摘要
基于远红外成像的行人检测已成为计算机视觉与模式识别领域的研究热点之一，远红外图像不依赖于场景的光照条件，反映的是具有不同热辐射率的场景目标的表面温度分布，能够捕捉黑暗和烟雾等环境中的行人目标，故基于远红外成像的夜间行人检测在汽车辅助驾驶系统和智能监控系统中具有重要的应用前景。由于行人目标的非刚体性质和较强的运动随意性，其外观模式通常呈现出复杂多变和尺度不一等特点，导致其具有较高的类内多样性；与可见光图像相比，远红外图像中的行人目标还具有纹理信息较少、分辨率较低的特点，因此，基于远红外成像的行人检测是一个极具挑战性的研究课题。
本文围绕面向汽车辅助驾驶的夜间行人检测问题展开研究，基于配备单目摄像头的车载平台，研究解决保障检测系统实时性、准确性、适合于变化场景的行人检测问题，涉及候选区域（Regions of Interest, ROIs）提取方法、远红外行人描述特征的提取方法、行人识别方法等方面关键技术的研究。论文的主要贡献如下：
1）提出一种基于概率模板匹配的远红外行人检测方法。根据行人的运动朝向建立多尺度概率模板，缓解因外观模式引起的行人类内方差较大的问题；进一步将目标跟踪/多帧校验方法融入概率模板匹配的过程，综合利用行人在多帧图像之间出现的连续性和检测结果的一致性滤除不稳定的误检模式，并填补部分由于ROIs提取精度不高所造成的漏检行人。实验表明该方法能够较好地保障检测系统的实时性，与基于行人步态模式的概率模板构建方法相比，文中方法归纳行人外观模式的能力更强。
2）基于统计学习的识别框架，提出熵加权方向梯度直方图特征（Entropy WeightedHistograms of Oriented Gradients, EWHOG）描述远红外行人，综合了所描述目标的局部形状信息和局部梯度分布的随机信息，确保其局部形状能够更好地被局部密集像素梯度或边缘方向表示；为了解决因成像尺度不一等因素引起的目标类内方差较大的问题，提出基于EWHOG特征的三分支结构支持向量机（Support Vector Machine, SVM）行人识别方法，并利用快速分类支持向量机（Fast Classification Support Vector Machine,FCSVM）对获得的支持向量进行优化，从而约简识别环节所需要的计算和存储开销；根据远红外行人头部及其与周围背景之间灰度分布的差异性，提出进一步抑制误检目标的行人头部校验方法。实验表明：EWHOG特征能有效区分远红外行人；快速分类方案以轻微降低行人识别准确率为代价，保证检测系统运行的实时性，在市区和郊区场景中均获取了较好的检测性能。
3）针对行人检测本质上属于“稀有事件检测”问题的特点，从ROIs提取的角度出发，提出一种基于像素梯度的垂直投影方法，根据远红外图像中天空与路面等背景区域通常具有大范围高灰度同质性的特点，利用图像梯度信息对可能包含行人的竖直带状图像区域进行初定位，避免对整幅输入图像进行搜索；实验表明该方法能够提高ROIs提取阶段的搜索效率，并能够抑制部分仅包含背景目标的候选区域。在行人识别阶段，将图像空间金字塔表示方法融入EWHOG特征的提取过程，在多层图像片（cell）划分方式下，利用局部方向梯度直方图的熵加权分布特性及其全局结构信息表征远红外行人，提出了金字塔熵加权方向梯度直方图（Pyramid Entropy Weighted Histograms of OrientedGradients, PEWHOG）特征；鉴于PEWHOG特征属于直方图统计特征，利用基于直方图交叉核（Histogram Intersection Kernel, HIK）的SVM分类器实现行人识别；针对收集具有代表性的训练数据较为困难、行人分类器的预测性能依赖于初始训练数据的问题，提出基于bootstrapping和提前终止策略的离线训练机制。
4）训练数据与测试数据之间通常存在不可避免的数据分布差异性，这可能导致多数基于传统统计学习的行人检测方法在场景因素变化较大时表现得不够理想。针对这个问题，提出一种Boosting-style的归纳迁移学习算法DTLBoost，用于高效且有效地处理变化场景中的远红外行人检测问题。显式地定义成员分类器对训练数据的预测不一致性程度，将其融入DTLBoost算法的样本权重更新规则，从辅助训练数据中选择具有正迁移能力的数据，并鼓励不同成员分类器学习目标训练数据的不同部分或方面。最后在本文收集的数据集和OSU远红外行人数据集上，通过面向行人识别和行人检测问题的扩展实验评估了该方法的性能，实验结果表明该方法在新场景和变化视角场景中均具有较好的行人检测性能。
Pedestrian detection based on far-infrared (FIR) imageries has become a hot spot incomputer vision and pattern recognition community. FIR imageries capture the targets withdifferent distribution of surface temperature and thermal radiation emissivity and do notdepend on the illumination conditions, which makes it suitable to capture pedestrians indarkness and scenarios permeated with smoke. So it gains important potential in automotivedriver assistance systems and transportation video surveillance in night time scenarios. Thewide variety of possible appearances and scales of pedestrians caused by their non-rigidcharacteristic and high arbitrariness of motions usually leads to higher within-class variance.And compared with the imageries in visible spectrum, pedestrians in FIR imageries alsopresent as blur targets with lower resolution and less texture information. Therefore,pedestrian detection based on FIR imageries is a challenging task.
