基于步态特征提取的ELM身份识别方法  

ELM Identification Method Based on Gait Feature Extraction

作　　者：马添力 肖文栋[1,2] MA Tianli;XIAO Wendong(School of Automation,University of Science and Technology Beijing,Beijing 100083,China;Beijing Engineering Research Center of Industrial Spectrum Imaging,Beijing 100083,China)

机构地区：[1]北京科技大学自动化学院,北京100083 [2]北京市工业波谱成像工程技术研究中心,北京100083

出　　处：《自动化仪表》2021年第1期17-23,共7页Process Automation Instrumentation

摘　　要：利用步态信息进行身份识别是一种新兴的生物识别技术。相较于其他的生物识别技术,其具有不易伪装、可在远距离情况下进行身份识别的优点。现有模型的识别方法计算量大、模型难以准确建立;现有的分类方法普遍存在训练时间长、分类准确率不高的问题。针对以上问题,对步态视频进行分帧处理,将分帧后的图像进行运动目标检测、形态学处理和图像归一化预处理,生成步态能量图(GEI),提出了对GEI进行特征提取并采用超限学习机(ELM)进行分类的方法。测试结果表明,该方法在保证身份识别准确率的前提下,训练模型的速度有大幅提升。所提方法对利用步态信息进行身份识别有一定指导意义,特别对大规模图像分类问题的训练速度提升有较大启发。The identification with the gait information is a new biometrics technology,compared with other biometrics,it has the advantages of not easy to disguise and can be identified at a distance.The gait recognition method based on the model is computationally expensive and difficult to establish the model accurately;the existing classification methods generally have the problems of long training time and low classification accuracy.In view of the above problems,this paper divided the gait video into frames.After preprocessing such as moving target detection,morphological processing and image normalization,gait energy image(GEI)was generated,this paper proposes a method of feature extraction of GEI and classification by using extreme learning machine(ELM).The experimental results show that this method can improve the speed of the training model greatly on the premise of ensuring the classification accuracy.The proposed method has certain guiding significance for the use of gait information for identity recognition,especially for the improvement of the training speed of large-scale image classification problem.

关 键 词：背景建模 步态能量图 特征提取 数据增强 超限学习机 步态信息 身份识别 

分 类 号：TH86[机械工程—仪器科学与技术]