基于特征点分类的实时多目标检测与跟踪-计算机应用专业论文.docx

浙江大学确士学位论文 摘要 摘要 多目标的检测与跟踪是计算机视觉里的一个很重要的问题，在安全监控等领 域有着广泛的应用，比如智能交通系统。目前流行的方法存在着各种局限性 z 基 于背景差分的方法会受遮挡、光照变化以及相机抖动的影响:基于特征点聚类的 方法因为目标物体的大小不间，很难达到稳定的跟踪:另外一些基于物体模型匹 配的方法，因为计算复杂而难以达到实时。本文提出了…种新颖的基于特征点分 类的多目标检测与跟踪方法，不依赖背景，也能达到实时稳定的检测跟踪效果。 算法主要包含两部分:预处理阶段，把目标物体分为若干部分，并提取每个部分 的特征来训练一个分类器;在线时，首先在整张阁片中提取特征点，通过训练好 的分类器来筛选属于目标物体的特征点，并确定特征点对应的物体周部，之后通 过一种离散化的投票方法快速检测出目标，并确定特征点与目标之间的对应关 系，最后在跟踪特征点的基础上进行目标的跟踪。由于采用了基于局部分块的策 略，本文的方法能够鲁棒地处理局部遮挡情况。文中的实验结果我明该方法用于 检测跟踪具有快速而且稳定的效果，在此基础上，本文还对目标进行了速度估计 及物体分割，都具有很好的效果。 关键词z 目标检测:跟踪:特征点:分类器;分割 浙江大学刷士学位论文 Abstracl Abstract Multi-Object detection and tracking is an important problem in Computer Vision，and has various applications in the security surveillance area，such 础 inte11igent transport system. The classíc algorithms have kinds of limitations. For instanω，background subtraction based methods，are very difficult to handle problems of 例clusions，illumination change，and camera shaking. Approaches based on feature trackíng and grouping is usually not stable enough due to di他rent sizes of objects. Other 叩proaches which based on appearanωmatching ， is usually time consuming and may not achieve real time performance. In this paper，we propose a fast object detection and tracking method based on feature point classification ，which is independent of background model. Our method ωnsists of two parts: 1) in offline module，we divide the object into several parts a∞ording to feature point Iωations，and extract each parts feature to train a classifier; 2) in the online module，we first get the feature points and use trained classifier to decide which part the feature points belong to. Then we employ a di阳ete voting scheme ωdetect target objects in a ve可如t speed. Final1y，after establishing the ωnespondence between feature points and objects，we track o时ects basing on feature point tracking. Our meth叫 is robust to partial occlusion since we use a p翩翩based strategy. Experiments show that our method is very fast and effective. We also estímate