联合生成与判别模型的目标检测与跟踪  

Object detection and tracking combining generative and discriminative model

作　　者：刘倩[1,2] 侯建华[1,2] 牟海军[1,2] 赵巍[3] 笪邦友[1,2] 

机构地区：[1]中南民族大学电子信息工程学院,武汉430074 [2]中南民族大学智能无线通信湖北省重点实验室,武汉430074 [3]华中农业大学食品科技学院,武汉430070

出　　处：《中国图象图形学报》2013年第10期1293-1301,共9页Journal of Image and Graphics

基　　金：国家自然科学基金项目(61141010);武汉市科技供需对接计划项目(201051824575);湖北省自然科学基金项目(2010CDB02001;2012FFA113);中南民族大学智能无线通信湖北省重点实验室开放课题(IWC2012005)

摘　　要：提出一种新的基于生成-判别模型的目标检测与跟踪方法。利用DAISY特征描述子所具有的对光照、形变、视角、尺度的不变性以及计算高效的特点,提取目标稳定的特征点并加以表达,形成生成模型;采用霍夫森林分类器作为判别模型,用以训练目标图像块。在后续视频序列中利用目标的检测结果和判别码本的相似性测量对模型进行更新,构建一个动态自适应的判别码本。实验结果表明这种将快速有效的DAISY描述子和识别率高、鲁棒性强的霍夫森林分类器相结合的算法,跟踪精度高、实时性较好,具有目标局部防遮挡能力和不同分辨率下的识别能力。A new object detection and tracking method is proposed based on generative and discriminative models. Firstl, stable features of object are extracted and represented by the DAISY feature descriptor, which has computational efficiency and is invariant to illumination, deformation, viewpoint, and scale. In this way, the object generative model is construc- ted. Second, the Hough forest classifier is adopted as discriminative model and the input patches of object are trained. Mo- reover, the discriminative codebook is updated by computing the similarity measurement between the detection results of the following video sequence and the codebook, and makes the codebook dynamic and adaptive. Experiment results show that the proposed algorithm, combining DAISY feature descriptor and Hough forest, has satisfactory tracking precision and good real-time performance, Additionally, it works well under the condition of partial occlusions and different image resolutions.

关 键 词：目标检测与跟踪 生成模型 判别模型 DAISY特征描述子 霍夫森林 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]