This dissertation focuses on the issues of night time pedestrian detection for automotivedriver assistance systems uisng monocular FIR camera, aiming at (1) guaranteeing reliableperformance for automotive applications, with both real time implementation and highdetection accuracy;(2) dealing with pedestrian detection across unseen scenarios and newviewpoints. The main contents refer to the extraction of regions of interest (ROIs), featurerepresentation for FIR pedestrians and the framework of pedestrian recognition, which can besummarized as follows:
1) A night time pedestrian detection method is proposed based on probabilistic templatematching, where the multi-scale probabilistic templates are established according to themoving directions of pedestrians and employ to recognize the potential pedestrians. Theprobabilistic templates alleviate the large within-class variability of pedestrians caused by thechanging appearance and thus improve the accuracy for describing appearance of pedestrians.Due to the characteristic of detection agreement of pedestrians among several successiveframes, an object tracking and multi-frame validation module is integrated in templatesmatching to suppress some false detection and fill the detection gap caused by the inaccurateextracted ROIs. The experimental results demonstrate that the proposed method meetsreal-time implementation criteria and the resulting probabilistic templates guarantees higheraccuracy for describing pedestrians’ appearance, compared to the ones based on gait patternsof pedestrians.
2) Following a learing-based detection framework, we first propose entropy weightedhistograms of oriented gradients (EWHOG) to describe FIR pedestrians effectively. Considering both the information of local object shape and microdistributed chaotic degreesof local oriented graident distribution, EWHOG aims to pay more emphasis on thedistribution of local intensity gradients provided by local object shape. To reduce thewithin-class variance of objects located at different distances, a three-branch classifiercombining EWHOG features and supoort vector machine (SVM) is presented to recognizepedestrians. To reduce the computational and storage overhead, the resulting support vectorsare optimized using fast classification supoort vector machine (FCSVM). A further validationphase is then proposed to suppress some flase detection according to the intensity differencebetween FIR pedestrians’ heads and their adjacent regions. Experiments show that theproposed EWHOG is more approapriate to distinguish FIR pedestrians; the fast pedestrianrecognition framework guarantees higher implementation efficiency and the results in bothurban and suburban scenarios demonstrate its acceptable detection performance, at the cost ofonly slightly decrease of detection accuracy.
3) Considering the rare-event-detection inherent in the tasks of pedestrian detectionwhere rare pedestrians need to be located from enormous background regions in the imagesequences, this dissertation proposes a pre-segmentation method called pixel-gradientoriented vertical projection to efficiently locate the vertical image stripes that probablycontain FIR pedestrians, which avoids the dense search within the whole input images. It isbased on the feature that the ground and sky in FIR images usually represent as largehomogeneous regions, which makes it possible to perform pixel-gradient oriented verticalprojection using the gradient information. Experimental results indicate that thepre-segmentation method significantly improves the speed of ROIs extraction and helps tofilter out some negative ROIs. In order to capture both the local object shape described by theentropy weighted distribution of oriented gradient histograms and its pyramid spatial layout, anovel pyramid entropy weighted histograms of oriented gradients (PEWHOG) is proposed todescribe FIR pedestrians. Then PEWHOG is fed to a three-branch structured SVM classifierusing histogram intersection kernel (HIK). An off-line training procedure combining both thebootstrapping and early-stopping strategy is proposed to generate a more robust classifier byexploiting hard negative samples iteratively, which also deals with the issue thatgeneralization ability of the resulting classifier depends on the initial training data.
