中文摘要：

用激光光致发光光谱技术对发票真伪的快速鉴别进行了研究。首先采集80个真假发票样本的激光光致发光光谱数据,然后利用遗传算法对谱线重叠严重的566～669nm波段进行高斯拟合,提取特征参数,再结合BP神经网络技术建立鉴别模型。结果表明,从104个光谱数据提取13个特征参数,真假发票的拟合确定系数R2分别为0.997 89和0.996 83,相对标准差RSD分别为0.017 052和0.022 362,有效实现对原始光谱特征信息的提取和简化;将13个特征参数和2个评估参数R2,RSD作为BP神经网络的输入变量,模型质量最高,对60个建模发票样本和20个未知发票样本的识别准确率均达到100%。说明文章提出的光谱分析方法具有很好的分类和鉴别作用,为发票真伪的快速鉴别提供了一种新方法。

英文摘要：

The rapid identification of invoice authenticity was studied based on laser-induced photoluminescence spectrum.First,the spectral curves of eighty invoice samples were obtained by laser-induced photoluminescence detection system,and genetic al-gorithm（GA）was applied to fit and separate overlapped spectral region between 566and 669nm by three Gaussian peaks.Spec-tral feature parameters extracted by GA were employed as the inputs of BP neural networks,and then an identification model wasbuilt.One hundred and four data were converted to 13Gaussian parameters,and for authentic and false invoices the coefficientsof determination（R2）were 0.997 89and 0.996 83and the relative standard deviations（RSD）were 0.017 052and 0.022 362,respectively.It was showed that Gaussian fitting algorithm could not only simplify the parameters of models,but also improvethe explanation of analysis models.Through comparison analysis of the results,it was found that the model,whose thirteen fea-ture parameters and two evaluated parameters were all applied as BP inputs,was the best,and the corrected identification ratesof sixty calibration samples and twenty validation samples were both 100%.So the identification method studied in the presentresearch played a good role in the classification and identification,and offered a new approach to the rapid identification of invoiceauthenticity.