基于大气反射-散射模型的复原图像中交通视频车灯检测

汤春明; 曹志升; 林祥清; 肖文娜; 耿磊

doi:10.16383/j.aas.2016.c150485

基于大气反射-散射模型的复原图像中交通视频车灯检测

doi: 10.16383/j.aas.2016.c150485

汤春明^1, ,,
曹志升^1,,
林祥清^1,,
肖文娜^1,,
耿磊^1,2,

1.
天津工业大学电子与信息工程学院天津 300387
2.
天津工业大学智能信息处理技术与系统实验室天津 300387

基金项目:

天津工业大学引进教师科研启动项目 030367

天津市科技支撑计划重点项目 14ZCZDGX00033

天津市第三批三年千人计划项目 62014511

详细信息

作者简介:
曹志升, 天津工业大学电子与信息工程学院硕士研究生.2013年获得天津工业大学电子与信息工程学院电子信息工程学士学位.主要研究方向为图像处理与模式识别, 视频中目标识别与追踪.E-mail:zhishengcao1999@163.co

林祥清, 天津工业大学电子与信息工程学院硕士研究生.2014年获得天津工业大学电子与信息工程学院电子信息科学与技术学士学位.主要研究方向为图像处理, 目标识别与追踪.E-mail:lxiangqing0311@163.com

肖文娜, 天津工业大学电子与信息工程学院硕士研究生. 2014 年获得大连民族大学信息与通信工程学院电子信息工程学士学位. 主要研究方向为目标识别与追踪. E-mail: xwna829@sina.cn

耿磊天津工业大学电子与信息工程学院副教授. 2012 年获天津大学博士学位. 主要研究方向为机器视觉, 图像处理. E-mail: genglei@tjpu.edu.cn

通讯作者:
汤春明, 天津工业大学电子与信息工程学院教授.2006年获哈尔滨工程大学信息与通信工程学院博士学位.2009年于哈尔滨工业大学计算机科学与技术学院博士后流动站出站.主要研究方向为视频中多目标识别与追踪, 图像处理与模式识别, 数据处理与信息挖掘, 材料无损检测技术与系统.本文通信作者.E-mail:tangchunminga@hotmail.com

计量
- 文章访问数: 1916
- HTML全文浏览量: 266
- PDF下载量: 1396
- 被引次数: 0
出版历程
- 收稿日期: 2015-07-28
- 录用日期: 2015-10-19
- 刊出日期: 2016-04-01

Headlights Detection in Traffic Videos Based on Atmospheric Reflection-scattering Model via Reconstructing Restoration Images

TANG Chun-Ming^{1
, ,},
CAO Zhi-Sheng^1
,,
LIN Xiang-Qing^1
,,
XIAO Wen-Na^1
,,
GENG Lei^{1,2
,}

1.
School of Electronic and Information Engineering, Tianjin Polytechnic University, Tianjin 300387
2.
Intelligent Information Processing Technology and Systems Laboratory, Tianjin Polytechnic University, Tianjin 300387

Funds:

Scientific Re-search Starting Project for Introduce Teachers of Tianjin Polytechnic University 030367

Science and Technology Supporting Key Project of Tianjin 14ZCZDGX00033

the Third Thousand Talents Plan of Tianjin over Three Years 62014511

More Information

Author Bio:
Master student at the College of Electronic and Information Engineering, Tianjin Polytechnic University. He received his bachelor degree from Tianjin Polytechnic University in 2013. His research interest covers image processing and pattern recognition, objects identification and tracking in video

Master student at the School of Electronic and Information Engineering, Tianjin Polytechnic University. He received his bachelor degree from Tianjin Polytechnic University in 2014. His research interest covers image and processing, objective recognition and tracking

Master student at the School of Electronic and Information Engineering, Tianjin Polytechnic University. She received her bachelor degree from Dalian Nationality University in 2014. Her research interest covers objects0 recognition and tracking

Associate professor at the School of Electronic and Information Engineering, Tianjin Polytechnic University. He received his Ph. D. degree from Tianjin University in 2012. His research interest covers computer vision and image processing

