The Effect of Spectral Pretreatment on the LSSVM Model of Nitrogen in Intertidal Sediments
LÜ Mei-rong1, REN Guo-xing1, 2, LI Xue-ying1, FAN Ping-ping1, LIU Jie1, SUN Zhong-liang1, HOU Guang-li1, LIU Yan1*
1. Institute of Oceanographic Instrmentation,Qilu University of Technology (Shandong Academy of Sciences),Shandong Provincial Key Laboratory of Marine Monitoring Instrument Equipment Technology,National Engineering and Technological Research Center of Marine Monitoring Equipment, Qingdao 266100, China
2. School of Information Science and Engineering, Ocean University of China, Qingdao 266100, China
Abstract:Spectral data transformation and feature wavelength extraction are two important spectral pretreatment methods, which play an important role in eliminating environmental interference. Previous literature mainly compared different spectral data transformation methods and there was less studyon the spectral feature wavelength extraction methods and the combination of the two methods. In order to obtain suitable spectral pretreatment method and improve the accuracy of LSSVM model of sediment nitrogen in the intertidal zone, the effect of 4 spectral transformation methods combined with 3 characteristic wavelength extraction methods on the accuracy of LSSVM model of sediment nitrogen is studied for accurate prediction of sediment nitrogen in the intertidal zone. The results showed that the spectral transformation methods of multivariate scattering correction (MSC) or normal distribution (SVN) increasedthe correlation between spectra and nitrogen content and the highest correlation reached 0.69 and 0.71 respectively. MSC and SVN improved the prediction accuracy of LSSVM model, and the prediction R2 and RPD are 0.88, 0.87 and 2.78, 2.69, respectively. The feature wavelength extraction method of uninformative variable elimination (UVE) also improved the prediction accuracy of LSSVM model, model test R2 and RPD were 0.89 and 2.70, respectively. However, not all of the characteristic wavelengths extracted by UVE have a high correlation with nitrogen content. In addition, the combination of UVE and MSC or SVN also improved the prediction accuracy of the model, but it is not better than UVE alone or MSC or SVN alone. The results of this paper can provide a technical reference for nitrogen estimation and spectral data preprocessing of intertidal sediments.
吕美蓉,任国兴,李雪莹,范萍萍,刘 杰,孙中梁,侯广利,刘 岩. 光谱数据预处理对潮间带沉积物氮LSSVM模型的影响研究[J]. 光谱学与光谱分析, 2020, 40(08): 2409-2414.
LÜ Mei-rong, REN Guo-xing, LI Xue-ying, FAN Ping-ping, LIU Jie, SUN Zhong-liang, HOU Guang-li, LIU Yan. The Effect of Spectral Pretreatment on the LSSVM Model of Nitrogen in Intertidal Sediments. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2020, 40(08): 2409-2414.