基于多特征混合模型的视觉目标跟踪

Visual object tracking based on collaborative model using multiple features

用单一特征训练跟踪模型进行跟踪鲁棒性较差,为解决这一问题,提出一种多特征表示的混合模型跟踪方法,将生成跟踪模型与判别跟踪模型结合.在生成模型中,利用金字塔结构计算基于颜色的直方图特征表示并以此来计算目标和候选之间的匹配度;判别模型则采用由灰度特征,HOG特征和LBP特征融合训练得到的SVM分类器来判别候选是否为跟踪目标,接着将匹配度和分类结果结合产生对候选的评估,最终评估最高的候选作为跟踪结果同时也用来更新判别模型的训练集.在CVPR2013跟踪数据集上的实验结果表明,该方法能有效克服局部遮挡和背景干扰等问题,实现在复杂背景下的目标跟踪.

In this work,we focus on short-term single object tracking,which is the most general type of tracking problems.Numerous significant trackers have been proposed over the past few decades.As we can see from these trackers,using a single type of feature to train the tracking model is a drawback,which usually leads to poor robustness in object tracking.On the other hand,the tracking result of a single tracker can sometimes be very unstable due to the complication of video sequences.Aiming at addressing these problems,we propose a collaborative tracking model fusing hybrid features in this paper.A generative model,which uses color-based features to compute the matching scores between target object and candidates,is combined with a discriminative model,which trains SVM with gray-scale,HOG and LBP features to classify the candidates as target or non-target objects.A spatial pyramid model is adopted in the feature calculation process of the generative model for a better description of the target and candidates.Then,the matching score and classification score are fused to rank the candidates.All the candidates are produced under the particle filter framework.The top-ranked candidate is regarded as the target object and used to update the training set for discriminative model as well.Besides,different update scheme and frequen-cy are used for the two parts of the collaborative model.The ensemble does not impair the performance of the tracking precision but slightly outperform the single model.Finally,we compare our results with other trackers along with some analysis.The experimental results on visual tracking benchmark CVPR2013 demonstrate that our approach is effective for partial occlusion and background clutter in tracking,thus realizes desirable tracking performance under complex conditions.