摘要
伪装人体检测在视频监控领域具有重要的研究和应用价值。本文针对目前基于图像表层特征提取的伪装人体检测方法无法有效检测出没有明显移动的伪装人体目标,充分利用了人体目标具有一定范围呼吸率的独特特性,提取并增强伪装人体的微振动特征,利用振动特征实现了伪装人体目标进行检测。首先,提出了自己的伪装人体视频数据集和基于微振动特征的伪装人体检测模型;其次利用训练集估计伪装人体目标的最优呼吸率区间,利用估计的最优呼吸率区间增强了视频中伪装人体的微振动;最后,根据增强了微振动的视频和提出的检测模型实现伪装人体目标的定位检测,并通过图像形态学去噪后处理降低检测结果的噪声。本文提取微振动特征的检测算法通过充分的消融实验和对比实验进行了验证,在提出的视频数据集上,检测效果IOU达到了0.526,Precision达到了0.738,优于其它图像特征最先进的检测方法。
Camouflaged human detection has important research and application value in the field of video surveillance.In view of the fact that the current detection methods for camouflaged human based on image surface feature extraction cannot effectively detect the target without obvious movement,this paper makes full use of the human’s unique characteristics with a certain range of respiration rates.This paper extracts and enhances the micro-vibration features of the camouflaged human to realize the detection of the target.First,a camouflaged human video data set and a camouflaged human detection model based on micro-vibration features are proposed.And then the training set is used to estimate the optimal respiration rate interval of camouflaged human.Next,the estimated optimal respiration rate interval is used to enhance the micro-vibration of the human.Finally,according to the enhanced micro-vibration video and the proposed detection model,the camouflaged human in video gets detected and segmented,and the post-processing of image morphology denoising is used to reduce the noise of the result.The detection algorithm for extracting micro-vibration features in this paper is verified by sufficient ablation experiments and comparative experiments.On the proposed video data set,the detection IOU reaches 0.526,and the precision reaches 0.738,which is better than other advanced detection methods.
作者
黄飞龙
杨学志
吴克伟
沈晶
陈鲸
HUANG Feilong;YANG Xuezhi;WU Kewei;SHEN Jing;CHEN Jing(School of Computer and Information,Hefei University of Technology,Hefei 230009,China;Anhui Province Key Laboratory of Industry Safety and Emergency Technology,Hefei 230009,China;School of Software,Hefei University of Technology,Hefei 230009,China;School of Electronic Information and Electrical Engineering,Hefei Normal University,Hefei 230061,China;Anhui Microwave and Communication Engineering Technology Research Center,Hefei 230061,China)
出处
《智能计算机与应用》
2021年第6期36-42,50,共8页
Intelligent Computer and Applications
关键词
伪装人体
振动特征
呼吸率区间
微振动
camouflaged human
vibration features
respiration rate interval
micro-vibration
