Remote Seed Identification 

Name: Dan Curry
Affiliation: Crop and Soil Science Department, OSU
Phone: 541 207-4247
Knowledge Required: Must be proficient with how an iPad functions and be familiar with the necessary programming languages to identify seed with machine learning. Must be able to put components together to develop a working model that stores and delivers a final document that accurately reports the identification of 2500 seeds.
Motivation: Currently, an Oregon grass seed cleaning facility begins cleaning a new seed lot and takes a sample of the clean seed from the machine and physically sends the seed over to a seed laboratory for analysis. This takes much time and is not very efficient. Sometimes the cleaning machines are shut down, to wait for the results of the analysis.

This new concept allows the grass seed cleaning facility to take a picture of 2500 seeds (5 grams), spread them out on a flat surface, take a picture using an iPad or other camera device and upload the image to a processor, with results coming back to the seed cleaning facility within minutes. These results inform the operator how many off-type seeds are still in his clean seed. The operator can then make the necessary changes to the cleaning machines to improve the quality of the final product. If the results of the image analysis indicate the seed is free of weeds, other crop seeds and inert material, the operator can continue to run the cleaning machines as previous set, which saves money for the grass seed grower, since less good seed is discarded.
Description: Seed Identification using an iPad to take a picture of 2500 seeds, send the image to a Jetson TX2 processor, program a computer using machine learning to identify the seeds, mark the off-type seed on the image, store the results in a database and email a final PDF of the results to a client.
Objectives: To use the latest equipment available to develop a system that will speed up the transfer of information in Oregon grass seed testing.

Deliverables: A working machine that can accept images from electronic cameras, identify each seed, label off-types and send a report via internet and store the results.
Other comments: Feel free to call me at my office on campus: 541 737-5094 or by email:

   D. Kevin McGrath
   Last modified: Tue Mar 6 10:29:08 2018