every potato is unique, randomly generated by nature, and its appearance (todo: internal appearance via x-ray) combined with mass and volume gives it a truely unique "fingerprint" dataset, that can be used to universally identify it among millions of other potatoes.
I believe this hypothesis is mostly true, or true enough to be demonstrated, if not practically employed.
A handheld scanner for potatoes could be simply done with consumer smartphone hardware and an app. An external (bluetooth) mass scale could be employed for accuracy, but volume and mass calculations could be done entirely in software.
-
there's no reason to "ship" each potato's ID with it individually -- the UIDs are database side and don't need to be literally attached to each potato itself
-
any lot or container of potatoes could be associated (in the database) with the potato UIDs, and their combined checksum as the lot/container's UID
-
pigeonholing: how does git not run out of hash space, how do I avoid it? hash( hash(a) ++ hash(b) ++ hash(c) ) ==
hash( hash(a') ++ hash(b) ++ hash(c)) -
error-correction and data-filling analysis allows a potato's identity to be recognised even after damage / cutting / breaking
-
(extra credit: sequential identity tagging equipment on harvesting equipment could allow tracking dirty potatoes to the square foot from which they were harvested)
-
stitching the surface images isn't necessary for proof-of-concept [ed.: source??]
non-biolab-grade data collection equipment may mean that some potato uniqueness is not captureable, even though it physically exists.
IOW, every potato is necessarily unique, but this does not guarantee that the digital quantization of the potato will preserve that uniqueness.
(later:
- correct lot location is not part of a potato's identity
- lot location history may be part of a potato's record, but must be done carefully
- identifiable potato features:
- evident (physical / "obvious"): physical source, timestamp first seen, volume, mass, dimensions and/or heightmap
- emergent: actual density, variety, size class (C/A/B/E S/M/L), grade (1/C/2/3)
- checksumming should be enough to ensure uniqueness
- checksumming the potato's surface could be combined/auxillary with checksumming the evident record (because the surface image is evident but not a single number),
- but a serial# could just be used to refer to the entire record (evident+emergent)