In order to sell grain lots in European countries, seed industries need a government certification. This certification requests purity testing, seed counting in order to quantify specified seed species and other impurities in lots, and germination testing.
These analyses are carried out within the framework of international trade according to the methods of the International Seed Testing Association.
Presently these different analyses are still achieved manually by skilled operators.
Previous works have already shown that seeds can be characterized by around 110 visual features (morphology, colour, texture), and thus have presented several identification algorithms.
Until now, most of the works in this domain are computer based.
The approach presented in this article is based on the design of dedicated electronic vision machine aimed to identify and sort seeds.
This machine is composed of a FPGA (Field Programmable Gate Array), a DSP (Digital Signal Processor) and a PC bearing the GUI (Human Machine Interface) of the system.
Its operation relies on the stroboscopic image acquisition of a seed falling in front of a camera.
A first machine was designed according to this approach, in order to simulate all the vision chain (image acquisition, feature extraction, identification) under the Matlab environment.
In order to perform this task into dedicated hardware, all these algorithms were developed without the use of the Matlab toolbox.
The objective of this article is to present a design methodology for a special purpose identification algorithm based on distance between groups into dedicated hardware machine for seed counting.