论文部分内容阅读
为了实现带壳褐变板栗的快速无损分选,研究了基于近红外光谱技术的褐变板栗检测方法。试验在1 000~2 500 nm波段范围内采集带壳板栗的反射光谱,利用竞争性自适应权重法(CARS)筛选波长变量,采用偏最小二乘-线性判别分析法(PLS-LDA)建立褐变板栗的识别模型,并与全光谱变量所建立的识别模型进行比较。结果显示,经二阶导数处理的光谱结合CARS-PLS-LDA方法可以有效识别褐变板栗,所建模型对测试集样本的敏感性、特异性和识别正确率分别为0.85%,0.90%和88.73%,优于未经变量优选的识别模型。筛选的波长变量个数为224个,仅占全部变量的14.39%,不仅降低了模型复杂度,而且提高了识别准确率。研究结果表明,近红外光谱技术结合CARS与PLS-LDA可以检测带壳褐变板栗,为板栗的快速自动分选提供了理论依据。
In order to achieve rapid and non-destructive sorting of shell-shorn chestnuts, the browning chestnut detection method based on near-infrared spectroscopy was studied. The reflectance spectra of chestnuts collected from 1 000 to 2 500 nm were collected, and the wavelength variations were screened by competitive adaptive weighting method (CARS). PLS-LDA was used to establish the browning index Change the chestnut identification model and compare with the recognition model established by the full-spectrum variable. The results showed that the second derivative treatment combined with CARS-PLS-LDA method could effectively identify browning chestnut. The sensitivity, specificity and recognition accuracy of the model to the test sample were 0.85%, 0.90% and 88.73 %, Better than the variable-free recognition model. The number of wavelength variables screened 224, accounting for only 14.39% of the total variables, not only reduces the complexity of the model, but also improves the recognition accuracy. The results showed that near infrared spectroscopy combined with CARS and PLS-LDA could detect browning chestnut shell, which provided a theoretical basis for the rapid automatic sorting of chestnut.