激光与光电子学进展, 2010, 47 (12): 123001, 网络出版: 2010-11-17
基于遗传算法的脐橙可溶性固形物的可见/近红外光谱无损检测
Nondestructive Detection of Soluble Solids Content on Navel Orange with Vis/NIR Based on Genetic Algorithm
光谱学 遗传算法 偏最小二乘法 可溶性固形物 脐橙 spectroscopy genetic algorithm partial least squares soluble solids content navel orange
摘要
应用可见/近红外光谱结合遗传偏最小二乘法(GA-PLS),建立了柑桔类水果可溶性固形物(SSC)的快速无损检测模型。应用光纤光谱仪采集脐橙的可见/近红外光谱,其光谱范围为350~1800 nm。把脐橙的可见/近红外光谱划分成15个光谱区间,通过GA-PLS方法,选出5个光谱区间(包含波段446个,对应波长范围为554~643 nm,1000~1088 nm,1089~1177 nm,1445~1533 nm和1623~1711 nm)建立了预测脐橙可溶性固形物的模型。验证组的最佳预测结果为相关系数和均方根误差分别为0.9132和1.2579。实验结果表明,应用GA-PLS方法选出的可见/近红外特征光谱区域,不仅提高了脐橙可溶性固形物模型的预测精度,而且使模型更加简洁。
Abstract
A rapid quantification technique is developed and validated for nondestructively quantifying the soluble solids content (SSC) of citrus fruits using Vis/NIR spectroscopy in conjunction with partial least squares regression (PLS) using genetic algorithm (GA). The spectral are recorded in the Vis/NIR region from 350 to 1800 nm using the fiber optic probe method. The navel orange Vis/NIR spectra data are divided into 15 intervals. Consequently, 5 subsets (wave regions 554~643 nm, 1000~1088 nm, 1089~1177 nm, 1445~1533 nm and 1623~1711 nm respectively) and 446 data points are selected quickly by GA-PLS. The best prediction results for the navel orange in predicted set are 0.9132 and 1.2579 for correlation coefficient and root mean square errors of prediction respectively. With the proposed method, a concise easily computed model can be built to select the characteristic region of Vis/NIR spectroscopy.
薛龙, 黎静, 刘木华, 王晓, 罗春生. 基于遗传算法的脐橙可溶性固形物的可见/近红外光谱无损检测[J]. 激光与光电子学进展, 2010, 47(12): 123001. Xue Long, Li Jing, Liu Muhua, Wang Xiao, Luo Chunsheng. Nondestructive Detection of Soluble Solids Content on Navel Orange with Vis/NIR Based on Genetic Algorithm[J]. Laser & Optoelectronics Progress, 2010, 47(12): 123001.