基于卷积神经网络的毛发显微图像分类
|
摘要
显微照相在现场毛发物证提取与采集中被广泛应用,而基于显微形态图像处理分析的毛发识别有助于进一步提高毛发检验鉴定的自动化程度。为实现毛发的无损检验,首先利用光学显微镜进行不同来源、不同条件的毛发形态显微图像的采集,使用Matlab软件预处理400倍显微镜下拍摄的样本图像数据;尝试多种卷积神经网络的结构和参数组合对毛发显微图像进行特征提取与训练学习,加入中心损失函数提升识别率及泛化能力来优化网络,将显微图像输入网络即可实现毛发的快速识别。本文对人类毛发、假发等四类共20 000张显微图像进行分类训练及测试实验。结果表明,在学习率为0. 000 7,训练迭代次数为30 000次时,识别率达98. 97%,泛化精度达98. 80%。该方法可实现毛发显微图像的高效、准确、自动分类识别,可提高毛发鉴定效率。
Photomicrography is widely used in the field of hair material evidence extraction and collection,and hair recognition based on microscopic image processing analysis helps to further improve the automation of hair inspection and identification. In order to achieve non-destructive testing of hair,firstly,optical microscopy was used to collect hair microscopic images from different sources and under different conditions,and sample image data taken under 400 times microscope was preprocessed using Matlab software. Extract and train the features of hair microscopy image combined with various structures and parameters groups of convolutional neural networks the parameters. Add the central loss function to improve the recognition rate and generalization ability to optimize the network,and input the microscopic image into the network to realize the rapid recognition of hair. In this paper,a total of 20,000 microscopic images of human hair,wig and other four types of classification were used in training and testing experiments. The results show that when the learning rate is 0. 000 7 and the number of training iterations is 30,000,the recognition rate is 98. 97%,and the generalization accuracy is 98. 80%. The method can realize efficient,accurate and automatic classification and recognition of hair microscopic images,and can improve hair identification efficiency.
Photomicrography is widely used in the field of hair material evidence extraction and collection,and hair recognition based on microscopic image processing analysis helps to further improve the automation of hair inspection and identification. In order to achieve non-destructive testing of hair,firstly,optical microscopy was used to collect hair microscopic images from different sources and under different conditions,and sample image data taken under 400 times microscope was preprocessed using Matlab software. Extract and train the features of hair microscopy image combined with various structures and parameters groups of convolutional neural networks the parameters. Add the central loss function to improve the recognition rate and generalization ability to optimize the network,and input the microscopic image into the network to realize the rapid recognition of hair. In this paper,a total of 20,000 microscopic images of human hair,wig and other four types of classification were used in training and testing experiments. The results show that when the learning rate is 0. 000 7 and the number of training iterations is 30,000,the recognition rate is 98. 97%,and the generalization accuracy is 98. 80%. The method can realize efficient,accurate and automatic classification and recognition of hair microscopic images,and can improve hair identification efficiency.
引文
[1]贺英,万立华.毛发检验在法医学中的应用和新进展[J].重庆医科大学学报,2006,(31):58-60.
[2]周飞燕,金林鹏,董军.卷积神经网络研究综述[J].计算机学报,2017,40(06):1229-1251.
[3]程国建,郭文惠,范鹏召.基于卷积神经网络的岩石图像分类[J].西安石油大学学报(自然科学版),2017,(4):116-122.
[4]宋欣益.基于卷积神经网络的高光谱数据分类方法研究[D].哈尔滨:哈尔滨工业大学,2016.
[5]PENG Min,WANG Chongyang,CHEN Tong,et al.NIR-Face Net:A Convolutional Neural Network for Near-Infrared Face Identification[J].Information,2016,7(4):1-14.
[6]李武军,王崇骏,张炜,等.人脸识别研究综述[J].模式识别与人工智能,2006,19(1):58-66.
[7]LAWRENCE S,GILES C L,TSOI A C,et al.Face Recognition-A Convolutional Neutral Network Approach.[J].IEEETrans on Neutral Networks,1997,8(1):98-113.
