CAJ | 학술논문

针对复杂环境下遮挡和视角变化导致目标跟踪漂移问题，提出一种新的目标代谢模型，并结合 LK（Lucas-Kanade）图像配准方法，通过二阶段递进的跟踪方式，实现目标复杂运动状态下的有效跟踪。二阶段分别使用代谢模型中的动态模板和准静态模板进行跟踪，实时更新动态模板的同时，周期性更新准静态模板，利用目标运动信息与模板变化信息之间的关系，实现对目标新旧特征的代谢更新。实验结果表明，该算法有效解决了模型更新带来的跟踪漂移问题，具有较好的鲁棒性和实时性。
침대복잡배경하차당화시각변화도치목표근종표이문제，제출일충신적목표대사모형，병결합 LK（Lucas-Kanade）도상배준방법，통과이계단체진적근종방식，실현목표복잡운동상태하적유효근종。이계단분별사용대사모형중적동태모판화준정태모판진행근종，실시경신동태모판적동시，주기성경신준정태모판，이용목표운동신식여모판변화신식지간적관계，실현대목표신구특정적대사경신。실험결과표명，해산법유효해결료모형경신대래적근종표이문제，구유교호적로봉성화실시성。
Aiming at the problem of target tracking drift caused by occlusion and visual angle change in complex environment,this paper proposes a new target metabolic model,and achieves effective target tracking in the state of complex target motion based on the combination of Lucas-Kanade (LK)image registration algorithm and through a two-stage progressive tracking means.The two-stage progressive tracking mode utilises dynamic template and quasi-static template in metabolic model for target tracking respectively.While the dynamic template is timely updated,the quasi-static template is updated periodically as well.The metabolism update of target from old characteristics to new char-acteristics is realised by the utilisation of the relationship between the target motion information and the template change information.Experi-mental results show that the proposed method effectively solves the tracking sift problem brought about the model update,and has good quali-ties in terms of robustness and real-time property.