0.008mm accuracy with the OpenScan Mini - Tutorial - Mesh analysis using CloudCompare
I scanned a feeler gauge with the OpenScan Mini, processed the data in the OpenScan Cloud and scaled the object with the help of the known dimension. Next, I digitally measured the thickness of each feeler and got a deviation of 0.008 mm from the claimed/real thicknesses (0.05 mm - 0.55 mm). There are several points that will be discussed in this article. But just to give you a perspective, that 0.008 mm is somewhat around a fifth of the thickness of the human hair.
Since the beginning of the OpenScan project, the question of how accurate the device can be had been haunting. I have been very hesitant to even try to answer this question, since I am no metrology expert and my methods might be a bit funky. But I finally convinced myself, that the following measurements and discussion seem at least worth sharing. And hopefully, this could even start a discussion with some more experienced users. Maybe even some other scanner users or developers could do a similar scan with their devices and share the (raw) results. I would even do the analysis :)
I will use this case to show my standard procedure of analyzing the mesh in CloudCompare (which are both free-to-use programs).
The scan object and procedure: Feeler gauge
The feeler gauge is normally used to measure gap widths. I got mine for around 10€ and this should be at least an indicator for the quality and accuracy. The feeler's thickness range from 0.05 mm to 1 mm in steps of 0.05 mm. But I only used those in the range of 0.05-0.55 mm plus the 1 mm piece.
This is probably the ugliest scan object arrangement I ever did, but it is highly functional:
I used some putty to give the object some depth, because the alignment of front and back seemed to be tricky/fail otherwise in earlier attempts
added the gauge block of known dimensions. The width along the blue line (see below) is 50.000(5) mm and accurate to 0.5 microns according to DIN EN ISO 3650
sprayed the whole arrangement with a light sprinkle of Aesub Orange to create enough surface features
used the polarizer module of the OpenScan Mini to filter reflective highlights
took a total of 150 photos with the Arducam IMX519
Raw scan result using the OpenScan Cloud, see and download the 3d model on Sketchfab
Cross-section along the red line (0.005 to 0.055 mm thickness)
the noise in the bottom area of the scan is caused by the photos being mostly blurry in that area (since the object exceeds the 8x8x8cm scanning volume of the OpenScan Mini)
in the area of interest (the feeler tips and the gauge block) have very low noise
you can even see my fingerprints in the putty ;)
Scaling was done in Autodesk Meshmixer and the procedure will be covered in an upcoming tutorial. But basically I took three measurements along the blue line and scaled the object accordingly.