基于卷积特征选择的红外目标跟踪

钱琨; 杨俊彦; 余跃; 赵东; 荣生辉

doi:10.11884/HPLPB201931.190133

基于卷积特征选择的红外目标跟踪

doi: 10.11884/HPLPB201931.190133

钱琨^{1, 2,},
杨俊彦^{1, 2},
余跃^{1, 2},
赵东³,
荣生辉⁴

1.
上海航天控制技术研究所, 上海 201109
2.
中国航天科技集团有限公司红外探测技术研发中心, 上海 201109
3.
西安电子科技大学物理与光电工程学院, 西安 710071
4.
中国海洋大学信息科学与工程学院, 山东青岛 266100

基金项目:

国家自然科学基金项目 51801142

中央高校基本科研业务费项目 201813019

中国博士后科学基金面上项目 2019M652472

详细信息

作者简介:
钱琨(1990—)，男，博士，工程师，主要从事红外目标跟踪方面的研究; jsdtqk@163.com

中图分类号: TP391.4
计量
- 文章访问数: 1076
- HTML全文浏览量: 234
- PDF下载量: 48
- 被引次数: 0
出版历程
- 收稿日期: 2019-04-17
- 修回日期: 2019-06-06
- 刊出日期: 2019-09-15

Infrared target tracking based on selective convolution features

Qian Kun^{1, 2
,},
Yang Junyan^{1, 2},
Yu Yue^{1, 2},
Zhao Dong³,
Rong Shenghui⁴

1.
Shanghai Institute of Spaceflight Control Technology, Shanghai 201109, China
2.
Infrared Detection Technology Research and Development Center of China Aerospace Science and Technology Corporation, Shanghai 201109, China
3.
School of Physics and Optoelectronic Engineering, Xidian University, Xi'an 710071, China
4.
School of Information Science and Engineering, Ocean University of China, Qingdao 266100, China

摘要

摘要: 对红外图像中的目标跟踪时，复杂的背景信息以及目标像素数较少等因素增加了红外目标跟踪难度，目标区域的图像块缺乏特征信息使得普通跟踪算法较易产生跟踪偏移问题。为解决此问题，提出了一种基于粒子滤波框架下的卷积特征选择的红外目标跟踪算法。首先，在初始目标块上提取少量图像块作为滤波器，进而获得表征能力更强的卷积特征。然后，采用在线提升算法对该特征进行选择，增加跟踪算法的精度和执行效率。最后，将贝叶斯分类器的响应作为粒子权值估计出目标状态。实验结果验证了所提算法的跟踪性能优于其他几种传统算法。
- 红外图像 /
- 目标跟踪 /
- 弱小目标 /
- 卷积特征 /
- 提升 /
- 粒子滤波
Abstract: Infrared target tracking is heavily influenced by illumination variation, small size and complex background, and the lack of target information makes the algorithm lose targets easily. Therefore, an algorithm based on convolution features and feature selection method is presented in this paper to track IR targets. First, several filters in target patches of the first frame are used to obtain strong features. Then, the boosting method is utilized to train the features with redundant information, thus, the algorithm performance of accuracy and execution efficiency can be improved. Finally, particle weights are represented by the response of the native Bayes classifier. Experimental results show that the presented algorithm obtains good performance.
- infrared image /
- target tracking /
- dim-small target /
- convolutional feature /
- boosting /
- particle filter