Corresponding author: TANG Chun-Ming Professor at the School of Electronics and Information Engineering, Tianjin Polytechnic University. She received her Ph. D. degree from the School of Information and Communication Engineering, Harbin Engineering University in 2006. She has been out of the postdoctoral research station of Harbin Institute of Technology in 2009. Her research interest covers multi-objects identification and tracking in video, image processing and pattern recognition, data processing and information mining, techniques and system of material0s nondestructive testing. Corresponding author of this paper

摘要

摘要: 针对夜间复杂照明环境导致车灯检测率低的问题, 提出了一种基于大气反射-散射模型的复原图像中夜间交通视频车灯检测算法. 首先根据漫反射原理抑制路面漫反射光, 在对大气散射模型做了改进之后, 估计了大气散射模型中的大气光, 再根据暗原色先验理论估计环境光, 重新定义透射率, 从而得到了只含有车灯及反射区域的复原图像.为了进一步抑制该复原图像中的强光光晕, 再次利用暗原色先验理论重新估计环境光, 得到最终的复原图像. 最后对复原图像中的所有亮斑根据四类几何特征逐步筛选, 排除视野中的非车灯. 实验结果表明, 该方法在复杂雨雪天气、高密度及高速等不同情况下, 与同类先进算法相比具有较高的检测率, 较低的漏检率和误检率.
- 车灯检测 /
- 复原图像 /
- 大气反射-散射模型 /
- 暗原色先验理论
Abstract: As the detection rate of headlights under complex lighting scenes in nighttime is relatively low, a novel algorithm of headlights detection based on atmospheric reflection-scattering model has been presented. The diffuse reflection light from road surface is firstly suppressed using the diffuse reflection principle. After improving the atmosphere scattering model, the airglow is estimated. The ambient light is then estimated with the dark-channel prior theory and the transmittance is defined again. A restoration image is obtained in which only headlights and reflection regions are contained. In order to suppress the halo produced by the highlights in the restoration image, the ambient light is estimated again with the dark-channel prior theory. All spots are then checked according to their four types of geometric characteristics step by step, non-headlights are finally filtered out. Experimental results show that under complicated weather, such as rain or snow, or under high population, or high-speed, the proposed algorithm has a higher detection rate, lower miss rate and false detection rate, compared with other similar advanced algorithms.
- Headlights detection /
- restoration images /
- atmospheric reflection-scattering model /
- dark-channel prior theory

HTML全文

图 1 光源基本组成框架示意图

Fig. 1 Basic composition diagram of light sources

下载: 全尺寸图片幻灯片

图 2 车灯提取

Fig. 2 Headlights extraction

下载: 全尺寸图片幻灯片

图 3 本文车灯检测算法流程

Fig. 3 Flow chart of headlights extraction algorithm

下载: 全尺寸图片幻灯片

图 4 车灯形状分类

Fig. 4 Headlight shapes classi-cation

下载: 全尺寸图片幻灯片

图 5 文献^[4]检测结果与本文检测结果对比

Fig. 5 Comparison of headlights extraction by our and Zhang et al.^[4]algorithms

下载: 全尺寸图片幻灯片

图 6 相同条件下, 同一视频特殊情况检测结果对

Fig. 6 Comparison of some special cases in the same video and conditions

下载: 全尺寸图片幻灯片

图 7 相同条件下, 不同视频特殊情况检测结果对比

Fig. 7 Comparison of special cases in di®erent videos and same conditions

下载: 全尺寸图片幻灯片

图 8 高密度情况下, 文献^[5]检测结果与本文检测结果对比

Fig. 8 Comparison of headlight extraction by our algorithm and Azimuthally-blur technique^[5] in high intensity

下载: 全尺寸图片幻灯片

图 9 雪后情况下, 文献^[5] 检测结果与本文检测结果对比

Fig. 9 Comparison of headlight extraction by our algorithm and Azimuthally-blur technique^[5] after sno

下载: 全尺寸图片幻灯片

图 10 雨后情况下, 文献^[5] 检测结果与本文检测结果对比

Fig. 10 Comparison of headlight extraction by our algorithm and Azimuthally-blur technique^[5] after rain

下载: 全尺寸图片幻灯片

图 11 高速情况下, 文献^[5] 检测结果与本文检测结果对比

Fig. 11 Comparison of headlight extraction by our algorithm and Azimuthally-blur technique^[5] in high speed

