Waag Society Fablab Amsterdam

Fabmoment 3D-Scanner
Splinescanner
Miguel & Oscar

Currently we are busy testing the 2nd 3D-scanning software called the Splinescanner. The setup of the scan is almost the same as the Davidscanning software. The main difference is that the reference of the scanner is done by the align-laser itself instead of implementing a 90˚ reference point framework, and that the scans are made by a series of photos instead of 1 stream. The advantage of the laser reference is that there is no limit to the bounds of your scanning object (except by the measures of the platform used) and the fixed laserline, which generates the optimal reflection dots on the object.
By generating a series of images instead of using a stream the scanner has a higher resolution, and therefore a better detailed result. In this setup different types of cameras can be used, which is nice for the open source and low-cost purpose of making our 3D scanner.

Creating the Setup for the Splinescanner
To make a scan with the Splinescanner software you will need an alignment laser, a rotary platform, (photo) camera and a dark room. We used the same camera (Sony 3CCD Handy Cam) and laser (5V alignment laser) as the Davidscanner software, but we had to build the rotary platform and the dark box.

Making a rotary platform
First you need a strong unipolar stepper motor which is capable of handling your average scan objects weight. Second you will need a circular platform with a centered hole which matches the axis of the stepper motor (we used acrylic plate with the laser cutter). Next you will need to plug the motor correctly to an arduino, to do so you will need the stepper motor driver schematic, which is provided by the deliverer or online and a 12V battery. Then you write a slow turning script for the arduino.
 

Setup Rotary Platform

 
Stepper Motor Schematic

This is how our rotary platform looks. 

Creating a dark scan space
We created a temporary solution because we did not wanted to waste to much time with this in the testing phase. We took out a cardboard box which fitted our minimal range for the object-camera distance, and hung up a black curtain all the way around it. At the bottom side of the box we drew 2 lines, with a 26˚ angle between so we can align our laser and camera with the object. The result works fine for our tests, but when you want to optimize your scanning process it is best to use a real dark room.

This is the total setup needed to test the Splinescanner software.

Scanning with the Splinescanner software
When you are ready to test the 3D-scanner you should first get an object which has no black colour in it. The laser points will not reflect well on the black surface, therefore the part will not be taken into the scan.
The Splinescanner uses a different method of scanning compared to the Davidscanner.

•    First you make a short footage of only the align laser in the center of the rotary platform (best to use a flat object to create a precise surface). At this point the rotary platform should be still. This is to create a reference for the program.
•    Second you make a 360˚ scan in full light, laser off. It is better if the background is plain, so the object comes out a little in the footage. This scan is made to fill in the colors later on. The scanned object needs to be fixed on the platform.
•    The third scan is similar to the second. Don’t move anything of the setup between these 2 phases. Next you create the dark scanning space (in our case, close the box and hang the black curtain). And make another 360˚ scan with the laser on.

After these 3 steps you basically have all the data you need to create a 3D image of your object. Next thing is to convert the files so that the software is able to render it. The Splinescanner software takes a series of scan images in daylight and matches them with the scan images of the laser in total darkness which it refers to the centre reference laser image.
With our scans the output files of the camera were MPEG, and we needed JPG images for the software. Therefore we used a free video to JPG converter.
Since the Splinescanner software is still in the starting phase, it only generates an output when you name the images right.
Currently there is only 1 input format for the images which are:

-    RightPic # (starting from  the first image you created in daylight till the last)
-    RightRed # (starting from 1 image of the centred laser, continued by the first image of the laser in darkness till the last)

 
Miguel created a script to convert these files automatically to the right name (ask for it if you are using the software miguel.jgz@gmail.com).

Using the Splinescanner software
Now you can run the Splinescanner software. First you have to select the first input image which is best to use the centre laser image (Files Right Camera). Also select the output directory for your result (Save File Directory).  Next you select Show Picture, click on the image to see the file.

Then you select the bounding box of the laser, with 2 right clicks on the corners of the image. Next you select Find Laser and the software generates a blue line right where your laser is. By selecting Update Text the software calculates the dots on the line. Naturally you want as many dots as possible, try changing the red colour bar to generate more.

Finally you select Geometry and the scanning process starts. The Splinescanner places the laser images on the object images, and matches the point locations and colours of the object with the laser lines (yellow line is the laser image). Therefore this software is able to not only render a detailed 3D image of the object, but also add the real colors to it.

Finally you will have 2 results in the selected output directory, a PLY file and a WRL file. The WRL file is the 3D image with color. To open this we use Meshlab, which is an open-source meshing software. 

The scan is generated by a lot of dots, which you want to connect in order to make a precise mesh. To do so you have to select: Filter – Remeshing, Simplification and Reconstruction – Ball Pivoting Surface Construction. From there you can input a value which resembles the max radius between the dots that will be linked. The more dots you have generated, the smaller this surface has to be in order to link them. When you take a large radius the 3D-mesh will turn out bad because it links points that were not mending to be combined.

After this you are able to render smooth meshes of your scan. Here are some of our results.

Some parts of the sheep object morphed into each other. Also the black colored surfaces are converted to gaps in the mesh. Now that we understand the Splinescanner we will try to optimize the 3D scanning process by using better components. Our Steppermotor did not suit our needs, because it was too weak to keep the object in a steady 360˚ turn. When this happens points will turn into each other and this causes our morphed sheep.
To solve the gaps we want to try using a better (stronger) laser so the dark surface won’t absorb the laser light. 
More results (better ones) are coming soon!