4) Under a traditional learning-based pedestrian detection framework, an FIR pedestrianclassifier trained by data extracted from one scenario may face difficulty in detectingpedestrians correctly in another distinct scenario due to the inevitable disparity in distributionsbetween the training data and test data. And it is expensive and sometimes difficult to label sufficient new training data from target domains to re-train a scenario-specific classifier. Tothis end, this dissertation proposes a novel Boosting-style algorithm for data-level transferlearning termed DTLBoost to detect FIR pedestrians towards distinct scenarios adaptationefficiently and effectively, which requires only a small amount of newly labeled training datafrom the target domains. To achieve better Boosting-style ensembles for inductive transferlearning, the degree of classification disagreement is formulated explicitly and incorporatedinto the weight updating rules of training samples. It helps to select the samples in auxiliarydata with positive transferability and encourage different base learners to learn different partsor aspects of target data. Extensive experiments including the performance evaluation of bothclassifier-level and system-level has been conducted to validate the effectiveness of theproposed method using our FIR pedestrian dataset and OSU thermal pedestrian dataset. Theresults demonstrate that the proposed method can impressively improve the detectionperformance across distinct scenarios, i.e. towards both new scenes and viewpointsadaptation.
引文
[1] Gerónimo D, López A M, Sappa A D, et al. Survey of pedestrian detection for advanceddriver assistance systems [J]. IEEE Transactions on Pattern Analysis and MachineIntelligence,2010,32(7):1239-1258
[2] Borhade S, Shah M, Jadhav P, et al. Advanced Driver Assistance System [A].Proceedings of International Conference on Sensing Technology (ICST)[C]. Kolkata,India: IEEE,2012:718-722
[3] Olmeda D, de la Escalera A, Armingol J M. Contrast invariant features for humandetection in far infrared images [A]. Proceedings of IEEE Intelligent VehiclesSymposium (IV)[C]. Alcala de Henares, Spain: IEEE,2012:117-122
[4] Sun X, Lu H P, Wu J. Bus detection based on sparse representation of transit signalpriority [J]. Neurocomputing,2013,118:1-9
[5] Huang S Q, Shalaby A, Bissessar R. Impacts of transit priority on signal coordinationcase study of Toronto, Ontario, Canada [J]. Transportation Research Record,2012,2311:29-43
[6] Wang J T, Chen D B, Chen H Y, et al. On pedestrian detection and tracking in infraredvideos [J], Pattern Recognition Letters,2012,33(6):775-785
[7] Zhang T Z, Liu S, Xu C S, et al. Mining semantic context information for intelligentvideo surveillance of traffic scenes [J]. IEEE Transactions on Industrial Informatics,2013,9(1):149-160
[8] Gavrila D M. Sensor-based pedestrian protection [J]. IEEE Intelligent Systems,2001,16(6):77-81
[9] Traffic Safety Facts: A compilation of motor vehicle crash data from the fatalityanalysis reporting system and the general estimates system [EB/OL].http://www-nrd.nhtsa.dot.gov/pdf/nrd-30/NCSA/TSFAnn/TSF2004.pdf,2004
[10] Fors C, Lundkvist S O. Night-time traffic in urban areas: a literature review on roaduser aspects [R]. Sweden: Swedish National Road and Transport Research Institute(VTI),2009
[11]公安部交通管理局,2009年全国道路交通事故情况[EB/OL].http://www.mps.gov.cn/n16/n1252/n1837/n2557/2276407.html,2010
[12]田广,戚飞虎,朱文佳等.单目移动拍摄下基于人体部位的行人检测[J].系统仿真学报.2006,18(10):2906-2910
[13] Dalal N, Triggs B. Histograms of oriented gradients for human detection [A].Proceeding of IEEE Conference on Computer Vision and Pattern Recognition (CVPR)[C]. San Diego, CA, USA: IEEE,2005:886-893
[14] Gerónimo D, Sappa A, López A M, et al. Adaptive image sampling and windowsclassification for onbord pedestrian detection [A]. Proceeding of InternationalConference on Computer Vision Systems (ICVS)[C]. Bielefeld University, Germany:Applied Computer Science Group,2007
[15] Mittal A, Paragios N. Motion-based background subtraction using adaptive kerneldensity estimation [A]. Proceedings of IEEE Conference on Computer Vision andPattern Recognition (CVPR)[C]. Washington DC, USA: IEEE,2004:302-309