下载: 全尺寸图片幻灯片

图 12 车灯配对及车辆轨迹

Fig. 12 Headlights pairing and vehicles tracking

下载: 全尺寸图片幻灯片

表 1 检测结果与本文检测结果对比

Table 1 Comparative results of headlights detection by our and Zhang et al.^[4] algorithm

视频	实际车灯	检测车灯TP	漏检FN	误检FP	检测率E
文献^[4]	44	44	0	5	89.80%
本文算法	44	44	0	2	95.65%

下载: 导出CSV

表 2 四种特殊情况下车灯检测结果

Table 2 Headlight detection under four special cases

图像		实际车灯	检测车灯TP		漏检FN		误检FP		检测率E (%)
图像		实际车灯	文献^[5]	本文	文献^[5]	本文	文献^[5]	本文	文献^[5]	本文
高密度	(a)	23	19	22	3	1	0	0	86.36	95.65
	(b)	21	15	20	6	1	0	1	71.43	90.91
	(c)	17	15	15	2	2	1	0	83.33	88.24
	(d)	12	10	11	2	1	0	0	83.33	91.67
雪后	(a)	14	14	14	0	0	3	1	82.35	93.33
	(b)	10	10	10	0	0	2	1	83.33	90.91
	(c)	12	10	10	2	2	2	1	71.43	76.92
	(d)	8	6	6	2	2	2	1	60	66.67
雨后	(a)	13	10	13	3	0	0	0	76.92	100
	(b)	13	11	13	2	0	0	0	84.62	100
	(c)	10	9	10	1	0	0	0	90	100
	(d)	11	9	11	2	0	0	1	81.82	91.67
高速	(a)	15	13	14	2	1	0	1	86.67	87.5
	(b)	11	9	10	2	1	0	0	81.82	90.91
	(c)	22	20	21	2	1	1	1	86.96	91.30
	(d)	22	21	22	1	0	1	1	91.30	95.45

下载: 导出CSV

表 3 车灯检测结果

Table 3 Headlight detection results

视频序列	参数	序列1	序列2	序列3	序列4	序列5	序列6	序列7	序列8	序列9
视频序列	实际车灯	66	92	52	52	116	212	76	144	105
文献^[5]	TP	64	87	51	51	114	208	75	135	99
	FN+FP	2+8	5+6	1+4	1+12	2+12	4+16	1+6	9+1	6+1
	E (%)	86.49	88.78	91.07	79.69	90.48	91.23	92.68	93.10	93.40
	FNR (%)	2.70	5.10	1.79	1.56	1.56	1.75	1.22	6.21	5.66
	FPR (%)	10.81	6.12	7.14	18.75	9.36	7.02	7.32	0.69	0.94
本文算法	TP	66	88	51	50	110	205	74	136	96
	FN+FP	0+4	3+5	1+3	2+8	6+8	7+12	2+2	6+3	4+2
	E (%)	94.29	91.67	94.44	83.33	93.22	91.52	94.87	93.79	94.12
	FNR (%)	0	3.13	1.82	3.33	4.84	3.13	2.56	4.14	3.92
	FPR (%)	5.71	5.21	5.45	13.33	6.45	5.36	2.56	2.07	1.96

下载: 导出CSV

参考文献(17)

