摘要
初步研究了基于可见-近红外光谱技术和模式识别快速鉴别家蚕品种的方法。采用偏最小二乘法(PLS)进行模式特征分析,完成特征提取后,将获得的主成分作为神经网络的输入变量,建立了三层反向传播人工神经网络(BPANN)。试验采取4个品种的蚕种,应用所建立的PLS-BP模型对样本进行分析预测,准确率接近100%。
A method based on visible-near infrared spectroscopy (NIRS) and pattern recognition for fast and nondestructive identification of silkworm (Bombyx mori) varieties was preliminarily studied. By using partial least squares (PLS), the spectrum was reduced to some principal components which was taken as the input of back propagation artificial neural network (BPANN), and a three-layer neutral network model (PLS-BP) was established. Four varieties of silkworm eggs were analysed by the PLS-BP model and the recognition accuracy achieved was near 100%.
出处
《蚕业科学》
CAS
CSCD
北大核心
2006年第3期436-438,447,共4页
ACTA SERICOLOGICA SINICA
基金
国家茧丝绸发展风险基金资助项目[编号国茧协办函(2005)48号]