[16] Leibe B, Cornelis N, Cornelis K, et al. Dynamic3D scene analysis from a movingvehicle [A]. Proceeding of IEEE Conference on Computer Vision and PatternRecognition (CVPR)[C]. Minneapolis, MN, USA: IEEE,2007:1423-1430
[17] Xu F L, Liu X, Fujimura K. Pedestrian detection and tracking with night vision [J].IEEE Transactions on Intelligent Transportation Systems,2005,6(1):63-71
[18] Nanda H, Davis L. Probabilistic template based pedestrian detection in infrared videos[A]. Proceedings of IEEE Intelligent Vehicle Symposium (IV)[C]. Versailles, France:IEEE,2002:15-20
[19]云廷进,郭永彩,高潮. K-均值聚类中心分析法实现红外人体目标分割[J].光电工程,2008,35(3):140-144
[20] Liu Q, Zhuang J J. A generalized thresholding algorithm of pedestrian segmentation forfar-infrared images [A]. Proceedings of IEEE International Conference on ImagingSystems and Techniques (IST)[C]. Manchester, UK: IEEE,2012:338-343
[21] Davis J, Vinary S. Robust background-subtraction for person detection in thermalimagery [A]. Proceeding of the Conference on Computer Vision and PatternRecognition Workshop (CVPRW)[C]. Washington DC, USA: IEEE,2004:713-716
[22]梁英宏.红外视频图像中的人体目标检测方法[J].红外与激光工程,2009,38(5):931-935
[23] Sezgin M, Sankuro B. Survey over image thresholding techniques and quantitativeperformance evaluation [J]. Journal of Electronic Imaging,2004,13(l):146-165
[24] Ge J F, Luo Y P, Gyomei T. Real-time pedestrian detection and tracking at nighttimefor driver-assistance systems [J]. IEEE Transactions on Intelligent TransportationSystems,2009,10(2):283-298
[25] Xia D, Sun H, Shen Z K. Real-time infrared pedestrian detection based on multi-blockLBP [A]. Proceeding of International Conference on Computer Application and SystemModeling (ICCASM)[C]. Taiyuan, China: IEEE,2010:139-142
[26] Sun H, Wang C, Wang B L, et al. Pyramid binary pattern features for real-timepedestrian detection from infrared videos [J]. Neurocomputing,2011,74(5):797-804
[27] O’Malley R, Jones E, Glavin M. Detection of pedestrians in far-infrared automotivenight vision using region-growing and clothing distortion compensation [J]. InfraredPhysics&Technology,2010,53(6):439-449
[28] Olmeda D, de la Escalera A, Armingol J M. Detection and tracking of pedestrians ininfrared images [A]. Proceeding of International Conference on Signals, Circuits andSystems (SCS)[C]. Medenine, Tunisia: IEEE,2009:1-6
[29] Bertozzi M, Broggi A, Fascioli A, et al. Pedestrian detection for driver assistance usingmultiresolution infrared vision [J]. IEEE Transactions on Vehicular Technology,2004,53(6):1666-1678
[30] Fang Y J, Yamada K, Ninomiya Y, et al. A shape-independent method for pedestriandetection with far-infrared images [J]. IEEE Transactions on Vehicular Technology,2004,53(6):1679-1697
[31] Li J F, Gong W G, Li W H, et al. Robust pedestrian detection in thermal infraredimagery using the wavelet transform [J]. Infrared Physics&Technology,2010,53(4):267-273
[32] Krotosky S, Trivedi M M. Multimodal stereo image registration for pedestrian detection[A]. Proceeding of IEEE Intelligent Transportation Systems Conference (ITSC)[C].Toronto, Ont., Canada: IEEE,2006:109-114
[33] Broggi A, Fascioli A, Fedriga I, et al. Stereo-based preprocessing for human shapelocalization in unstructured environments [A]. Proceeding of IEEE Intelligent VehiclesSymposium (IV)[C]. Columbus, OH: IEEE,2003:410-415
[34] Zhao L, Thorpe C E. Stereo-and neural network-based pedestrian detection [J]. IEEETransaction on Intelligent Transportation Systems,2000,1(3):148-154
[35] Fardi B, Schuenert U, Wanielik G. Shape and motion-based pedestrian detection ininfrared images: A multi sensor approach [A]. Proceedings of IEEE Intelligent VehiclesSymposium (IV)[C]. Las Vegas, NV, USA: IEEE,2005:18-23
[36] Gavrila D M, Giebel J, Munder S. Vision-based pedestrian detection: the PROTECTORsystem [A]. Proceeding of IEEE Intelligent Vehicles Symposium (IV)[C]. Parma, Italy:IEEE,2004:410-415
[37] Tons M, Doerfler R, Meinecke, M M. Radar sensors and sensor platform used forpedestrian protection in the EC-funded project SAVE-U [A]. Proceeding of IEEEIntelligent Vehicles Symposium (IV)[C]. Parma, Italy: IEEE,2004:813-818