[8]TAIGMAN Y,YANG M,RANZATO M A,et al.Deep Face:Closingthe Gap to Human-Level Performance in Face Verification[C]//2014 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).IEEE Computer Society,Columbus,Ohio,USA,2014:1701-1708.
[9]SUN Yi,WANG Xiaogang,TANG Xiaoou.Deep Learning Face Representation from Predicting10000 Classes[C]//2014 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).IEEE Computer Society,Columbus,Ohio,USA,2014:1891-1898.
[10]SUN Yi,CHEN Yuheng,WANG Xiaogang,et al.Deep learning face representation by joint identification-verification[C]//2014 International Conference on Neural Information Processing Systems.Montreal,Canada,2014:1988-1996.
[11]SUN Yi,Wang Xiaogang,TANG Xiaoou.Deeply learned face representations are sparse,selective,and robust[C]//2015IEEE Conference On Computer Vision And Pattern Recognition(CVPR).IEEE,Boston,USA,2015:2892-2900.
[12]SUN Yi,LIANG Ding,WANG Xiaogang,et al.Deep ID3:Face recognition with very deep neural networks[J].Computer Science,2015,2(3):1-5.
[13]SCHROFF F,KALENICHENKO D,PHILBIN J.Face Net:Aunified embedding for face recognition and clustering[C]//2015 IEEE Conference On Computer Vision And Pattern Recognition(CVPR).IEEE,Boston,USA,2015:815-823.
[14]李亚,王广润,王青.基于深度卷积神经网络的跨年龄人脸识别[J].北京邮电大学学报,2017,40(01):84-88+110.
[15]刘维达.基于笔迹的性别识别方法研究[D].哈尔滨:哈尔滨工业大学,2017.
[16]LUIZ G Hafemann,ROBERT Sabourin,LUIZ S Olivera.Analyzing features learned for offine signature verification using deep cnns[J].Internation Conference on Pattern Recognition(ICPR),2016 23rd,2016:2989-2994.
[17]LUIZ G Hafemann,ROBERT Sabourin,LUIZ S Olivera.Written dependent feature learning for offine signature verification using deep convolutional neutral networks[J].Pattern Recognition,2017(70):163-176.
[18]邓辉瑶.基于深度卷积神经网络的离线笔迹鉴别研究[D].武汉:华中科技大学,2016.
[19]梁曦璐.基于改进神经网络的离线签名笔迹识别[D].北京:中国政法大学,2017.
[20]魏海宏,刘新.基于深度卷积神经网络的离线手写签名鉴定[J].网络安全技术与应用,2018(03):40-41.
[21]王星,郑湙彬,朱枫怡,等.基于非参数回归和卷积神经网络的在线手写签名身份认证模型研究[J].数理统计与管理,2018,37(04):610-623.
[22]ABDELHAMID O,MOHAMED A R,JIANG Hui,et al.Applying Convolutional Neural Networks concepts to hybrid NN-HMM model for speech recognition[C]∥IEEE International Conference on Acoustics.IEEE,Speech and Signal Processing(ICASSP).Kyoto,Japan,2012:4277-4280.
[23]ABDELHAMID O,MOHAMED A R,JIANG Hui,et al.Convolutional Neural Networks for Speech Recognition[J].IEEE/ACM Transactions on Audio,Speech,and Language Processing,2014,22(10):1533-1545.
[24]胡涛,张超,程炳,等.卷积神经网络在异常声音识别中的研究[J].信号处理,2018,34(03):357-367.
[25]张文宇,刘畅.卷积神经网络算法在语音识别中的应用[J].信息技术,2018(10):147-152.
[26]张文涛,韩莹莹,黎恒.基于卷积神经网络的交通声音事件识别方法[J].现代电子技术,2018,41(14):70-73+78.
[27]徐宏伟,严迪群,阳帆,等.基于卷积神经网络的电子变调语音检测算法[J].电信科学,2018,(2):46-57.
