基于团块模型的行人阴影抑制算法

J4 ›› 2014, Vol. 36 ›› Issue (11): 2203-2209.

基于团块模型的行人阴影抑制算法

金圣韬1，孟朝晖1，刘伟2

(1.河海大学计算机与信息学院，江苏南京 211100;2.63780部队，海南三亚 572000)

收稿日期:2013-05-23 修回日期:2013-09-22 出版日期:2014-11-25 发布日期:2014-11-25

A pedestrian shadow eliminating algorithm
based on blob model

JIN Shengtao1,MENG Zhaohui1,LIU Wei2

(1.School of Computer and Information,Hohai University,Nanjing 211100;2.Troop 63780,Sanya 572000,China)

Received:2013-05-23 Revised:2013-09-22 Online:2014-11-25 Published:2014-11-25

摘要/Abstract

摘要：

为有效解决行人阴影对行人跟踪产生的影响，提出一种基于团块模型的阴影抑制算法。该算法利用团块结构作为行人模型，通过计算团块的矩特征和团块垂直直方图，借助几何的方法获得疑似阴影区域。然后，将该疑似阴影区域的灰度、方向、中心位置作为参数对整个前景区域进行阴影建模，使得阴影区域更精确。最后，从行人模型中剔除阴影团块达到阴影抑制的效果。经过视频序列测试，本文算法对不同运动方向、不同数量的行人均有良好的阴影抑制效果，表明了该设计方案的有效性和准确性。

关键词: 阴影抑制, 团块直方图, 阴影建模, 行人跟踪

Abstract:

To solve the problem of negative effect of shadows in pedestrian detection,a shadow eliminating algorithm based on blob model is proposed. We first describe the pedestrian model with blob structures,and then compute the blob’s central moments and vertical histogram. With the help of these data and the a geometric method,the suspected shadow region can be obtained.After that,the rough approximation of the shadow region can be further refined through shadow modeling with several features including the intensity,orientation and center position of the shadow region. Finally,shadow blobs can be precisely eliminated from the pedestrian blob model.Experimental results demonstrate that different shadows under different conditions can be eliminated accurately and effectively.

Key words: shadow eliminating;blob histogram;shadow modeling;pedestrian detection

金圣韬1，孟朝晖1，刘伟2. 基于团块模型的行人阴影抑制算法[J]. J4, 2014, 36(11): 2203-2209.

JIN Shengtao1,MENG Zhaohui1,LIU Wei2. A pedestrian shadow eliminating algorithm
based on blob model [J]. J4, 2014, 36(11): 2203-2209.

参考文献［10］

［1］	Cristani M,Delbue A. Human behavior analysis in video surveillance:A social signal processing perspective［J］.Neurocomputing,2013(100):8697.
［2］	Yoneyama A, Yeh C H, Kuo C. Movingcast shadow elimination for robust vehicle extraction based on 2D joint vehicle/shadow models［C］∥Proc of IEEE Conference on Advanced Video and Signal Based Surveillance,2003:229236.
［3］	Salvador E, Cavallaro A, Ebrahimi T. Cast shadow segmentation using invariant color features［J］. Computer Vision and Image Understanding,2004,95(2):238259.
［4］	Salvador E, Cavallaro A, Ebrahimi T. Shadowaware objectbased video processing［J］. IEEE Proceedings on Vision, Image and Signal Processing,2005,152(4):398406.
［5］	Huang Jiabin, Chen Chusong. Moving cast shadow detection using physicsbased features［C］∥Proc of IEEE Conference on Computer Vision and Pattern Recognition,2009:23102317.
［6］	Hampapur A. Robust and efficient foreground analysis for realtime video surveillance［C］∥Proc of IEEE Conference on Computer Vision and Pattern Recognition,2005:11821187.
［7］	Leone A,Distante D.Shadow detection for moving objects based on texture analysis ［J］. Pattern Recognition,2007,40(4):12221233.
［8］	Zhang Wei, Fang Xiangzhong, Xu Yi. Detection of moving cast shadows using image orthogonal transform［C］∥Proc of International Conference on Pattern Recognition, 2006:626629.
［9］	Wang Shupeng, Ji Hongbing. Appearance model based on blobs for object tracking ［J］. Journal of Xidian University,2008,35(5):799804.(in Chinese)
［10］	Shan Yong, Wang Runsheng, Yang Fan. An algorithm of motion detection based on multifeature background model［J］. Computer Engineering & Science,2007,29(8):4042.(in Chinese)
［11］	Wang Xiaoying. A novel fuzzy clustering algorithm for the analysis of axillary lymph node tissue sections ［J］. Applied Intelligence,2007,27(3):237248.
	附中文参考文献：
［9］	王书朋,姬红兵.利用团块模型进行目标跟踪［J］.西安电子科技大学学报(自然版),2008 35(5):799804.
［10］	单勇,王润生,杨凡.基于多特征背景模型的运动目标检测算法［J］.计算机工程与科学,2007,29(8):4042.

[1]	曹玉东, 陈冬昊, 曹睿, 赵朗. 融合Mask R-CNN的在线多目标行人跟踪方法[J]. 计算机工程与科学, 2023, 45(07): 1216-1225.
[2]	李博川1，丁轲2. 结合阴影抑制的混合高斯模型改进算法[J]. J4, 2016, 38(03): 556-561.