论文部分内容阅读
提出了一种遗传区间偏最小二乘法(GA-iPLS),并用该方法快速提取苹果糖度近红外光谱的特征区域,在此基础上采用遗传偏最小二乘法(GA-PLS)提取苹果糖度近红外光谱的特征波长,进行苹果糖度预测。结果表明,整个光谱等分为40个子区间,遗传区间偏最小二乘法能快速寻找出5个特征子区间(第4,6,8,11,18号);在5个特征子区间的基础上用遗传偏最小二乘法继续优化,从中提取44个特征波长。建立在5个特征子区间和44个特征波长上的偏最小二乘法模型精度均优于全光谱偏最小二乘法模型,对预测集的预测相关系数提高了近10%;且模型得到了很大的简化,用于建模的主因子数减少了7个。这些结果表明,用这两种方法不但可以建立简洁、数据运算量少的模型,还可以快速地提取近红外光谱的特征区域和特征波长。
A Genetic Least Squares Partial Least Squares (GA-iPLS) method was proposed and used to rapidly extract the characteristic regions of near-infrared spectra of apple. Based on GA-PLS, The characteristic wavelength of the spectrum, apple brix prediction. The results showed that the whole spectrum was equally divided into 40 sub-intervals. The genetic interval partial least-squares method could quickly find out 5 sub-intervals (No.4, No.6, No.8, No.11 and No.8). Based on the five sub-intervals Genetic optimization partial least square method to continue optimization, extracted from 44 characteristic wavelength. The accuracy of the PLS models built on the 5 feature subsets and 44 feature wavelengths are better than the full-spectrum PLS model, and the prediction correlation coefficient of the prediction set is increased by nearly 10%; and the model is very large The number of principal factors used for modeling has been reduced by seven. These results show that using these two methods not only can establish a simple model with less data, but also can quickly extract the characteristic region and characteristic wavelength of near-infrared spectroscopy.