[38] Curio C, Edelbrunner J, Kalineke T, et al. Walking pedestrian recognition [J]. IEEETransactions on Intelligent Transportation Systems,2000,1(3):155-163
[39] Cutler R, Davis L S. Robust real-time periodic motion detection, analysis andapplications [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2000,22(8):781-796
[40] Broggi A, Bertozzi M, Fascioli A, et al. Shape-based pedestrian detection [A].Proceeding of IEEE Intelligent Vehicles Symposium (IV)[C]. Dearborn, MI: IEEE,2000:215-220
[41] Gavrila D M. Pedestrian detection from a moving vehicle [J]. Lecture Notes inComputer Science,2000,1843:37-49
[42] Gavrila D M, Munder S. Multi-cue pedestrian detection and tracking from a movingvehicle [J]. International Journal of Computer Vision,2007,73(1):41-59
[43] Gavrila D M. A bayesian, exemplar-based approach to hierarchical shape matching [J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2007,29(8):1408-1421
[44] Mahlisch M, Oberlander M, Lohlein O, et al. A multiple detector approach tolow-resolution FIR pedestrian recognition [A]. Proceeding of IEEE Intelligent VehiclesSymposium (IV)[C]. Las Vegas, NV, USA: IEEE,2005:325-330
[45] Bertozzi M, Broggi A, Gomez C H, et al. Pedestrian detection in far infrared imagesbased on the use of probabilistic templates [A]. Proceeding of IEEE Intelligent VehiclesSymposium (IV)[C]. Istanbul, Turkey: IEEE,2007:327-332
[46] Munder S, Gavrila D M. An experimental study on pedestrian classification [J]. IEEETransactions on Pattern Analysis and Machine Intelligence,2006,28(11):1863-1868
[47] Papageorgiou C, Poggio T. A trainalbe system for object detection [J]. InternationalJournal of Computer Vision,2000,38,(1):15-33
[48] Viola P, Jones M. Rapid object detection using a boosted cascade of simple features [A].Proceeding of IEEE Conference on Computer Vision and Pattern Recognition (CVPR)[C]. Kauai, HI, USA: IEEE,2001:511-518
[49] Viola P, Jones M, Snow D. Detecting pedestrians using patterns of motion andappearance [A]. Proceeding of IEEE International Conference on Computer Vision(CVPR)[C]. Nice, France: IEEE,2003:734-741
[50] Rakate G R, Borhade S R, Jadhav P S, et al. Advanced pedestrian detection systemusing combination of Haar-like features, Adaboost algorithm and Edgelet-Shapelet [A].Proceedings of IEEE International Conference on Computational Intelligence andComputing Research (ICCIC)[C]. Coimbatore, India: IEEE,2012:1-5
[51] Zhang Y, Liu W M, Mo C, et al. Pedestrian detection in complex scene using full binarytree classifiers based on locally assembled Binary Haar-like features [A]. Proceedingsof World Congress on Intelligent Control and Automation (WCICA)[C]. Taipei,Taiwan, China: IEEE,2011:1180-1184
[52] Miron A, Besbes B, Rogozan A, et al. Intensity self similarity features for pedestriandetection in Far-Infrared images [A]. Proceedings of IEEE Intelligent VehiclesSymposium (IV)[C]. Alcala de Henares, Spain: IEEE,2012:1120–1125
[53] Kim D S, Kim M, Kim B S, et al. Histograms of local intensity differences forpedestrian classification in far-infrared images [J]. Electronics Letters,2013,49(4):258-260
[54] Ojala T, Pietikainen M, Maenpaa T. Multiresolution gray-scale and rotation invarianttexture classification with local binary patterns [J]. IEEE Transactions on PatternAnalysis and Machine Intelligence,2002,24(7):971-987
[55] Boudissa, A, Tan J K, Kim H, et al. A simple pedestrian detection using LBP-basedpatterns of oriented edges [A]. Proceedings of IEEE International Conference on ImageProcessing (ICIP)[C]. Lake Buena Vista, FL, USA: IEEE,2012:469-472
[56] Nguyen D T, Ogunbona, P, Li W Q. Human detection with contour-based local motionbinary patterns [A]. Proceedings of IEEE International Conference on Image Processing(ICIP)[C]. Brussels, Belgium: IEEE,2011:3609-3612
[57] Zheng Y B, Shen C H, Hartley R, et al. Pyramid center-symmetric local binary/trinarypatterns for effective pedestrian detection [J]. Lecture Notes in Computer Science,2011,6495:281-292
[58] Jiajia Luo, Hairong Qi. Motion Local Ternary Pattern for distributed multi-view humanaction recognition [A]. Proceedings of International Conference on Distributed SmartCameras (ICDSC)[C]. Hong Kong, China: IEEE,2012:1-6
