基于融合残差网络的人体异常行为识别方法

Human Abnormal Behavior Recognition Method Based on Fusion Residual Network

在视频监控系统中实时检测并识别异常行为的能力是智能监控系统的关键问题。本文提出了一种基于融合残差网络的网络框架,实现从视频中检测并识别人体异常行为。该残差网络框架主要包括人体检测模块和行为识别模块。基于前者本文提出检测残差网络(d-Res)采用多尺度目标检测策略来保证人体的检测速度以及检测效果;后者用于提取异常行为空间特征,使用基于迁移学习的识别残差网络(r-Res)提取图像的深层次特征,从而高效地对异常行为进行分类。最后,在UTI数据集上进行了实验,对本文所提算法的性能进行了评估。实验结果表明,所提出的方法在检测识别现实场景中的异常行为方面取得了令人满意的效果。

The ability to detect and identify abnormal behavior of video surveillance systems in real-time is the key problem for intelligent surveillance systems.This paper proposes a network framework based on fusion residual networks to detect and identify abnormal human behavior from video.The residual network framework mainly includes a human detection module and a behavior recognition module.Based on the former,this paper proposes that the detection residual network(d-Res)adopts a multi-scale object detection strategy to ensure the speed and effect of human detection.The latter is used to extract the spatial features of the abnormal behavior and uses the recognition residual network(r-Res)based on transfer learning to extract the deep features of the image so that effectively classifies the abnormal behavior.Finally,experiments are performed on the UTI data set to evaluate the performance of the proposed algorithm.The experimental results show that the proposed method achieves satisfactory results in detecting and recognizing abnormal behaviors in real-world scenarios.