基于TOF相机和级联卷积网络的人头检测  

Head Detection Based on TOF Camera and Cascade Convolutional Network

作　　者：胡文俊 马秀丽[1] HU Wenjun;MA Xiuli(School of Information and Communication Engineering,Shanghai University,Shanghai 200444,China)

机构地区：[1]上海大学信息与通信工程学院,上海200444

出　　处：《电视技术》2020年第7期41-44,共4页Video Engineering

摘　　要：图像序列中遮挡和角度变化等因素会造成人头漏检。TOF(Time-of-Filght)相机可以捕获场景中物体间的空间位置关系,因此将TOF相机置于俯视视角采集的深度图像信息,可以采集丰富的人体头部特征。鉴于此,提出了一种基于级联卷积网络的深度图像快速人头检测方法。Occlusion and Angle change in image sequence can cause capitation omission.The TOF(Time-of-Filght)camera can capture the spatial position relationship between objects in the scene.The depth image information collected by placing the TOF camera in a overlooking perspective can collect abundant human head features.In view of this,this paper proposes a fast head detection method for depth images based on cascading convolutional network.

关 键 词：深度图像 人头检测 级联网络 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]