基于特征融合的背景差分改进算法

doi:10.16180/j.cnki.issn1007-7820.2020.12.005

摘要/Abstract

摘要：

背景差分法可完整快速地分割出目标图像,但其在背景扰动与光照变化等情况下检测效果不佳。文中提出一种基于特征融合的背景差分改进算法。该算法将时空局部二值模式纹理特征以及颜色特征相融合,同时考虑两特征的置信度和相似性得分得出背景概率,继而进行前景分割,并将当前检测出的背景像素用于背景模板更新,以便更好地解决复杂背景下的目标检测问题。实验结果表明,新算法的检测效果优于其他同类算法,在保持背景差分算法鲁棒性与复杂度的同时,在背景扰动与光照变化等情况下表现出了良好的检测效果。

关键词: 背景差分, 目标检测, 特征融合, 置信度, 背景概率, 模板更新

Abstract:

The background difference method has higher reliability in completely and quickly segmenting the target image, but the detection effect is not good in the case of background disturbance and illumination variation. This study proposes an improved background difference algorithm based on feature fusion. This algorithm fuses spatio-temporal local binary pattern texture features and color features, and consideres the confidence and similarity scores of the two features to obtain the background probability, following by foreground segmentation. The background pixels are used for background template updates, which is designed to better solve the problem of target detection in complex backgrounds. The experimental results show that the detection effect of this algorithm is better than other similar algorithms. While maintaining the robustness and complexity of the background difference algorithm, it displayes good detection effect under the background disturbance and illumination changes.

Key words: background subtraction, target detection, feature fusion, confidence, background probability, template update

中图分类号:

TP391.41

官洪运,苏振涛,汪晨. 基于特征融合的背景差分改进算法[J]. 电子科技, 2020, 33(12): 22-27.

GUAN Hongyun,SU Zhentao,WANG Chen. Improved Background Subtraction Based on Feature Fusion[J]. Electronic Science and Technology, 2020, 33(12): 22-27.

