I am thinking to estimate the uncertainty of holes photographed from the bottom and top side. The difference of the same hole photographed from two sides is distributed as Gaussian. The length of the plane is $10.00 \pm 0.01 cm$ where are holes distributed hexagonally with $50 \mu m$ difference in distance. I have attributed these factors in affecting the uncertainty of the hole in photographing it in small magnitudes:
- length of the plane
- viewing angle of the camera (by Taylor expansion getting 3.6% systematic uncertainty)
- slicing of holes from the plane
I am uncertain how much can the algorithmic approach of detecting the border of the hole in the image affect the uncertainty of the diameter of the hole. Some steps of algorithmic approach:
- Canny algorithm for edges so allowing finding circles
- circle's diameter can be measured as maximal distance between two points in one circle
The hole plane of 10cm is divided into smaller regions for the algorithmic approach, like the following for one sample where the Canny is applied:
which is a simplification, since the original data is 26.1 Mb (tiff) in a single sample picture.
How can you approximate the uncertainty in slicing a hole from a plane?