HTML全文

图 1 小目标的卷积特征

Figure 1. Convolutional feature of small targets

下载: 全尺寸图片幻灯片

图 2 面目标的卷积特征

Figure 2. Convolutional features of area targets

下载: 全尺寸图片幻灯片

图 3 跟踪算法的流程图

Figure 3. Flowchart of the proposed tracking algorithm

下载: 全尺寸图片幻灯片

图 4 Girl-Griffith序列上的跟踪结果

Figure 4. Tracking results on Girl-Griffith

下载: 全尺寸图片幻灯片

图 5 People-Griffith序列上的跟踪结果

Figure 5. Tracking results on People-Griffith

下载: 全尺寸图片幻灯片

图 6 Car-Griffith序列上的跟踪结果

Figure 6. Tracking results on Gar-Griffith

下载: 全尺寸图片幻灯片

图 7 Plane-Griffith序列上的跟踪结果

Figure 7. Tracking results on Plane-Griffith

下载: 全尺寸图片幻灯片

图 8 Plane1序列上的跟踪结果

Figure 8. Tracking results on Plane1

下载: 全尺寸图片幻灯片

图 9 Plane2序列上的跟踪结果

Figure 9. Tracking results on Plane2

下载: 全尺寸图片幻灯片

图 10 六组序列的所有算法的中心位置误差

Figure 10. Center location error of all trackers over the six sequences

下载: 全尺寸图片幻灯片

表 1 六组红外序列描述

Table 1. Description of six IR sequences

sequence	frame	image size	target size	description
Girl-Griffith	150	640×512	40×110	people, around the tree and stone
People-Griffith	200	640×512	35×75	people, in front of the jungle
Car-Griffith	147	640×512	40×35	car, on the road
Plane-Griffith	239	640×512	15×8	plane, influenced by the cloud
Plane1	86	256×256	2×2~4×4	plane, bright cloud
Plane2	120	256×256	3×3~5×5	plane, strong edge

下载: 导出CSV

表 2 跟踪性能对比

Table 2. Comparison of all tracking algorithms

CLE							average SR	FPS
sequence	Girl-Griffith	People-Griffith	Car-Griffith	Plane-Griffith	Plane1	Plane2	average SR	FPS
CT	0.06	0.86	0.40	0.39	1.00	0.54	0.54	40
L1	0.59	0.20	0.05	0.96	0.65	0.58	0.51	12
TMT	0.33	0.38	0.54	0.95	0.77	0.08	0.51	24
STC	0.23	0.99	0.71	0.32	0.10	0.34	0.45	100
MSPF	0.47	0.04	0.76	0.84	0.15	0.65	0.49	20
CNB	1.00	0.97	1.00	0.97	1.00	0.95	0.98	19

下载: 导出CSV

参考文献(16)

