基于近红外光谱的玉米品种鉴别方法

Qualitative Identification Method of Maize Varieties Based on Near-Infrared Spectroscopy

玉米因其耐旱、产量高、抗倒伏等优点,被广泛种植于我国各地。但因不同玉米品种间价格和品质差异较大,人工分辨其品种较为困难。基于此,本文利用近红外光谱技术结合机器学习建立预测模型,提出了一种快速鉴别玉米品种的方法。实验将采集到的玉米粒近红外光谱数据经过多元散射校正预处理后,建立核极限学习机模型用于玉米品种预测实验。结果表明,核极限学习机在玉米品种鉴别中能够表现出较好的效果,其预测准确率和F1值可以达到85.66%和90%。为了进一步提高预测准确率,实验还针对建模中的两个重要参数引入了灰狼优化算法,即核函数γ和惩罚因子C的寻优,该算法有效提升了模型准确率和F1值,达到了实际应用标准。该方法为食用玉米品种分类提供了技术保障,同时也对有关部门的管理和监督提供了借鉴。

Corn is widely planted in various parts of China due to its advantages such as drought resistance,high yield,and lodging resistance.However,due to significant differences in price and quality among different corn varieties,it is difficult to manually distinguish them.Based on this,this article uses near-infrared spectroscopy technology combined with machine learning to establish a prediction model and proposes a fast method for identifying corn varieties.The experiment will preprocess the collected near-infrared spectral data of corn kernels through multiple scattering correction,and establish a kernel limit learning machine model for corn variety prediction experiments.The results show that the kernel limit learning machine can perform well in identifying corn varieties,with prediction accuracy and F1 values reaching 85.66%and 90%,respectively.In order to further improve the prediction accuracy,the experiment also introduced the Grey Wolf optimization algorithm for two important parameters in modeling,it’s the optimization of kernel function parameterγand penalty factor C,this algorithm effectively improves the model’s prediction accuracy and F1 value,meeting the practical application standards.This method provides technical support for the classification of edible corn varieties,and also provides reference for the management and supervision of relevant departments.