摘要
为了实现大豆品种的快速无损鉴别,对16份大豆品种的近红外透射光谱(NITS)进行分析。首先通过平滑和马氏距离的光谱预处理方法消除噪声和去除奇异光谱。然后分别用主成分分析(PCA)和离散多带小波变换(DWT)提取光谱特征,作为BP神经网络的输入,构建PCA-BP和DWT-BP大豆品种识别模型。结果表明:PCA-BP模型的识别准确率为98.125%,平均识别时间为9.3 ms;DWT-BP模型的识别准确率为95.93%,平均识别时间为6.4 ms。研究结果为大豆品种的快速无损鉴别提供了理论依据和实用方法。
In order to realize rapid nondestructive recognition of soybean varieties, near infrared transmittance spectrum （NITS） of 16 soybean samples were analyzed. Smoothing treatment and Mahalanobis distance were used to filter noise and wipe off singular spectrum. Principal component analysis（PCA） and discrete wavelet transform（DWT） were respectively used to ex- tract spectral features which act as the input of BP neural network. PCA-BP and DWT-BP identification model were built. The accuracy rate of PCA-BP model and DWT-BP model were 98. 125% and 95.93%, in addition, the average recognition time were 9.3 ms and 6.4 ms. The results of the investigation provided the theoretical support and practical method for rapid nonde- structive recognition of soybean varieties.
出处
《大豆科学》
CAS
CSCD
北大核心
2013年第2期249-253,共5页
Soybean Science
基金
引进国际先进农业科学技术计划"948计划"(2008-Z24)
关键词
近红外透射光谱
主成分分析
离散小波变换
BP神经网络
大豆
Near infrared transmittance spectroscopy
Principal component analysis
Discrete wavelet transform
BP Neuralnetwork
Soybean