[28]胡涛,张超,程炳,等.卷积神经网络在异常声音识别中的研究[J].信号处理,2018,(3):357-367.
[29]王欣,唐俊,王年.基于双层卷积神经网络的步态识别算法[J].安徽大学学报(自然科学版),2015,39(01):32-36.
[30]王欣.基于卷积神经网络与足底压力信息的步态识别[D].合肥:安徽大学,2015.
[31]刘璐璐.基于卷积神经网络的人体行为识别研究[D].合肥:中国科学技术大学,2017.
[32]李影.基于三维卷积神经网络步态识别方法的研究[D].北京:北京交通大学,2018.
[33]邵禹铭.基于深度学习方法的人体微动特征识别[D].合肥:中国科学技术大学,2018.
[34]景军锋,范晓婷,李鹏飞,等.应用深度卷积神经网络的色织物缺陷检测[J].纺织学报,2017,38(2):68-74.
[35]张宏伟,张凌婕,李鹏飞.基于深度卷积神经网络的织物花型分类[J].纺织高校基础科学学报,2017,30(2):261-265.
[36]冀中,刘青,聂林红,等.基于卷积神经网络的纹理分类方法研究[J].计算机科学与探索,2016,10(3):389-397.
[37]张玮,张华熊.基于卷积神经网络的纺织面料主成分分类[J/OL].浙江理工大学学报(自然科学版):1-7[2018-11-07].
[38]高惠琳.基于卷积神经网络的军事图像分类[J].计算机应用研究,2017,34(11):3518-3520.
[39]HINTON G E,SALAKHUTDINOV R R.Reducing the dimensionality of data with neural networks[J].Science,2006,313(5786):504-507.
[40]龚丁禧,曹长荣.基于卷积神经网络的植物叶片分类[J].计算机与现代化,2014(04):12-15+19.
[41]ALEX K,LLYA S,GEOFFREY H.Image Net classification with deep convolutional neural networks[J].Advances in Neural Inforrnation Processing Systems,2012,25(2):1106-1114.
[42]杨帅,张有芬,李玉惠,等.基于深度卷积神经网络的车标分类[J].工业仪表与自动化装置,2017(05):75-78.
[43]刘园园.基于卷积神经网络的花卉图像分类算法的研究[D].北京:华北电力大学,2017.
[44]张弛.基于卷积神经网络的鞋印图像分类算法研究[D].大连:大连海事大学,2016.
[45]杨小林,席焕久,温有锋,等.人类毛发的形态学研究及法医学意义[J].解剖科学进展,2011,17(01):86-89.
[46]张伟,徐艳春.毛发微观结构研究的回顾与展望[J].兽类学报,2003(04):339-345.
[47]周用武,王泽威.利用偏光显微镜检验鉴定兽类毛发的初步报告[J].四川动物,2006,25(3):614.
[48]闫立强.显微镜在毛发物证检验中的应用[J].辽宁警专学报,2014(5):59-62.
[49]高野,闫立强.基于相差显微镜无损检验是否染色头发的研究[J].辽宁警专学报,2014(1):57-61.
[50]郭海涛,薛晓明,侯森林.红外光谱分析在野生动物毛发鉴定中的应用[J].江苏农业科学,2011,39(6):495-497.
[51]詹重万,肖开琪,陈年芳.正常人头发的扫描电镜观察[J].解剖学报,1983(3):334-336.
[52]邹友,陶克明,李立新,等.毛发常见机械性损伤形态的环境扫描电镜研究[J].刑事技术,2006(1):4-6.
[2]周飞燕,金林鹏,董军.卷积神经网络研究综述[J].计算机学报,2017,40(06):1229-1251.
[3]程国建,郭文惠,范鹏召.基于卷积神经网络的岩石图像分类[J].西安石油大学学报(自然科学版),2017,(4):116-122.
[4]宋欣益.基于卷积神经网络的高光谱数据分类方法研究[D].哈尔滨:哈尔滨工业大学,2016.
