Quickbird遥感影像的车辆自动检测与运动参数估计-论文
来源:易妖游戏网
第3O卷第4期 液 晶与显示 Chinese Journa1 of Liquid Crystals and Displays V01.3O No.4 Aug.2015 2015年8月 文章编号:1007—2780(2015)04—0687—08 Quickbird遥感影像的车辆自动检测 与运动参数估计 张博研 ,李广泽,武星星 (中国科学院长春光学精密机械与物理研究所,吉林长春130033) 摘要:遥感图像的车辆目标提取与运动参数估计在交通管理、战场态势分析等领域具有广阔的应用前景,但目前相关算 法均需要人工参与或借助GIS信息,针对上述问题提出了一种基于计算机视觉的全自动车辆检测与运动参数估计算法。 分析了Quickbird全色与多光谱传感器的焦平面结构特征以及该结构造成的“鬼影”现象;针对全色与多光谱遥感影像的 分辨率高、光谱信息丰富的特点,利用植被指数归一化、图像分割、形态学灰度重建等图像处理过程,实现了全色图像中 运动车辆的自动检测,在此基础上检测低分辨率的多光谱图像中的目标。利用全色与多光谱图像的成像时间差估计运 动参数。在Quickbird遥感影像的验证实验中充分证明了算法的可行性与正确性。 关键词:全色图像;多光谱图像;车辆检测;运动参数估计;形态学重建 文献标识码:A doi:10.3788/YJYXS20153004.0687 中图分类号:TP79 Speed estimation and automatic detection of moving vehicle from Quickbird satellite images ZHANG Bo—yan ,LI Guang—ze,WU Xing—xing (Changchun Institute of Optics,Fine Mechanics and Physics,Chinese Academy of Science, Changchun 130033,China) Abstract:Velocity estimation and vehicle detection from satellite images is widely used in the domain of traffic monitoring,battlefield analysis etc.But most of the algorithms in this area are either manual or incorporating ancillary data.A new method has been developed to extract vehicles automatically and determine their speeds based on computer vision.First of all,the“Ghost”phenomenon of moving tar— gets is analyzed by considering the structure of Quickbird panchromatic and muhispectral focal plane. Normalized differential vegetation index,image segmentation and morph0logical gray scale reconstruc— tion are combined to realize accurate vehicles detection in panchromatic image;subsequently image matching is applied to extract vehicles in the multispectral image based on the vehicle positions in pan— chromatic image.The speed can be calculated by using the vehicle extraction results and time intervals between panchromatic and multispectral images.Finally,this approach was tested on several images of Quickbird covering different backgrounds and can obtain a detection rate as high as 90 .The 收稿日期:2014-10-21;修订日期:2014一ll一15. 基金项目:国家自然科学基金资助项目(No.61108066);吉林省科技发展计划项目(No.2O13O101028jc) *通信联系人,E—mail:boyanl021@163.corn 第4期 张博研,等:Quickbird遥感影像的车辆自动检测与运动参数估计 693 表3场景2的车辆检测与速度参数估计结果 Tab.3 Speed estimation and vehicle detection results of scene2 表4场景3的车辆检测与速度参数估计结果 Tab.4 Speed estimation and vehicle detection results of scene3 成像时本身存在一个推扫速度,图像中的速度为 目标自身的运动速度与推扫速度的合成结果,也 为算法的进一步改进提供了新的方向。 图10(b)为孟加拉国的一条乡村道路,通过 表3中的数据可见,车辆目标的平均速度在6O 5 结 论 km/h左右,符合车辆在乡村道路上的行驶情况; 图10(C)为印度契尔卡湖附近高速公路截 图,由于该地区没有车辆速度上限,所以车辆速度 均在116。72 km/h以上,如表4所示。 通过上述验证实验可知,对于不同区域背景 的Quickbird遥感影像,通过本算法检测车辆目 标不仅不需要人工参与,而且目标的正确探测率 针对Quickbird全色影像的分辨率高、多光 谱图像光谱信息丰富的特点,提出了一种全自动 的运动车辆检测算法,不仅实现了较好的车辆检 测结果,而且突破了以往算法需要借助人工帮助 或者GIS信息的,然后结合全色与多光谱图 像的成像时间间隔完成了运动参数估计,最后对 高,虚警率低,计算的运动参数值符合实际情况, 具有相当高的参考价值,但是有些车辆的运动方 向与道路的方向稍有偏差,这是因为TDI CCD在 Quickbird遥感影像的验证试验证明了该算法的 普适性强,正确探测率高,虚警率低,是一种性能 值得推广的运动车辆检测与运动参数估计算法。 参 考 文 献: [1]代科学,李国辉,涂丹,等.监控视频运动目标检测减背景技术的研究现状与展望EJ].中国国象图形学报,2006,ll (7):919-927. Dai K X,Li G H,Tu D,et a1.Prospects and current studies on background subtraction Techniques for moving ob— jects detection from durveillance video[J].Journal of Image and Graphics,2006,11(7):919—927.