[59] Nanni L, Brahnam S, Lumini A. Local Ternary Patterns from Three Orthogonal Planesfor human action classification [J]. Expert Systems with Applications,2011,38(5):5125-5128
[60] Sabzmeydani P, Mori G. Detecting pedestrians by learning shapelet features [A].Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR)[C]. Minneapolis, MN, USA: IEEE,2007:1–8
[61] Wang Y Z, Xing J P, Luo X L, et al. Pedestrian detection using coarse-to-fine methodwith Haar-like and Shapelet features [A]. Proceedings of International Conference onMultimedia Technology (ICMT)[C]. Ningbo, China: IEEE,2010:1-4
[62] Zhang L, Wu B, Nevatia R, Pedestrian detection in infrared images based on local shapefeatures [A]. Proceedings of IEEE Conference on Computer Vision and PatternRecognition (CVPR)[C]. Minneapolis, MN, USA: IEEE,2007:1–8
[63] Wu B, Nevatia R. Detection and tracking of multiple, partially occluded humans byBayesian combination of edgelet based part detectors [J]. International Journal ofComputer Vision,2007,75(2):247–266
[64] Hu B, Wang S J, Ding X Q. Multi features combination for pedestrian detection [J].Journal of Multimedia,2010,5(1):79-84
[65] Levi K, Weiss Y. Learning object detection from a small number of examples: theimportance of good features [A]. Proceeding of IEEE Conference on Computer Visionand Pattern Recognition (CVPR)[C]. Washington DC, USA: IEEE,2004:53-60
[66] Yun T J, Guo Y C, Gao C. Human detection in far-infrared iamges based on histogramsof maximal oriented energy map [A]. Proceedings of International Conference onWavelet Analysis adn Pattern Recognition (ICWAPR)[C]. Beijing, China: IEEE,2007:933-938
[67] Gerónimo D, López A, Ponsa D, et al. Haar wavelets and edge orientation histogramsfor on-board pedestrian detection [J]. Lecture Notes in Computer Science,2007,4477:418-425
[68] Lowe D G. Distinctive image features from scale-invariant keypoints [J]. InternationalJournal of Computer Vision,2004,60(2):91-110
[69] Du J H, Guan Y P, Shi Y J. Pedestrian detection based on rapid SIFT matching [J].Chinese Journal of Electron Devices,2012,35(5):601-606
[70] Leoputra W S., Venkatesh S, Tan T. Pedestrian Detection for Mobile Bus Surveillance[A]. Proceedings of International Conference on Control, Automation, Robotics andVision (ICASSP)[C]. Hanoi, Vietnam: IEEE,2008:726-732
[71] Mikolajczyk K, Leibe B, Schiele B. Local features for object class recognition [A].Proceedings of IEEE International Conference on Computer Vision (ICCV)[C]. Beijing,China: IEEE,2005:1792-1799
[72] Zhu Q, Yeh M C, Cheng K T, et al. Fast human detection using a cascade of histogramsof oriented gradients [A]. Proceeding of IEEE Conference on Computer Vision andPattern Recognition (CVPR)[C]. New York, NY, USA: IEEE,2006:1491-1498
[73] Shen J F, Sun C Y, Yang W K. Fast human detection based on enhanced variable sizeHOG features [J]. Lecture Notes in Computer Science,2011,6676:342-349
[74] Watanabe T, Ito S, Yokoi K. Co-occurrence histograms of oriented gradients for humandetection [J]. Information and Media Technologies,2010,5(2):659-667
[75] Ren H Y, Heng C K, Zheng W, et al. Fast object detection using boosted co-occurrencehistograms of oriented gradients [A]. Proceeding of IEEE International Conference onImage Processing (ICIP)[C]. Hong Kong, China: IEEE,2010:2705-2708
[76] Satpathy A, Jiang X D, Eng H L. Extended histogram of gradients with asymmetricprincipal component and discriminant analyses for human detection [A]. Proceeding ofCanadian Conference on Computer and Robot Vision (CRV)[C]. St. Johns, NL,Canada: IEEE,2011:64-71
[77] Pang Y W, Yuan Y, Li X L, et al. Efficient HOG human detection [J]. SignalProcessing,2011,91(4):773-781
[78] Hu J, Guan C, Zhou X. Improved HOG descriptor for pedestrian detection based onrough set [J]. Journal of Convergence Information Technology,2013,8(10):325-332
[79] Guo L, Ge P S, Zhao Y B, et al. Pedestrian Detection based on HOG features optimizedby gentle AdaBoost in ROI [J]. Journal of Convergence Information Technology,2013,8(2):554-562
[80] Pang Y W, Yan H, Yuan Y, et al. Robust CoHOG feature extraction in human-centeredimage/video management system [J]. IEEE Transactions on Systems, Man, andCybernetics, Part B: Cybernetics,2012,42(2):458-468
