Tech note: automatic probe counting — the box material decides the lighting
Transparent boxes count most reliably with backlight; opaque black boxes must be opened and lit from the front; black boxes with a clear film only work with external coaxial light. Key findings for four box types.
Probes are thin and numerous, and manual counting is error-prone. Image counting looks simple, but the biggest variable is not the algorithm — it is the box the probes sit in. Here are the findings from two tests across four box types.
Transparent boxes: backlight is most stable
With a flat panel lighting from below, probe outlines are crisp and counting is most stable. Front lighting also works, but the detected edges are jagged and less complete. To count with front light without opening the lid, probes need to be spaced further apart (only 3–5 pixels apart in our images).
- If the lid carries labels or writing, it must be opened
- No writing on or near probes; consistent orientation improves accuracy
Opaque black boxes
Backlight cannot pass through, so the lid must be opened and a front light used.
Black boxes with a clear film
The film reflects, so only external coaxial front light works, and probes must be aligned consistently for uniform colour in the image. Large dust particles inside also disturb counting.
Upright-probe boxes
Expected to use external coaxial front light, with fixtures designed around the probe dimensions.
System planning
The front light must be mounted on the moving camera module, so payload needs checking first. Locating individual probes with a low-magnification camera only suits transparent boxes; for others the box area is located first and then imaged as a whole. Shin Shou's probe counter integrates operator permissions, counting screens and custom reports.
