中文摘要：

将拉曼光谱分析法与数理统计方法有机结合,构建人血与犬血种属判别模型,实现了不同种属血液样本的高效无损鉴别.采用拉曼光谱的无损测试模式对血液样本进行测试,考察了抗凝管管材、聚焦位置及曝光时间等对血液样本拉曼光谱的影响,在激发波长为632.8 nm,光谱扫描范围为200~1800 cm-1,功率衰减率50%,曝光时间5 s及累加次数为2次的优化条件下,获得了无损检测条件下的血液样本拉曼光谱图.针对血液样本组分复杂、拉曼光谱信号基底背景高等问题,提出了基于小波变换去噪,进行分段多项式基线校正的预处理方法,有效解决了血液样本拉曼光谱谱图的高噪音和基线漂移问题.实验选择30例正常人血和33例比格犬血为样本训练集,5例正常人血和5例比格犬血为测试集,基于主成分分析法（PCA）联合线性判别法（LDA）模型,训练集分类正确率达到95.23%,盲测集分类正确率达90.00%.这种基于非接触式血液样本拉曼光谱和PCA-LDA判断模型的测试方法在进出口检验检疫等涉及血液无损鉴别的领域具有广泛的应用价值和前景.

英文摘要：

Efficient and rapid non-destructive screening method for human and animal blood has great significance and practical prospects in export inspection and quarantine,forensic testing and other areas. By taking full advantages of Raman spectroscopy and combination of mathematical and statistical methods,species discriminant model for human blood and canine blood was established in this paper. First of all,the nondestructive Raman spectroscopy test for blood samples was focused. The factors which affected the non-destructive test were investigated in details,such as the material of anticoagulant tubes,the focus position and exposure time in Raman detecting and so on. Raman spectra of human and canine blood samples were obtained under the optimal conditions. The effective spectra were accumulated with 633 nm excitation in spectral range of 200—1800 cm-1with a 5 s integration time and 2 accumulating times. Laser power was selected approximately 8. 85 m W. In view of the complexity of the blood sample composition and the high signal background of the Raman spectrum,the pretreatment method of wavelet denoising and piecewise polynomial baseline correction for the series of Raman spectra was proposed and solved effectively the recognition rate of the Raman spectrum. The model of multivariate statistical algorithm was established from the training data which set consisted of 30 spectra from human blood and 33 spectra from dog blood and the testing data which included five spectra from human blood and five spectra from dog blood by combining principal component analysis（ PCA） with linear discriminant analysis（ LDA） in MATLAB to identified the species of human and dog blood. The classification accuracy of proposed model for the calibration set and the blind test are 95. 23% and 90. 00%. Studies have shown that based on the non-contact method of testing blood samples Raman and PCA-LDA classification model has broad application value and prospect in the fields of import and export inspection and quarantine blood nondes