基于DST-PCR5多目标自适应视觉跟踪方法

复杂场景下实现快速稳定地自适应跟踪是视觉领域亟需解决的课题之一, 利用目标的多特征信息进行高效融合是提升跟踪算法鲁棒性能的重要途径。本文首先基于DST(Dempster-Shafer Theory)和PCR5(Proportional Conflict Redistribution No.5)设计一种新的合并策略融合运动目标的颜色和纹理特征，其次在粒子滤波框架下建立复杂场景下的多目标自适应跟踪模型，最终实现了复杂场景下多特征信息融合的自适应视觉跟踪。实验结果及性能分析表明，该方法在不良的跟踪条件下，高冲突证据的自适应处理能力得到明显改善，有效提高了粒子的使用效率和跟踪的鲁棒性，可以较好实现复杂场景下准确、稳定地多目标跟踪。

Adaptive visual tracking method of multi-targets based on DST-PCR5

In the field of computer vision, one of the most urgent issues is the realization of adaptive tracking fast and stably in complex scene, and the efficient combination by using multiple cues of moving targets is an important approach to improve the robust performance of tracking algorithms. Firstly, combined the color cues and texture cues of targets by designing a new combination strategy based on DST(Dempster-Shafer Theory) and PCR5(Proportional Conflict Redistribution No.5). Secondly, established the adaptive tracking model of multi-targets in the framework of particle filters. As a result, achieved the adaptive visual tracking of fusion multiple cues in complex scene. Two sets of experiments including difficult tracking scenes validated the proposed approach, and the results show that the adaptive ability of fusion high conflict evidences has obviously improvement based on the condition of bad tracking surroundings, and the efficiency of particles and the robustness of tracking targets perform well. The introduced approach exhibits an excellent ability for tracking multi-targets accurately and stably in complex scene.