基于近红外光谱技术和极限学习机算法的血迹种属快速鉴定  

Rapid Identification of Bloodstain Based on Near Infrared Spectroscopy and Extreme Learning Machine Algorithm

作　　者：毕福轮 王伟 祁月莹 谢嘉懿 纳嫚 吴加权[3] 梁莹 张建强 BI Fulun;WANG Wei;QI Yueying;XIE Jiayi;NA Man;WU Jiaquan;LIANG Ying;ZHANG Jianqiang(Judicial Expertise Center of Xundian Public Security Bureau,Xundian 655200,Yunnan,China;Yunnan Police College,Kunming 650223,China;College of Science,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]云南寻甸回族彝族自治县公安司法鉴定中心,云南寻甸655200 [2]云南警官学院,昆明650223 [3]昆明理工大学理学院,昆明650500

出　　处：《刑事技术》2024年第5期507-513,共7页Forensic Science and Technology

基　　金：云南警官学院物证光谱技术省创新团队(202105AE160007);云南省刑事科学技术重点实验室开放课题(2020zz02)。

摘　　要：血迹是刑事案件案发现场最重要的物证之一,快速识别血迹、获取潜在证据,对刑事案件侦破具有重要意义。本文采用手持式近红外光谱仪采集了不同颜色纯棉纺织物上的人血、鸡血和猪血等不同类型的血迹样本的近红外光谱数据。利用标准正态变量(SNV)的预处理方式对采集的近红外光谱数据进行预操作,结合极限学习机(ELM)算法建立血迹种属快速识别模型。接下来,对建立的血迹种属快速识别模型进行了样本测试,并与使用传统支持向量机(SVM)和遗传算法-反向传播(GA-BP)所建立识别模型的测试结果进行了对比和分析。实验结果表明,ELM算法的预测准确率为98.48%,高于GA-BP算法的预测准确率(84.62%)和SVM算法的预测准确率(73.84%)。同时,ELM算法的精准度、灵敏度、特异度和F1-score也均远高于SVM和GA-BP算法。ELM算法所建立的血迹种属识别模型达到了较高的准确率,具有最佳的建模效果。本文的研究结果表明基于手持式近红外光谱仪和ELM算法结合的快速检测方法能够高效、无损、快速、准确地识别血迹类型,为刑事案件中血迹种属的快速鉴定和识别提供了新的技术参考。Bloodstain is one of the most important forensic evidences in criminal cases.How to identify the bloodstains and obtain some potential evidence is of great signifi cance to solve the criminal case.In this paper,a hand-held near-infrared(NIR)spectrometer was used to collect the spectral data of different species of bloodstains samples on cotton fabrics with different colors including human blood,chicken blood and pig blood.After collecting the spectral data,standard normal variables(SNV)pre-processing operation was implemented on the spectral data to eliminate the common offset and scaling effects.Then,the training models were established via extreme learning machine(ELM)algorithm to identify the species of bloodstain.Next,the testing samples were predicted by means of using the built specie identifi cation bloodstain model.Meanwhile,the traditional support vector machine(SVM)and genetic algorithm-back propagation(GA-BP)classifi cation algorithms were also used to build the identifi cation model and the prediction results were also compared with ELM algorithm.The experimental results showed that the prediction accuracy of ELM algorithm was 98.48%,which was higher than that of GA-BP algorithm(84.62%)and SVM algorithm(73.84%).Meanwhile,the precision,sensitivity and specificity of the prediction results using ELM algorithm were also much higher than those of SVM and GA-BP algorithms.The above results showed that the accuracy of the identifi cation model built by ELM algorithm was the highest and the overall performance of the model was the best.The research results of the paper showed that he rapid detection method based on a handheld NIR spectrometer and ELM algorithm could identify the types of the bloodstains effi ciently,non-destructively,quickly and accurately and it provided a new technical reference for bloodstains detection and identifi cation in criminal cases.

关 键 词：近红外 极限学习机 血迹种属 无损 快速 

分 类 号：DF795.1[医药卫生—法医学]