[5]PENG Min,WANG Chongyang,CHEN Tong,et al.NIR-Face Net:A Convolutional Neural Network for Near-Infrared Face Identification[J].Information,2016,7(4):1-14.
[6]李武军,王崇骏,张炜,等.人脸识别研究综述[J].模式识别与人工智能,2006,19(1):58-66.
[7]LAWRENCE S,GILES C L,TSOI A C,et al.Face Recognition-A Convolutional Neutral Network Approach.[J].IEEETrans on Neutral Networks,1997,8(1):98-113.
[8]TAIGMAN Y,YANG M,RANZATO M A,et al.Deep Face:Closingthe Gap to Human-Level Performance in Face Verification[C]//2014 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).IEEE Computer Society,Columbus,Ohio,USA,2014:1701-1708.
[9]SUN Yi,WANG Xiaogang,TANG Xiaoou.Deep Learning Face Representation from Predicting10000 Classes[C]//2014 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).IEEE Computer Society,Columbus,Ohio,USA,2014:1891-1898.
[10]SUN Yi,CHEN Yuheng,WANG Xiaogang,et al.Deep learning face representation by joint identification-verification[C]//2014 International Conference on Neural Information Processing Systems.Montreal,Canada,2014:1988-1996.
[11]SUN Yi,Wang Xiaogang,TANG Xiaoou.Deeply learned face representations are sparse,selective,and robust[C]//2015IEEE Conference On Computer Vision And Pattern Recognition(CVPR).IEEE,Boston,USA,2015:2892-2900.
[12]SUN Yi,LIANG Ding,WANG Xiaogang,et al.Deep ID3:Face recognition with very deep neural networks[J].Computer Science,2015,2(3):1-5.
[13]SCHROFF F,KALENICHENKO D,PHILBIN J.Face Net:Aunified embedding for face recognition and clustering[C]//2015 IEEE Conference On Computer Vision And Pattern Recognition(CVPR).IEEE,Boston,USA,2015:815-823.
[14]李亚,王广润,王青.基于深度卷积神经网络的跨年龄人脸识别[J].北京邮电大学学报,2017,40(01):84-88+110.
[15]刘维达.基于笔迹的性别识别方法研究[D].哈尔滨:哈尔滨工业大学,2017.
[16]LUIZ G Hafemann,ROBERT Sabourin,LUIZ S Olivera.Analyzing features learned for offine signature verification using deep cnns[J].Internation Conference on Pattern Recognition(ICPR),2016 23rd,2016:2989-2994.
[17]LUIZ G Hafemann,ROBERT Sabourin,LUIZ S Olivera.Written dependent feature learning for offine signature verification using deep convolutional neutral networks[J].Pattern Recognition,2017(70):163-176.
[18]邓辉瑶.基于深度卷积神经网络的离线笔迹鉴别研究[D].武汉:华中科技大学,2016.
[19]梁曦璐.基于改进神经网络的离线签名笔迹识别[D].北京:中国政法大学,2017.
[20]魏海宏,刘新.基于深度卷积神经网络的离线手写签名鉴定[J].网络安全技术与应用,2018(03):40-41.
[21]王星,郑湙彬,朱枫怡,等.基于非参数回归和卷积神经网络的在线手写签名身份认证模型研究[J].数理统计与管理,2018,37(04):610-623.
[22]ABDELHAMID O,MOHAMED A R,JIANG Hui,et al.Applying Convolutional Neural Networks concepts to hybrid NN-HMM model for speech recognition[C]∥IEEE International Conference on Acoustics.IEEE,Speech and Signal Processing(ICASSP).Kyoto,Japan,2012:4277-4280.
[23]ABDELHAMID O,MOHAMED A R,JIANG Hui,et al.Convolutional Neural Networks for Speech Recognition[J].IEEE/ACM Transactions on Audio,Speech,and Language Processing,2014,22(10):1533-1545.
[24]胡涛,张超,程炳,等.卷积神经网络在异常声音识别中的研究[J].信号处理,2018,34(03):357-367.
