近红外光谱技术在玉米种子活力检测中的应用研究

摘要

In order to realize rapid nondestructive recognition of maize vigor,a maize vigor intelligent detection model was put forward by combining NIR and BP neural network.At first,test samples were aged into three grades by artificial aging,and near infrared spectroscopy of which was collected.The original data were prerated using statitics method of normalization in order to eliminate noise and improve the efficiency of models. Principal component analysis(PCA)was respectively used to extract spectral features which acted as the input of BP neural network and the model was constructed according to preprocessing and feature extraction.The results showed the accuracy rate of the model was 90.3% and its average identification time was 27.36 ms.This investigation provided the theoretical support and practical method for rapid nondestructive recognition of maize vigor.%利用近红外光谱和BP神经网络建立玉米种子活力的快速无损检测模型。首先通过人工老化将样本按老化程度分为3种级别，分别采集样本的近红外光谱。对原始光谱进行矢量归一化预处理以消除光谱噪声。然后利用主成分分析（PCA）方法提取光谱特征，作为BP神经网络的输入，依据预处理及特征提取构建出BP神经网络种子活力检测模型。试验结果表明，该识别方法的准确率为90.3%，平均识别时间为27.36 ms。研究结果为玉米种子活力的快速无损检测提供了理论依据和实用方法。