摘要
行人检测是计算机视觉中的关键技术,是安防监控、视频行人行为分析乃至无人驾驶和驾驶辅助系统等后续研究的基础,因此,对行人检测的精确准度、实时性、鲁棒性都提出了较高的要求。文中给出了一个基于深度混合高斯背景建模和深度神经网络的检测方案,能够快速精准地检测出视频场景中的行人,并采用缩放、滑动的组合窗口策略,有效消除了因光线产生的阴影影响,定位出视频中行人的最佳位置。
As the key technology of computer vision, pedestrian detection is the foundation of research in security surveillance, human behavior analysis, autonomous vehicle systems and advanced driver assistant systems. So , it is important to make sure the accuracy, real-time performance and robustness of detection algorithm. This paper proposes a solution using Gauss mixture background modeling ( GMM )and deep neural networks,also applies strategy of resizing window and sliding window to locate the best position of pedestrians in surveillance video.
作者
丁文祥
朱孔凡
DING Wen-xiang;ZHU Kong-fan(School of Information Security Engineering,Shanghai Jiaotong University,Shanghai 200240,China;Yunnan Prison Adm inistration,Kunming 650031,China)
出处
《信息技术》
2016年第12期44-47,52,共5页
Information Technology
基金
国家科技支撑计划项目(2014BAK06B02)