[25]张文宇,刘畅.卷积神经网络算法在语音识别中的应用[J].信息技术,2018(10):147-152.
[26]张文涛,韩莹莹,黎恒.基于卷积神经网络的交通声音事件识别方法[J].现代电子技术,2018,41(14):70-73+78.
[27]徐宏伟,严迪群,阳帆,等.基于卷积神经网络的电子变调语音检测算法[J].电信科学,2018,(2):46-57.
[28]胡涛,张超,程炳,等.卷积神经网络在异常声音识别中的研究[J].信号处理,2018,(3):357-367.
[29]王欣,唐俊,王年.基于双层卷积神经网络的步态识别算法[J].安徽大学学报(自然科学版),2015,39(01):32-36.
[30]王欣.基于卷积神经网络与足底压力信息的步态识别[D].合肥:安徽大学,2015.
[31]刘璐璐.基于卷积神经网络的人体行为识别研究[D].合肥:中国科学技术大学,2017.
[32]李影.基于三维卷积神经网络步态识别方法的研究[D].北京:北京交通大学,2018.
[33]邵禹铭.基于深度学习方法的人体微动特征识别[D].合肥:中国科学技术大学,2018.
[34]景军锋,范晓婷,李鹏飞,等.应用深度卷积神经网络的色织物缺陷检测[J].纺织学报,2017,38(2):68-74.
[35]张宏伟,张凌婕,李鹏飞.基于深度卷积神经网络的织物花型分类[J].纺织高校基础科学学报,2017,30(2):261-265.
[36]冀中,刘青,聂林红,等.基于卷积神经网络的纹理分类方法研究[J].计算机科学与探索,2016,10(3):389-397.
[37]张玮,张华熊.基于卷积神经网络的纺织面料主成分分类[J/OL].浙江理工大学学报(自然科学版):1-7[2018-11-07].
[38]高惠琳.基于卷积神经网络的军事图像分类[J].计算机应用研究,2017,34(11):3518-3520.
[39]HINTON G E,SALAKHUTDINOV R R.Reducing the dimensionality of data with neural networks[J].Science,2006,313(5786):504-507.
[40]龚丁禧,曹长荣.基于卷积神经网络的植物叶片分类[J].计算机与现代化,2014(04):12-15+19.
[41]ALEX K,LLYA S,GEOFFREY H.Image Net classification with deep convolutional neural networks[J].Advances in Neural Inforrnation Processing Systems,2012,25(2):1106-1114.
[42]杨帅,张有芬,李玉惠,等.基于深度卷积神经网络的车标分类[J].工业仪表与自动化装置,2017(05):75-78.
[43]刘园园.基于卷积神经网络的花卉图像分类算法的研究[D].北京:华北电力大学,2017.
[44]张弛.基于卷积神经网络的鞋印图像分类算法研究[D].大连:大连海事大学,2016.
[45]杨小林,席焕久,温有锋,等.人类毛发的形态学研究及法医学意义[J].解剖科学进展,2011,17(01):86-89.
[46]张伟,徐艳春.毛发微观结构研究的回顾与展望[J].兽类学报,2003(04):339-345.
[47]周用武,王泽威.利用偏光显微镜检验鉴定兽类毛发的初步报告[J].四川动物,2006,25(3):614.
[48]闫立强.显微镜在毛发物证检验中的应用[J].辽宁警专学报,2014(5):59-62.
[49]高野,闫立强.基于相差显微镜无损检验是否染色头发的研究[J].辽宁警专学报,2014(1):57-61.
[50]郭海涛,薛晓明,侯森林.红外光谱分析在野生动物毛发鉴定中的应用[J].江苏农业科学,2011,39(6):495-497.
[51]詹重万,肖开琪,陈年芳.正常人头发的扫描电镜观察[J].解剖学报,1983(3):334-336.
[52]邹友,陶克明,李立新,等.毛发常见机械性损伤形态的环境扫描电镜研究[J].刑事技术,2006(1):4-6.