(in Chinese) [2]ChengJ,Yang J,Zhou Y,et a1.Flexible background mixture models for foreground segmentation口].Image and Vision Computing,2006,24(5):473—482. [33 Dickinson P,Hunter A,Appiah K.A spatially distributed model for foreground segmentation EJ].Image and 一 sion Computing,2009,27(9):1326—1335. [4]Zha Y F,Bi D Y,Yang Y.Learning complex background by multi-scale discriminative model EJ].Pattern Recogni— tion Letters,2009,30(11):1003—1014. Es] Bicego M,Castellani U,Murino V.A hidden Markov model approach for appearance—based 3D object recognition 694 液晶与显示 第3O卷 [J].Pattern Recognition Letters,2005,26(16):2588—2599. [6] HuJ S,Juan C W,Wang J J.A spatial—color mean—shift object tracking algorithm with scale and orientation estima— tion EJ].PatternRecognition Letters,2008,29(16):2165—2173. [7] Tao J,Yu W A.Preliminary study on imaging time difference among bands of WorldView~2 and its potential appli— cations[C]//GeDsc c and Remote Sensing Symposium(IGARSS),2011 IEEE International,2011 1 198—200. [8] 张小建,徐慧.基于视频处理的运动车辆检测算法的研究[J].液晶与显示,2012,27(1):108 113. Zhang X J,xu H.Moving vehicle detection algorithm based on video processing[J].Chinese Journal of Liquid Crystals and Display,2012,27(1):108—113.(in Chinese) E9] 张红颖,胡正.CenSurE特征和时空信息相结合的运动目标检测[J].光学精密工程,2013,21(9):2452—2463 Zhang H Y,Hu Z.Moving object detection in combination of CenSurE and spatial—temporal information EJ].Edito— rial Office of Optics and Precision Engineeri.2013,21(9):2452—2463.(in Chinese) ehi B,Zhang Y,Zhong M.Automatic moving vehicles information extraction from single—pass WorldView一2 im— [1o] Salagery l-J].Selected Topics in Applied Earth Observations and Remote Sensing,IEEE,2012,5(1):135—145. [11] Larsen S O,Koren H,Solberg R.Trafficmonitoring using very-high-resolution satellite Imagery[J].Photogram metric Engineering and Remote Sensing,2009,75(7):859—869. ong Z,Zhang Y.An initial study on vehicle information extraction from single pass QuickBird satellite imagery [12] Xi[J].Journal of Photogrammetric Engineering&Remote Sensing,2008,74(11):1401—1412. [13] Pesaresi M,Gutjahr K H,Pagot E.Estimating the velocity and direction of moving targets using a single optical VHR satellite sensor image[J].International Journal of Remote Sensing,2008,29(4):122卜1228. n X,Davis C H.Vehicle detection from high—resolution satellite imagery using morphological shared—weight neu— [14] Jiral networks[J].Image s.Compute,2007,25(9):1422—1431. [15] 王灿进,孙涛,王挺峰,等.基于轮廓特征的神经网络目标识别研究EJ].液晶与显示,2013,28(4):641—648. Wang C J,Sun T,Wang T F,et a1.Target recognition using neural network based on contour features[J].Chi— nese Journal of Liquid Crystals and Display,2013,28(4):641—648.(in Chinese) [16] ZhengH,Li L,An artificial immune approach for vehicle detection from high resolution space imagery[J].Int.J. Comput.Sci.Netw.Secur,2007,7(2):67—72. [17] 余勇,郑宏.基于形态神经网络的高分辨率卫星影像车辆检测口].哈尔滨工程大学学报,2006,27(B07):189—193. Yu Y,Zheng H.Vehicle detection from high resolution satellite imagery based on the morphological neural net— work[J].Journal of Harbin Engineering University,2006,27(B07):189—193.(in Chinese) F。Vu T T.Automated vehicle extraction and speed determination from QuickBird satellite ima— [18] Liu W,Yamazakiges[J].IEEE,2010,4(1):75-82. [19] Leitloff J,Hinz S,Stilla U.Vehicle detection in very high resolution satellite images of city areas[J].IEEE, 2010,48(7):2795-2806. [2O] Vincent L.Morphological grayscale reconstruction in image analysis:Applications and efficient algorithms[J]. IEEE,1993,2(2):176—2O1. E21] Dirk C.Borghys,M Idrissa,Michal S,et a1.Fusion of multispectral and stereo information for unsupervised tar— get detection in VHR airborne data[c].SPIE,2013,8745:874514—1—874514—12. 作者简介:张博研(1987一),女,吉林白城人,硕士,研究实习员,主要从事立体测绘和图像处理方向的研究。E-mail:boy an1021@163.corn 李广泽(1976一),男,吉林长春人,博士,研究员,主要从事空间遥感相机嵌入式控制系统设计与图像处理方 向的研究。E—mail:Lgzandrew@tom.corn 武星星(1980一),男,河南洛阳人,博士,副研究员,主要从事空间光电技术及智能控制方向的研究。E—mail: starglare@126.corn