[1]	Kim S, Lee J. Scale invariant small target detection by optimizing signal-to-clutter ratio in heterogeneous background for infrared search and track[J]. Pattern Recognition, 2012, 45(1): 393-406. doi: 10.1016/j.patcog.2011.06.009
[2]	Wang Xin, Ning Chen, Xu Lizhong. Spatiotemporal difference-of-Gaussians filters for robust infrared small target tracking in various complex scenes[J]. Applied Optics, 2015, 54(7): 1573-1586. doi: 10.1364/AO.54.001573
[3]	张素, 安博文, 潘胜达. 基于时空非局部相似性的海上红外目标检测[J]. 光子学报, 2018, 47: 1110001. https://www.cnki.com.cn/Article/CJFDTOTAL-GZXB201811031.htm Zhang Su, An Bowen, Pan Shengda. Infrared dim target detection based on temporal-spatial non-local similarity. Acta Photonica Sinica, 2018, 47: 1110001 https://www.cnki.com.cn/Article/CJFDTOTAL-GZXB201811031.htm
[4]	董维科, 张建奇, 刘德连, 等. 基于运动方向估计的管道滤波算法[J]. 光子学报, 2013, 42(4): 471-474. https://www.cnki.com.cn/Article/CJFDTOTAL-GZXB201304018.htm Dong Weike, Zhang Jianqi, Liu Delian, et al. Pipeline filter algorithm based on movement direction estimation. Acta Photonica Sinica, 2013, 42(4): 471-474 https://www.cnki.com.cn/Article/CJFDTOTAL-GZXB201304018.htm
[5]	程建, 周越, 蔡念, 等. 基于粒子滤波的红外目标跟踪[J]. 红外与毫米波学报, 2006, 25(2): 113-117. doi: 10.3321/j.issn:1001-9014.2006.02.008 Chen Jian, Zhou Yue, Cai Nian, et al. Infrared object tracking based on particle filter. Journal of Infrared and Millimeter Waves, 2006, 25(2): 113-117 doi: 10.3321/j.issn:1001-9014.2006.02.008
[6]	He Yujie, Li Min, Zhang Jinli, et al. Infrared target tracking via weighted correlation filter[J]. Infrared Physics and Technology, 2015, 73: 103-114. doi: 10.1016/j.infrared.2015.09.010
[7]	赵东, 周慧鑫, 于君娜, 等. 复杂天空背景下的红外弱小目标跟踪[J]. 强激光与粒子束, 2018, 30: 061002. doi: 10.11884/HPLPB201830.170511 Zhao Dong, Zhou Huixin, Yu Junna, et al. Tracking of infrared dim small target in complex sky background. High Power Laser and Particle Beams, 2018, 30: 061002 doi: 10.11884/HPLPB201830.170511
[8]	王晓卫, 王旭东, 贺明. 基于直方图比的背景加权的Mean Shift目标跟踪算法[J]. 强激光与粒子束, 2016, 28: 051001. doi: 10.11884/HPLPB201628.051001 Wang Xiaowei, Wang Xudong, He Ming. Target tracking algorithm based on Mean Shift and histogram ratio background weighted. High Power Laser and Particle Beams, 2016, 28: 051001 doi: 10.11884/HPLPB201628.051001
[9]	Zhang Kaihua, Zhang Lei, Yang M H. Fast compressive tracking[J]. IEEE Trans Pattern Analysis and Machine Intelligence, 2014, 36(10): 2002-2015. doi: 10.1109/TPAMI.2014.2315808
[10]	Babenko B, Yang M, Belongie S. Robust object tracking with online multiple instance learning[J]. IEEE Trans Pattern Analysis and Machine Intelligence, 2011, 33(8): 1619-1632. doi: 10.1109/TPAMI.2010.226
[11]	Henriques J F, Caseiro R, Martins P, et al. High-speed tracking with kernelized correlation filters[J]. IEEE Trans Pattern Analysis and Machine Intelligence, 2015, 37(3): 583-596. doi: 10.1109/TPAMI.2014.2345390
[12]	Qiao Huang, Yang Jie. A multistage target tracker in IR image sequences[J]. Infrared Physics and Technology, 2014, 65: 122-128. doi: 10.1016/j.infrared.2014.03.005
[13]	Li Ying, Li Pengcheng, Shen Qiang. Real-time infrared target tracking based on l₁ minimization and compressive features[J]. Applied Optics, 2014, 53(28): 6518-6526. doi: 10.1364/AO.53.006518
[14]	Wang Li, Liu Ting, Wang Gang, et al. Video tracking using learned hierarchical features[J]. IEEE Trans Image Processing, 2015, 24(4): 1424-1435. doi: 10.1109/TIP.2015.2403231
[15]	Zhang Kaihua, Liu Qingshan, Wu Yi, et al. Robust visual tracking via convolutional networks without training[J]. IEEE Trans Image Processing, 2016, 25(4): 1779-1792.
[16]	陈皓, 马彩文, 陈岳承, 等. 基于灰度统计的快速模板匹配算法[J]. 光子学报, 2009, 38(6): 1586-1589. https://www.cnki.com.cn/Article/CJFDTOTAL-GZXB200906065.htm Chen Hao, Ma Caiwen, Chen Yuecheng, et al. Fast template matching arithmetic based on image gray statistic. Acta Photonica Sinica, 2009, 38(6): 1586-1589 https://www.cnki.com.cn/Article/CJFDTOTAL-GZXB200906065.htm