| 摘要: |
| 为了解决复杂场景下，基于整体表观模型的目标跟踪算法容易丢失目标的问题，提出了一种多模型协作的分块目标跟踪算法.融合基于局部敏感直方图的产生式模型和基于超像素分割的判别式模型构建目标表观模型，提取局部敏感直方图的亮度不变特征来抵制光照变化的影响；引入目标模型的自适应分块划分策略以解决局部敏感直方图算法缺少有效遮挡处理机制的问题，提高目标的抗遮挡性；通过相对熵和均值聚类度量子块的局部差异置信度和目标背景置信度，建立双权值约束机制和子块异步更新策略，在粒子滤波框架下，选择置信度高的子块定位目标.实验结果表明，该方法在复杂场景下具有良好的跟踪精度和稳定性. |
| 关键词: 协作模型 局部敏感直方图 粒子滤波 分块目标跟踪 超像素 |
| DOI：10.13328/j.cnki.jos.005656 |
| 分类号:TP391 |
| 基金项目:国家自然科学基金（61472196，61672305）；山东省重点研发项目（2017GGX10133） |
|
| Part-based Object Tracking Based on Multi Collaborative Model |
|
LIU Ming-Hua1, WANG Chuan-Sheng2, HU Qiang1, WANG Chuan-Xu1, CUI Xue-Hong1
|
|
1.College of Information Science and Technology, Qingdao University of Science and Technology, Qingdao 266061, China;2.College of Mechanical and Electrical Engineering, Qingdao University of Science and Technology, Qingdao 266061, China
|
| Abstract: |
| A part-based tracking approach based on multi collaborative model is proposed that can address the problem of losing object based on the holistic appearance model in complex scenarios. Object appearance model is constructed by fusing the generative model based on local sensitive histogram (LSH) and discriminative model based on superpixel segmentation, by extracting the illumination invariant feature of the LSH resist the influence of the illumination changes on the object model effectively; for the lack of effective occlusion handling mechanism of the LSH algorithm, the part-based adaptive model segmentation method is introduced to improve the performance of resistance occlusion; by through the relative entropy and mean shift cluster method, measuring the differences confidence value and the foreground-background confidence value of the local part, establish the dual weights constraint mechanism and asynchronous update strategy for the part model, the partes with high confidence are selected to locate object in the particle filter framework. Experimental results on challenging sequences confirm that the proposed approach outperforms the related tracking algorithm in complex scenarios. |
| Key words: collaborative model local sensitive histogram particle filter part-based object tracking superpixel |