[81] Xu R, Jiao J B, Zhang B C, et al. Pedestrian detection in images via cascaded L1-normminimization learning method [J]. Pattern Recognition,2012,45(7):2573-2583
[82] Neagoe V E, Ciotec A, Barar A. A concurrent neural network approach to pedestriandetection in thermal imagery [A]. Proceedings of International Conference onCommunications (COMM)[C]. Bucharest, Romania: IEEE,2012:133-136
[83] Walk S, Majer N, Schindler K, et al. New features and insights for pedestrian detection[A]. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition(CVPR)[C]. San Francisco, CA, USA: IEEE,2010:1030–1037
[84] Dollár P, Wojek C, Schiele B, et al. Pedestrian detection: an evaluation of the state ofthe art [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2012,34(4):743-761
[85] Wojek C, Schiele B. A performance evaluation of single and multi-feature peopledetection [J]. Lecture Notes in Computer Science,2008,5096:82-91
[86] Geismann P, Knoll A. Speeding up HOG and LBP features for pedestrian detection bymultiresolution techniques [J]. Lecture Notes in Computer Science,2010,6453:243-252
[87] Ma Y D, Chen X K, Chen G. Pedestrian detection and tracking using HOG andoriented-LBP features [J]. Lecture Notes in Computer Science,2011,6985:176-184
[88] Enzweiler M, Gavrila D M. A Multilevel mixture-of-experts framework for pedestrianclassification [J]. IEEE Transactions on Image Processing,2011,20(10):2967-2979
[89]郭立君,刘曦,赵杰煜.结合运动信息与表观特征的行人检测方法[J].软件学报,2012,23(2):299-309
[90] Cheng W C, Jhan D M. A self-constructing cascade classifier with AdaBoost and SVMfor pedestrian detection [J]. Engineering Applications of Artificial Intelligence,2013,26(3):1016-1028
[91] Su S Z, Chen S Y. Analysis of feature fusion based on HIK SVM and its application forpedestrian detection [J]. Abstract and Applied Analysis,2013:436062
[92] Cortes C, Vapnik V. Support vector networks [J]. Machine Learning,1995,20(3):273-297
[93] Freund Y, Schapire R E. A decision-theoretic generalization of on-line learning and anapplication to boosting [J]. Journal of Computer and System Sciences,1997,55(1):119-139
[94] Maji S, Berg A C, Malik J. Classification using intersection kernel support vectormachines is efficient [A]. Proceeding of IEEE Conference on Computer Vision andPattern Recognition (CVPR)[C]. Anchorage, AK, USA: IEEE,2008:1-8
[95] Tuzel O, Porikli F, Meer P. Pedestrian detection via classification on riemannianmanifolds [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2008,30(10):1713-1727
[96] Shashua A, Gdalyahu Y, Hayun G. Pedestrian detection for driving assistance systems:single-frame classification and system level performance [A]. Proceeding of IEEEIntelligent Vehicles Symposium (IV)[C]. Parma, Italy: IEEE,2004:1-6
[97] Mohan A, Papageorgiou C, Poggio T. Example-based object detection in images bycomponents [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2001,23(4):349-361
[98] Bo W, Nevatia R. Detection of multiple, partially occluded humans in a single image byBayesian combination of edgelet part detectors [A]. Proceeding of IEEE InternationalConference on Computer Vision (ICCV)[C]. Beijing, China: IEEE,2005:90-97
[99] Xiong Y, Peng J X, Ding M Y, et al. An extended track-before-detect algorithm forinfrared target detection [J]. IEEE Transactions on Aerospace and Electronic Systems,1997,33(3):1087-1092
[100] Long Y L, Xu H, An W, et al. Track-before-detect for Infrared Maneuvering DimMulti-target via MM-PHD [J]. Chinese Journal of Aeronautics,2012,25(2):252-261
[101] Felzenszwalb P F, Girshick R B, McAllester D, et al. Object detection withdiscriminatively trained part-based models [J]. IEEE Transactions on Pattern Analysisand Machine Intelligence,2010,32(9):1627-1645
[102] Grauman K, Darrell T. The pyramid match kernel: discriminative classification withsets of image features [A]. Proceedings of IEEE International Conference on ComputerVision (ICCV)[C]. Beijing, China: IEEE,2005:1458-1465
[103] Wu J X. A fast dual method for HIK SVM learning [J]. Lecture Notes in ComputerScience,2010,6312:552-565
