This week I mostly worked on making a shareholding script to isolate the scanned object. We realized that, especially with the acrylic cover sheet, it’s impossible to center the object on the rotary arm. Since we’re still guaranteed that the scans are 90 degrees apart regardless of the center of rotation, we can pretty trivially use thresholding to find the bounding box of the object and align them in x/y, then rotation adjust as planned. This shouldn’t be too challenging but we have to overcome the fact that the manipulator legs are visible, so we need to crop them out somehow. Another possible approach would be to take the arithmetic mean of the four images, find the bounding box of the mean of the four objects, then take the center of rotation as the center of that bounding box.
Next week we will test both threshold approaches and proceed from there. I also still have to print and scan the test object, but it’s already designed so that should be quick.