图/表 5

图1

图2

表1

表2

表3

参考文献 16

[1]	方路平, 何杭江, 周国民. 目标检测算法研究综述[J]. 计算机工程与应用, 2018,54(13):11-18.
	Fang Luping, He Hangjiang, Zhou Guomin. Research overview of object detection methods[J]. Computer Engineering and Applications, 2018,54(13):11-18.
[2]	汪欣, 吴薇, 曾照. 基于视频的人脸检测算法[J]. 电子科技, 2019,31(1):1-7.
	Wang Xin, Wu Wei, Zeng Zhao. Video-based face detection algorithm[J]. Electronic Science and Technology, 2019,31(1):1-7.
[3]	Yao J, Odobez J. Multi-layer background subtraction based on color and texture[C]. Minneapolis:IEEE Conference on Computer Vision and Pattern Recognition, 2007.
[4]	Xu J, Ding X Q, Wang S J, et al. Background subtraction based on a combination of texture, color and intensity[C]. Beijing:The Ninth International Conference on Signal Processing, 2008.
[5]	Tuzel O, Porikli F, Meer P. A Bayesian approach to background modeling[C]. San Diego:IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2005.
[6]	Kim K, Chalidabhongse T H, Harwood D, et al. Real-time foreground-background segmentation using codebook model[J]. Real-Time Imaging, 2005,11(3):172-185. doi: 10.1016/j.rti.2004.12.004
[7]	Maddalena L, Petrosino A. A self-organizing approach to background subtraction for visual surveillance applications[J]. IEEE Transactions on Image Processing, 2008,17(7):1168-1177. doi: 10.1109/TIP.2008.924285 pmid: 18586624
[8]	Heikkila M, Pietikainen M. A texture-based method for modeling the background and detecting moving objects[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2006,28(4):657-662. pmid: 16566514
[9]	Liao S C, Zhao G Y, Kellokumpu V, et al. Modeling pixel process with scale invariant local patterns for background subtraction in complex scenes[C]. San Francisco:IEEE Conference on Computer Vision and Pattern Recognition, 2010.
[10]	St-Charles P L, Bilodeau G A, Bergevin R. Flexible background subtraction with self-balanced local sensitivity[C]. Columbus:IEEE Conference on Computer Vision and Pattern Recognition, 2014.
[11]	Zhang S P, Yao H X, Liu S H. Dynamic background modeling and subtraction using spatio-temporal local binary patterns[C]. San Diego:The Fifteenth IEEE International Conference on Image Processing, 2008.
[12]	杜鹃, 吴芬芬. 高斯混合模型的运动目标检测与跟踪算法[J]. 南京理工大学学报, 2017,41(1):41-46.
	Du Juan, Wu Fenfen. Movement target tracking algorithm by using Gaussian mixture model[J]. Journal of Nanjing University of Science and Technology, 2017,41(1):41-46.
[13]	杨大勇, 杨建华, 卢伟基. 于动态阈值的核密度估计前景检测算法[J]. 计算机应用, 2015,37(5):2033-2038.
	Yang Dayong, Yang Jianhua, Lu Wei. Foreground detection algorithm based on dynamic threshold kernel density estimation[J]. Journal of Computer Applications, 2015,37(5):2033-2038.
[14]	Zhang Z B, Yuan X B. An improved PBAS algorithm for dynamic background[J]. Electronic Design Engineering, 2017,25(3):35-40.
[15]	Wang Y, Jodoin P M, Porikli F, et al. CDnet 2014:an expanded change detection benchmark dataset[C]. Columbus:IEEE Conference on Computer Vision and Pattern Recognition Workshops, 2014.
[16]	郭治成, 党建武, 王阳萍, 等. 基于多特征融合的背景建模方法[J]. 光电工程, 2018,45(12):180-186.
	Guo Zhicheng, Dang Jianwu, Wang Yangping, et al. Background modeling method based on multi-feature fusion[J]. Opto-Electronic Engineering, 2018,45(12):180-186.

Method	Blizzard		Cubilce		winterStreet		Badminton		Canoe		Overpass		Average
Method	CT	HT	CT	HT	CT	HT	CT	HT	CT	HT	CT	HT	CT	HT
GMM	145.1	32	101.6	9	133.4	26	113.1	36	43.8	8	111.9	8	107.6	19.8
KDE	101.2	45	118.2	12	203.8	31	129.5	42	38.1	13	140.5	13	121.9	26.0
PBAS	275.7	56	136.4	19	259.5	51	147.6	60	71.6	18	158.9	16	174.9	36.7
本文算法	103.4	41	110.7	10	196.3	29	135.9	44	32.5	11	120.8	9	116.6	24.0

Method	Blizzard		Cubilce		WinterStreet		Badminton		Canoe		Overpass		Average
Method	R	P	R	P	R	P	R	P	R	P	R	P	R	P
GMM	0.81	0.52	0.81	0.59	0.74	0.42	0.79	0.70	0.76	0.86	0.73	0.82	0.77	0.65
KDE	0.78	0.73	0.83	0.55	0.71	0.35	0.73	0.62	0.75	0.83	0.80	0.91	0.76	0.67
PBAS	0.72	0.87	0.86	0.67	0.66	0.38	0.76	0.64	0.86	0.90	0.83	0.92	0.78	0.73
本文算法	0.80	0.74	0.90	0.74	0.75	0.56	0.83	0.85	0.80	0.87	0.87	0.92	0.83	0.78

Method	Blizzard	Cubilce	WinterStreet	Badminton	Canoe	Overpass
GMM	0.63	0.68	0.53	0.74	0.80	0.77
KDE	0.72	0.66	0.47	0.67	0.78	0.85
PBAS	0.79	0.75	0.48	0.69	0.88	0.87
本文算法	0.76	0.81	0.64	0.84	0.83	0.89