基于轻量级卷积网络的步态识别研究  

Gait recognition based on lightweight convolutional networks

作　　者：赵鑫泽 代雪晶[1] Zhao Xinze;Dai Xuejing(School of Public Security Information Technology and Intelligence,Criminal Investigation Police University of China,Shenyang 110000,China)

机构地区：[1]中国刑事警察学院公安信息技术与情报学院,沈阳110000

出　　处：《现代计算机》2023年第13期60-64,共5页Modern Computer

摘　　要：针对目前深度学习网络参数量大、成本较高等问题,提出了一种将轻量级卷积网络应用在步态识别中的研究,实现对行人步态的准确识别。该方法对数据集图像进行背景减除和归一化等预处理操作,根据每个人的步态特征提取出相应的步态能量图,使用Mobilenet网络模型进行特征学习并完成分类任务。实验结果表明,Mobilenet网络在减少模型参数量的同时,可以提高模型的识别精度,平均准确率和召回率可以达到90.7%、89.99%,能够有效地实现步态的识别及分类。In view of the large number of parameters and high cost of deep learning network,a lightweight convolutional network is proposed in this paper to achieve accurate recognition of pedestrian gait.In this method,the data set images were preprocessed by background subtraction and normalization,and the corresponding gait energy graph was extracted according to each person’s gait features.The Mobilenet network model was used for feature learning and classification.The experimental results show that the Mobilenet network can improve the recognition accuracy of the model while reducing the number of model parameters,and the average accuracy and recall rate can reach 90.7%and 89.99%,which can effectively realize the recognition and classification of gait.

关 键 词：深度学习 步态识别 卷积神经网络 步态能量图 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP18[自动化与计算机技术—计算机科学与技术]