[104] Davis J W, Keck M A. A two-stage template approach to person detection in thermalimagery [A]. Proceedings of IEEE Workshops on Application of Computer Vision(WACV)[C]. Breckenridge, CO, USA: IEEE,2005:364-369
[105] Giebel J，Gavrila D M，Schn rr C. A Bayesian framework for multi-cue3D objecttracking [J]. Lecture Notes in Computer Science,2004,3024:241-252
[106]赵运基,裴海龙.基于增量学习的关节式目标跟踪算法[J].华南理工大学学报(自然科学版),2012,40(3):88-93
[107] Gao J. Designing real-time multimedia applications on mobile devices [A]. Proceedingsof ACM SOSP Workshop on Networking, Systems, and Applications on MobileHandhelds [C]. New York, NY, USA: ACM,2011:1-6
[108]刘向东,陈兆乾.一种快速支持向量机分类算法的研究[J].计算机研究与发展,2004,41(8):1327-1332
[109]许言午,曹先彬,乔红.行人检测系统研究新进展及关键技术展望[J].电子学报,2008,36(5):962-968
[110] Xu Y W, Cao X B, Qiao H. An efficient tree classifier ensemble-based approach forpedestrian detection [J]. IEEE Transactions on Systems, Man and Cybernetics, Part B:Cybernetics,2011,41(1):107–117
[111] Saini L M, Soni M K. Artificial neural network-based peak load forecasting usingconjugate gradient methods [J]. IEEE Transactions on Power Systems,2002,17(3):907-912
[112] Wang X Y, Hua G, Han T X. Detection by detections: non-parametric detectoradaptation for a video [A]. Proceedings of IEEE Conference on Computer Vision andPattern Recognition (CVPR)[C]. Providence, RI, USA: IEEE,2012:350-357
[113] Sharma P, Huang C, Nevatia R. Unsupervised incremental learning for improved objectdetection in a video [A]. Proceedings of IEEE Conference on Computer Vision andPattern Recognition (CVPR)[C]. Providence, RI, USA: IEEE,2012:3298-3305
[114] Roth P M, Sternig S, Grabner H, et al. Classifier grids for robust adaptive objectdetection [A]. Proceedings of IEEE Conference on Computer Vision and PatternRecognition (CVPR)[C]. Miami, FL, USA: IEEE,2009:2727-2734
[115] Stalder S, Grabner H, Gool L V. Exploring context to learn scene specific objectdetectors [A]. Proceedings of International Conference on Performance Evaluation ofTracking and Surveillance (PETS)[C]. Miami, USA: IEEE,2009:63-70
[116] Rosenberg C, Hebert M, Schneiderman H. Semi-supervised self-training of objectdetection models [A]. Proceedings of IEEE Workshops on Application of ComputerVision (WACV)[C]. Breckenridge, CO, USA: IEEE,2005:29-36
[117] Pan S J, Yang Q. A survey on transfer learning [J]. IEEE Transactions on Knowledgeand Data Engineering,2010,22(10):1345-1359
[118] Dai W Y, Yang Q, Xue G R, et al. Boosting for transfer learning [A]. Proceedings ofInternational Conference on Machine Learning (ICML)[C]. New York, NY, USA:ACM,2007:193-200
[119] Cao X B, Wang Z, Yan P K, et al. Transfer learning for pedestrian detection [J].Neurocomputing,2013,100:51-57
[120] Eaton E, desJardins M. Selective Transfer between learning tasks using task-basedboosting [A]. Proceedings of AAAI Conference on Artificial Intelligence [C]. SanFrancisco, California, USA: AAAI Press,2011,8:337-342
[121]于立萍,唐焕玲.基于分类一致性的迁移学习及其在行人检测中的应用[J].山东大学学报(工学版),2013,43(4):26-31,45
[122] Pang J B, Huang Q M, Yan S C, et al. Transferring boosted detectors towards viewpointand scene adaptiveness [J]. IEEE Transactions on Image Processing,2011,20(5):1388-1400
[123]谢尧芳,苏松志,李绍滋.基于稀疏编码的迁移学习及其在行人检测中的应用[J].厦门大学学报(自然科学版),2010,49(2):186-192
[124] Wang M, Li W, Wang X G. Transferring a generic pedestrian detector towards specificscenes [A]. Proceedings of IEEE Conference on Computer Vision and PatternRecognition (CVPR)[C]. Providence, RI, USA: IEEE,2012:3274-3281
[125] Yu G X, Zhang G J, Yu Z W, et al. Semi-supervised ensemble classification insubspaces [J]. Applied Soft Computing,2012,12(5):1511-1522
[126] Wang S, Chen H H, Yao X. Negative correlation learning for classification ensembles[A]. Proceedings of International Joint Conference on Neural Networks (IJCNN)[C].Barcelona, Spain: IEEE,2010:1-8
[127] Liu Y, Yao X. Simultaneous training of negatively correlated neural networks in anensemble [J]. IEEE Transactions on Systems, Man, and Cybernetics, Part B:Cybernetics,1999,29(6):716-725
[128] Everingham M, Gool L V, Williams C K I, et al. The Pascal visual object classes (VOC)challenge [J]. International Journal of Computer Vision,2010,88(2):303-338