Machine Learning Identification of Saline-Alkali-Tolerant Japonica Rice Varieties Based on Raman Spectroscopy and Python Visual Analysis

文献类型： 外文期刊

第一作者： Liu, Rui

作者： Liu, Rui;Tan, Feng;Wang, Yaxuan;Ma, Bo;Yuan, Ming;Wang, Lianxia;Zhao, Xin

作者机构： Heilongjiang Bayi Agr Univ, Coll Agr Engn, Daqing 163000, Peoples R China;Heilongjiang Bayi Agr Univ, Coll Elect & Informat, Daqing 163000, Peoples R China;Heilongjiang Bayi Agr Univ, Coll Civil Engn & Water Conservancy, Daqing 163000, Peoples R China;Heilongjiang Acad Agr Sci, Qiqihar Branch, Qiqihar 161006, Peoples R China;Qiqihar Univ, Coll Comp & Control Engn, Qiqihar 161006, Peoples R China

关键词： japonica rice; saline-alkali-tolerant; Raman spectroscopy; Python visual; RFE; typical nonlinear; SVM

期刊名称：AGRICULTURE-BASEL （ 2021影响因子：3.408； 五年影响因子：3.459 ）

ISSN：

年卷期： 2022 年 12 卷 7 期

页码：

收录情况： SCI

摘要： The core of saline-alkali land improvement is planting suitable plants. Planting rice in saline-alkali land can not only effectively improve saline-alkali soil, but also increase grain yield. However, traditional identification methods for saline-alkali-tolerant rice varieties require tedious and time-consuming field investigations based on growth indicators by rice breeders. In this study, the visualization method of Python data processing was used to analyze the Raman spectroscopy of japonica rice in order to study a simple and efficient identification method of saline-alkali-tolerant japonica rice varieties. Three saline-alkali-tolerant japonica varieties and three saline-alkali-sensitive japonica varieties were collected from control and saline-alkali-treated fields, respectively, and the Raman spectra of 432 samples were obtained. The data preprocessing stage used filtering-difference method to process Raman spectral data to complete interference reduction and crests extraction. In the feature selection stage, scipy.signal.find_peaks (SSFP), SelectKBest (SKB) and recursive feature elimination (RFE) were used for machine feature selection of spectral data. According to the feature dimension obtained by machine feature selection, dataset partitioning by K-fold CV, the typical linear logistic regression (LR) and typical nonlinear support vector machine (SVM) models were established for classification. Experimental results showed that the typical nonlinear SVM identification model based on both RFE machine feature selection and six-fold CV dataset partitioning had the best identification rate, which was 94%. Therefore, the SVM classification model proposed in this study could provide help in the intelligent identification of saline-alkali-tolerant japonica rice varieties.

分类号：

  • 相关文献
作者其他论文 更多>>

微信小程序