Spud Cutter

Potatoes aren’t like some other plants. There are no tiny seeds that come inside a pod. Instead, the potatoes themselves are the seeds. Rather than plant entire potatoes, however, it is much more efficient to slice them into pieces. That way one potato might give you several “seeds” rather than just one. The tricky thing about that is, though, if you cut the pieces too small, the potato chunk isn’t a viable seed because there’s not enough energy in the chunk to get the potato plant sprouted and established. There’s a lot more to it, but I’ll let you read more about that if you want. What we care about here is what the problems are with current potato cutting methods. Most modern potato seed cutting machines feed the potatoes through a set of blades that are a fixed width apart. The problem is that this often slices off the ends. It leaves a high number of chunks that are either too small to grow, or don’t have eyes on them (no eye, no sprout).

imageProcessingResults1Cropped
The trick to improving this scenario is to build a machine capable of engaging or disengaging cutting knives depending on the size and position of the potato coming through. Once such a machine is built, however, there is also the challenge of detecting the size and position of the incoming spuds. That gets to my part of this project. For a proof-of-concept, I used a usb webcam to take a picture of a row of potatoes, and then wrote a script to segment out the potatoes and estimate their size and weight. The size and weight (volume, actually) was estimated based on only a 2-D picture. I used an ellipsoid with three radii. The first two can be estimated directly from the segmented image. The third was inferred based on the typical aspect ratio of a sample of potatoes. The volume of the ellipsoid was then computed, and the weight estimated based on the typical density of a potato. That was the first phase.
The second phase was to decide, given the size and location of the potatoes (determined in step 1), which knives to engage so that the potatoes in the line are cut in the most appropriate locations.

For more information on what makes a good seed potato, there are several of websites and scholarly publications out there. Here is one to get you started:  http://umaine.edu/publications/2412e/

Simulation Results

spudCutterComparison

Results comparison. Vertical axis depicts number of chunks, horizontal axis represents number of ounces. Notice how the existing cutter has many chunks below 1.5 oz. These are non-viable seeds, and almost always represent a waste of the most valuable part of the seed potato (the skin).

inputSpudsCharacteristicsSubHist

The input characteristics for the random potatoes used in this simulation. The random potato generator used values based on actual potato measurements, with random Gaussian noise added.