基于耦合神经网络的步态识别方法  

Gait Recognition Method Based on Coupled Neural Network

作　　者：沈喆 盛阳 SHEN Zhe;SHENG Yang(College of Civil Aviation,Shenyang Aerospace University,Shenyang 110136,China)

机构地区：[1]沈阳航空航天大学民用航空学院,沈阳110136

出　　处：《科技和产业》2024年第9期124-127,共4页Science Technology and Industry

基　　金：辽宁省教育厅青年科技人才“育苗”项目(JYT2020130);痕迹检验鉴定技术公安部重点实验室开放课题(HJKF201907);公安部文件检验重点实验室开放课题(FTKF202102)。

摘　　要：主流步态识别过程中识别率易受到视角变化的影响,使用耦合网络模型解决最小类间距离大于最大类内距离的问题,利用步态样本比对解决识别问题。用二值图像序列合成步态能量图,联合使用逻辑回归与对比损失函数对模型进行训练优化,实验验证耦合神经网络的步态识别性能,并与普通卷积神经网络模型(CNNs)的识别结果进行对比。在背包与大衣遮挡情况下的识别率达到73.7%和60.5%,高于CNNs的识别率,并提高了遮挡情况下步态识别精确度。In order to solve the problem that the recognition rate is easily affected by the change of view Angle in the mainstream gait recognition process,the coupled network model is used to solve the problem that the minimum distance between classes is greater than the maximum distance within classes,and the comparison of gait samples is used to solve the recognition problem.The gait energy diagram is synthesized by binary image sequences,and the model is trained and optimized by logistic regression and contrast loss function.The gait recognition performance of the coupled neural network was verified by experiments and compared with the recognition results of the common convolutional neural network(CNNs).The recognition rate reached 73.7%and 60.5%in the case of backpack and overcoat occlusion,which is higher than CNNs,and the accuracy of gait recognition under occlusion is improved.

关 键 词：步态识别 步态能量图 视角变化 耦合神经网络模型 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]