Research On Key Techniques Of Near-infrared Spectroscopy And Development Of Related Application Software

In order to improve the stability of calibration model, with the 160 wheat powder samples'near-infrared spectra, four calibration models based on different regions were establishedusing partial least square regression (PLSR). The models were used to calculate theconcentration of protein and the results were compared with each other. It was found that themodel based on the combination region was the best. From this, some basic principles ofwavelength selection for calibration were concluded.The discrete scanning spectrometer based on the first overtone had a lot of advantage, inorder to construct the spectrometer; Simulated Annealing Algorithm (SAA) was introduced inthe paper as a wavelength point selection method. In the first overtone, four wavelengthpoints were selected using optimized SAA and another calibration model was establishedusing MLR. Comparing with the model established using PLSR, these two models almost hadthe same performance. This was to say, optimized SAA was powerful to locate thewavelength points of calibration for discrete scanning spectrometer and to simplify thecalculation of calibration.Lastly, some key and difficult Techniques of the development calibration andmeasurement software for near-infrared spectroscopy were presented. In addition, someuseful solutions and the original software were exhibited in the paper, it was very importantfor doing a deeper investigation about the development calibration and measurement softwarefor near-infrared spectroscopy.