目标跟踪是计算机视觉研究中的热点与难点。本项目面向目标表观建模核心问题，针对现有方法不能有效克服目标光照变化、姿势变化、遮挡和前背景区分能力弱等挑战，以联合稀疏表示为理论基础，研究多线索特征联合的表观建模理论与方法，完成以下目标：建立并完善联合稀疏表示框架下多线索特征鲁棒学习和融合策略；兼顾多线索特征的表示能力与判别能力，有效利用目标的邻近约束和背景上下文信息，改善多线索目标表观模型的判别能力；综合以上，建立统一的多线索特征联合表示和判别学习框架，扩展现有表观建模理论与方法，实现复杂场景下鲁棒高效的目标跟踪。本项目预期改变目前表观模型单线索依赖、表示能力与判别能力彼此孤立的现状，对目标跟踪理论与方法实用化意义重大。

Object tracking is one of the most active and difficult problems in computer vision. This project faces the core problems of target appearance modeling in object tracking. Aiming at solving the tracking failures of existing methods when handling illumination changes, pose changes, occlusion and low foreground/background contrast, this project studies the theories and methods of appearance modeling based on the fusion of multi-cue features. Based on these studies, this project achieves the following goals: building and improving robust learning and fusion straggles of multi-cue features under the joint sparse representation framework; Taking into account of both representation abilities and discriminative abilities, effectively exploiting the proximity constraint and background context information, improving the discriminative abilities of multi-cue features. Taking these together, building an unified joint representation and discriminative learning framework of multi-cue features, expanding the exiting theories and methods of appearance modeling, achieving robust and efficient object tracking in complex scenes. This project is expected to change the status of the existing appearance modeling methods that depend single cue feature and independently consider the representation abilities and discriminative abilities. This project is of great importance to theories and methods of object tracking as well as their practical applications.