基于OPO脉冲激光激发光声光谱的真假血液分类鉴别

为了实现快速准确并可回收再利用地鉴别真血和假血，采用光声光谱技术构建了一套血液光声检测系统并获取血样的光声信号。选取三种动物真血（马血、牛血和兔血）和两种假血（道具假血和红墨汁）共125组血样作为实验样本。获取了700～1 064 nm波段内所有样本的光声信号和光声峰峰值谱。实验表明，真血和假血的光声信号幅度、轮廓、峰值时间点和光声峰峰值均存在差异。为了实现高准确度的真假血液分类识别，采用了遗传优化的小波神经网络（WNN-GA）算法，对全波段100组样本进行训练，并构建了类Morlet小波基函数，然后对25组测试血样进行分类识别。利用遗传算法对WNN网络的权值、阈值和小波基函数平移、伸缩因子进行了优化，同时通过调节两个学习率因子，将真假血液的分类识别率提高了24%。采用主成分分析（PCA）对全波段血样光声峰峰值进行特征提取，再利用WNN-GA算法进行训练和分类识别。结果表明，在主成分个数为6时，PCA-WNN-GA融合算法可以使真假血液的分类识别率提高到100%。与另外6种分类识别算法相对比，该融合算法的识别准确率明显占优。光声光谱技术联合PCA-WNN-GA算法，可以准确地实现真假血液的分类鉴别。

Classification and Identification of Real or Fake Blood Based on OPO Pulsed Laser Induced Photoacoustic Spectroscopy

In order to rapidly and accurately achieve the identification of the real or fake blood, as well as recycled usage of blood, the photoacoustic spectroscopy was used in this work to establish a set of blood photoacoustic detection systems and to capture the photoacoustic signal of blood samples. Three kinds of animal blood (horse blood, cow blood, and rabbit blood), two kinds of fake blood (props blood and red ink), the total number of blood samples are 125 groups, were used as the experimental samples. The photoacoustic signals and photoacoustic peak-to-peak spectral of all blood samples at 700～1 064 nm were obtained. Photoacoustic experimental results show that the amplitude, profile, peak-point time, and peak-to-peak values of real and fake blood samples are different. To achieve the classification and identification of the real and fake blood with high precision, we used the wavelet neural network optimized by a genetic algorithm (WNN-GA) to train the 100 groups of samples for five kinds of blood in full wavelengths. Moreover, a kind of Morlet-like wavelet basis function was built. Then, 25 groups of blood samples were tested. Meanwhile, the GA algorithm was used to optimize the weights and thresholds of WNN network and the shift factor and stretch factor of wavelet basis function, and two learn factors can be adjusted. Compared with WNN, the correction rate of classification and identification for real and fake blood based on WNN-GA improved by 24%. Then, the principle components analysis (PCA) algorithm was used to extract the characteristic information of real or fake blood from the photoacoustic peak-to-peak full spectral. After that, the chosen principle components were trained and test by the WNN-GA algorithm. Results show that under 6 principle components, the algorithm of PCA-WNN-GA algorithm improves the correction rate of classification and identification for real and fake blood to 100%. Finally, compared with other the six algorithms, the correction rate of classification and identification for PCA-WNN-GA was superior to others. Therefore, the classification and identification of the real and fake blood can be well achieved via photoacoustic spectroscopy combined with the PCA-WNN-GA algorithm.