基于近红外光谱技术的小麦专用粉品质特性快速检测研究OA北大核心

Based on near-infrared spectroscopy technology,combined with different preprocessing and characteristic wavelength screening methods,partial least squares(PLS)prediction models and overall prediction model were established for indicators such as damaged starch content,falling number,water absorption rate,stability time,stretching area,extensibility and maximum resistance.The results showed that detrend(DT)was the best preprocessing method for the prediction model of damaged starch content and water absorption rate,savitzky-gloay(SG)convolutional smoothing was the best preprocessing method for the prediction model of falling number and stretching area,and standard normal variable transformation(SNV)was the best preprocessing method for the prediction model of extensibility and maximum resistance.Competitive adaptive reweighted sampling(CARS)could effectively improve the prediction accuracy of models for damaged starch content,falling number,water absorption rate,stretching area and maximum resistance,with prediction determination coefficients of 0.964 1,0.714 0,0.975 5,0.943 4 and 0.828 3,respectively,successive projections algorithm(SPA)had improved the performance of stability time and extensibility prediction models,with prediction determination coefficients of 0.713 5 and 0.953 0,respectively.The overall prediction model had improved its predictive performance for stability time,stretching area and maximum resistance,their residual predictive deviation increased from 1.86,4.27 and 2.51 to 2.43,5.26 and 3.11,respectively.In summary,near-infrared spectroscopy technology was effective and feasible for a non-destructive and rapid detection of the quality characteristics of wheat flour.