[1]	Zou Q, Ling H B, Luo S W, Huang Y P, Mei T. Robust nighttime vehicle detection by tracking and grouping headlights. IEEE Transactions on Intelligent Transportation Systems, 2015, 16(5): 2838-2849 doi: 10.1109/TITS.2015.2425229
[2]	Hajimolahoseini H, Soltanian-Zadeh H, Amirfattahi R. Robust vehicle tracking algorithm for nighttime videos captured by fixed cameras in highly reflective environments. IET Computer Vision, 2014, 8(6): 535-544 doi: 10.1049/iet-cvi.2013.0267
[3]	Chen Y L, Wu B F, Huang H Y, Fan C J. A real-time vision system for nighttime vehicle detection and traffic surveillance. IEEE Transactions on Industrial Electronics, 2011, 58(5): 2030-2044 doi: 10.1109/TIE.2010.2055771
[4]	Zhang W, Wu Q M J, Wang G H, You X G. Tracking and pairing vehicle headlight in night scenes. IEEE Transactions on Intelligent Transportation Systems, 2012, 13(1): 140-153 doi: 10.1109/TITS.2011.2165338
[5]	Tang C M, Hussain A. Robust vehicle surveillance in night traffic videos using an Azimuthally blur technique. IEEE Transactions on Vehicular Technology, 2015, 64(10): 4432-4440 doi: 10.1109/TVT.2014.2371067
[6]	汤春明, 聂美玲, 丁腾飞, 浩欢飞, 韩旭. 基于最小特征匹配代价的夜间车辆检测和追踪. 计算机应用与软件, 2015, 32(4): 292-296 http://www.cnki.com.cn/Article/CJFDTOTAL-JYRJ201504069.htm Tang Chun-Ming, Nie Mei-Ling, Ding Teng-Fei, Hao Huan-Fei, Han Xu. Nighttime vehicle detection and tracking based on minimum feature matching cost. Computer Applications and Software, 2015, 32(4): 292-296 http://www.cnki.com.cn/Article/CJFDTOTAL-JYRJ201504069.htm
[7]	Rubio J C, Serrat J, Lopez A M, Ponsa D. Multiple-target tracking for intelligent headlights control. IEEE Transactions on Intelligent Transportation Systems, 2012, 13(2): 594-605 doi: 10.1109/TITS.2011.2175219
[8]	Jazayeri A, Cai H Y, Zheng J Y, Tuceryan M. Vehicle detection and tracking in car video based on motion model. IEEE Transactions on Intelligent Transportation Systems, 2011, 12(2): 583-595 doi: 10.1109/TITS.2011.2113340
[9]	O'Malley R, Jones E, Glavin M. Rear-lamp vehicle detection and tracking in low-exposure color video for night conditions. IEEE Transactions on Intelligent Transportation Systems, 2010, 11(2): 453-462 doi: 10.1109/TITS.2010.2045375
[10]	齐保军. 大气散射模型在雾天图像增强和海面小目标检测中的应用 [博士学位论文], 国防科学技术大学, 中国, 2013. Qi Bao-Jun. The Application of Atmospheric Scattering Model in Image Contrast Enhancement and Surface of the Small Target Detection [Ph.D. dissertation], National University of Defense Technology, China, 2013.
[11]	Tan R T, Ikeuchi K. Separating reflection components of textured surfaces using a single image. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2005, 27(2): 178-193 doi: 10.1109/TPAMI.2005.36
[12]	Bouguer P. Traité d'Optique Sur La Gradation De La Lumiére, 1729. 93-115
[13]	He K M, Sun J, Tang X O. Single image haze removal using dark channel prior. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2011, 33(12): 2341-2353 doi: 10.1109/TPAMI.2010.168
[14]	Sneath P H A. Numerical Taxonomy: The Principles and Practice of Numerical Classification. New York, USA: W H Freeman Co, 1973. 53-85
[15]	Kosaka N, Ohashi G. Vision-based nighttime vehicle detection using CenSurE and SVM. IEEE Transactions on Intelligent Transportation Systems, 2015, 16(5): 2599-2608 doi: 10.1109/TITS.2015.2413971
[16]	吴迪, 朱青松. 图像去雾的最新研究进展. 自动化学报, 2015, 41(2): 221-239 http://www.aas.net.cn/CN/abstract/abstract18603.shtml Wu Di, Zhu Qing-Song. The latest research progress of image dehazing. Acta Automatica Sinica, 2015, 41(2): 221-239 http://www.aas.net.cn/CN/abstract/abstract18603.shtml
[17]	刘海波, 杨杰, 吴正平, 张庆年, 邓勇. 基于暗通道先验和Retinex理论的快速单幅图像去雾方法. 自动化学报, 2015, 41(7): 1264-1273 http://www.aas.net.cn/CN/abstract/abstract18700.shtml Liu Hai-Bo, Yang Jie, Wu Zheng-Ping, Zhang Qing-Nian, Deng Yong. A fast single image dehazing method based on dark channel prior and Retinex theory. Acta Automatica Sinica, 2015, 41(7): 1264-1273 http://www.aas.net.cn/CN/abstract/abstract18700.shtml