一种用于电力监控的行人运动检测与跟踪算法
|
摘要
在传统的行人检测跟踪算法中,主要采用HOG+SVM对视频中的行人直接进行检测,针对传统算法在复杂背景及多行人条件下,行人检测效果较差,且实时性低等缺点,提出一种融合目标运动检测与目标跟踪的行人检测算法。首先采用Vibe算法提取视频中的运动目标,并通过对Vibe算法的改进消除初始帧存在的阴影问题。针对视频中的运动目标采用Adaboost算法对运动目标区域进行行人检测,减小视频中背景的干扰,加快检测速度。最后采用卡尔曼滤波算法和匈牙利最优匹配算法对视频监控中的行人进行跟踪。仿真实验结果显示,该算法能够对电力监控视频中存在的行人进行检测跟踪。
As a traditional pedestrian detection and tracking algorithm,HOG+SVM is mainly used to detect pedestrians in video directly. In view of the disadvantages of the traditional algorithm in complex background and multi-pedestrian conditions,such as poor pedestrian detection effect and low real-time performance,this paper proposes a pedestrian detection algorithm that integrates target motion detection and target tracking. Firstly,the Vibe algorithm is used to extract the moving objects in video,and the shadow problem in the initial frame is eliminated by improving the Vibe algorithm.Aiming at moving objects in videos,the Adaboost algorithm is used to detect pedestrians in the moving target area,which can reduce background interference and accelerate detection speed. Finally,the Kalman filter algorithm and Hungarian optimal matching algorithm are used to track pedestrians in video surveillance. The simulation results show that the algorithm can detect and track pedestrians in power surveillance videos.
As a traditional pedestrian detection and tracking algorithm,HOG+SVM is mainly used to detect pedestrians in video directly. In view of the disadvantages of the traditional algorithm in complex background and multi-pedestrian conditions,such as poor pedestrian detection effect and low real-time performance,this paper proposes a pedestrian detection algorithm that integrates target motion detection and target tracking. Firstly,the Vibe algorithm is used to extract the moving objects in video,and the shadow problem in the initial frame is eliminated by improving the Vibe algorithm.Aiming at moving objects in videos,the Adaboost algorithm is used to detect pedestrians in the moving target area,which can reduce background interference and accelerate detection speed. Finally,the Kalman filter algorithm and Hungarian optimal matching algorithm are used to track pedestrians in video surveillance. The simulation results show that the algorithm can detect and track pedestrians in power surveillance videos.
引文
[1]柯天兵,李永倩.电力系统视频监控中目标检测与跟踪算法研究[J].电力科学与工程,2014,30(1):42-46.
[2]王青,杨耀权.电力监控系统中运动目标检测算法研究[J].电力科学与工程,2018,34(5):53-57.
[3]龚露鸣,徐美华,刘冬军,等.基于混合高斯和HOG+SVM的行人检测模型[J].上海大学学报(自然科学版),2018,24(3):341-351.
[4]邬厚民,吴卫祖.融合运动检测与目标跟踪的目标检测方法[J].计算机工程与设计,2018,39(7):2084-2089.
[5]邹薇,赵勋杰,李权,等.一种基于运动检测的行人多目标跟踪算法[J].计算机应用与软件,2014,31(8):132-135.
[6] WANG Y,PIERARD SEBASTIEN,SU S Z. Improving pedestrian detection using motion-guided filtering[J].Pattern Recognition Letters,2017,96(1):106-112.
[7] LI C,WANG X G,LIU W Y. Neural features for pedestrian detection[J]. Neurocomputing, 2017,238:420-432.
[8] JUNG S I,HONG K S. Deep network aided by guiding network for pedestrian detection[J]. Pattern Recognition Letters,2017,90:43-49.
[9] ZHANG Z G,TAO W B,SUN K,et al. Pedestrian detection aided by fusion of binocular information[J].Pattern Recognition,2016,60:227-238.
[10]马金鹏.改进的基于人头检测的行人跟踪算法[J].电子测量技术,2017,40(12):233-237.
[2]王青,杨耀权.电力监控系统中运动目标检测算法研究[J].电力科学与工程,2018,34(5):53-57.
[3]龚露鸣,徐美华,刘冬军,等.基于混合高斯和HOG+SVM的行人检测模型[J].上海大学学报(自然科学版),2018,24(3):341-351.
[4]邬厚民,吴卫祖.融合运动检测与目标跟踪的目标检测方法[J].计算机工程与设计,2018,39(7):2084-2089.
[5]邹薇,赵勋杰,李权,等.一种基于运动检测的行人多目标跟踪算法[J].计算机应用与软件,2014,31(8):132-135.
[6] WANG Y,PIERARD SEBASTIEN,SU S Z. Improving pedestrian detection using motion-guided filtering[J].Pattern Recognition Letters,2017,96(1):106-112.
[7] LI C,WANG X G,LIU W Y. Neural features for pedestrian detection[J]. Neurocomputing, 2017,238:420-432.
[8] JUNG S I,HONG K S. Deep network aided by guiding network for pedestrian detection[J]. Pattern Recognition Letters,2017,90:43-49.
[9] ZHANG Z G,TAO W B,SUN K,et al. Pedestrian detection aided by fusion of binocular information[J].Pattern Recognition,2016,60:227-238.
[10]马金鹏.改进的基于人头检测的行人跟踪算法[J].电子测量技术,2017,40(12):233-237.