Prototype Design of Variety Discriminator of Farm Products Based on Multi-color LEDs and BP-ANN

A new optical instrument for the variety discrimination of farm products was designed and fabricated. Multi-color LEDs were used as light source, Vis/NIR spectrometer as light detector, and optic fiber as light transmission medium. In the paper, the principle of using multi-color LEDs for variety discrimination of produces was first introduced. Then, the method of using error back propagation artificial neural network (BP-ANN) in the modeling of optical data was elaborated. Reflective light intensities of Multi-color LEDs were taken as the incoming signals to BP-ANN. The structure of BP-ANN with three layers has been optimized to minimize its calibration error. In the test, total 210 samples of three varieties of fragrant mushrooms were examined. Among them, 150 samples were picked randomly out as for model-calibration and other 60 for model-verification. With 52 samples judged correctly, variety discrimination rate reaches 86.7%. Finally, a universal variety discriminator of produces based on microprocessor MSP430 CPU was illustrated. The result showed that the new kind of optical instrument integrating multi-color LEDs with BP-ANN is promising in farm products information acquisition.