农药活性成分的快速测定已经成为农药质量监控的一个大趋势。 通过融合甲维盐制剂近红外和中红外得光谱数据, 旨在用数据融合的方法建立一种快速可靠的测定甲维盐制剂活性成分的方法。 采用了将偏最小二乘回归法与数据融合相结合, 以及用竞争自适应重加权采样法来选择偏最小二乘回归中的有效变量的方法。 与近红外和中红外各自建立的模型相比, 数据融合在吸取了近红外光谱和中红外光谱相互补充的信息后, 具有协同效应的模型效果有了很大的提高。 同时, 证实了竞争自适应重加权采样法在建模过程中是一个使得模型更加简单高效的有效的变量选择技术。 研究结果表明在吸收了不同来源的多种信息之后的数据融合是一种能提高模型效果的很有效的建模方法。 数据融合策略的可行性使得测定低浓度(0.1%～1.0%)样品能获得更好的结果, 而且结合了变量筛选算法的对近红外和中红外光谱的数据融合, 是一个很有前景的测定商业农药制剂中有效成分的方法。 最后建立了一种基于近红外光谱和中红外光谱数据融合来测定商业甲维盐制剂的有效成分的方法。

Rapid determination of pesticide active ingredient has been a trend in pesticide quality control. In this paper, we aimed to use data fusion strategy to develop a rapid and reliable method to determine the active ingredient in the emamectin benzoate formulation by fusing the information of NIR and MIR. Data fusion strategy combined with partial least squares (PLS) regression was applied. Competitive adaptive reweighted sampling was engaged to investigate effective variables in the PLS regression. Compared with the models established by independent NIR or MIR, there was a significant improvement provided by data fusion strategy, which benefited from the synergistic effect of complementary information obtained from NIR and MIR spectra. In the meantime, CARS was proved to be an effective variable selection technique in the modeling process that makes the model simpler and more efficient. The results in this work showed that data fusion is an effective modeling strategy that improves the model performance by utilizing more information from different sources. The feasibility of data fusion strategy can obtain better results in determination of low concentration samples (0.1%～1.0%), and data fusion of NIR and MIR spectra combined with a variable selection algorithm could be a promising strategy to determine the active ingredient in commercial pesticide formulation. Eventually, a data fusion method based on near infrared (NIR) spectra and mid infrared (MIR) spectra for determination the active ingredient of emamectin benzoate in the commercial formulation was developed.