukx

10mm is a minimum feature distance

The minimum don’t really matters , how more features at shorter distance how better in your case even if they are just at 5mm or less but available , for example while scanning a coin .

@grukx

Do X number of features have to be in FOV at once? If so what is X? I guess this will be something like 3 to provide ability to move scanner without loosing track between frames

For a stability you need minimum 3 of them at a time for greater stabilization , over time you will just learn from practice .
Remember everything that destroys the flat surface of the pattern projected on your object surface is a feature .

@grukx

Is the “shape” of a feature used to distinguish it from other features in FOV? Or is it the constellation (relative placement) of visible features that is tracked? Or both?

The shape don’t matters , the distortion of the pattern do , while you scanning the tracking switching from one feature to another keeping the frames aligned , while adding it like a shingles on a roof to build the point cloud surface .
Scanning with MINI need more practice because of the smaller FOV where you need to have at least 3 features available at any time while rotation .

To make it more clear for you to understand how the sensor works , get your object with some light and project the shadow on a wall … if you take a ball and project the shadow on a wall and rotate it , the shadow will not change , it will be always round no matter the rotation , object like that is featureless to the sensors .
Now you take a small cube , same thing from a front point it will always be the same , however if you move it at 45 degrees it actually change while rotation as the sensor can see more than 2 features , now it can see 3 features moving … recognizing it as 3D object .

So that is how it works … if you are not sure about your object features just add some additional bits , you can purchase them on Amazon or print them yourself … since you use MINI I recommend you get smaller bits or pyramids ( as long not made from resin or dark plastic , preferably white plastic or wood)and this will allows you to scan anything you want without even to think about .

And there are not dumb questions Geir … who asking for directions, don’t get lost … so don’t hesitate to ask any questions you may have .