CAJ | 학술논문

基于 Dezert‐Smarandache theory（DSmT ）提出一种自然场景下多运动目标遮挡的跟踪方法。在 DSmT 和粒子滤波框架下，通过设计新的合并策略来融合运动目标的颜色和位置信息，建立起融合位置和颜色信息的多目标跟踪仿真实验平台，最终实现自然场景下基于 DSmT 多运动目标遮挡的跟踪算法。通过两组不同场景下的跟踪实验测试，结果表明论文给出的证据合并策略和目标跟踪模型快速、高效。特别是对于完全交叉、遮挡目标的跟踪，最大粒子数达到40就能有效处理高冲突信息。进一步分析每个目标估计位置的 RMS 误差，对于自然场景下多目标跟踪过程中所涉及的证据间高冲突问题，采用文中方法显示出良好的跟踪能力和精度，其研究成果可为视觉跟踪技术的应用提供新的研究方法和思路。
기우 Dezert‐Smarandache theory（DSmT ）제출일충자연장경하다운동목표차당적근종방법。재 DSmT 화입자려파광가하，통과설계신적합병책략래융합운동목표적안색화위치신식，건립기융합위치화안색신식적다목표근종방진실험평태，최종실현자연장경하기우 DSmT 다운동목표차당적근종산법。통과량조불동장경하적근종실험측시，결과표명논문급출적증거합병책략화목표근종모형쾌속、고효。특별시대우완전교차、차당목표적근종，최대입자수체도40취능유효처리고충돌신식。진일보분석매개목표고계위치적 RMS 오차，대우자연장경하다목표근종과정중소섭급적증거간고충돌문제，채용문중방법현시출량호적근종능력화정도，기연구성과가위시각근종기술적응용제공신적연구방법화사로。
The aim of this article is to present a tracking approach for occluded objects in natural environment based on Dezert‐Smarandache theory(DSmT) .In the framework of DSmT and particle filters(PF) ,location and color cues of moving objects are combined by using new combination rules .As a result ,targets tracking platform which embeds location and color cues into the PF and DSmT is developed ,and the corresponding tracking algorithm in natural environment is realized based on DSmT .Two sets of experiments including many difficult tracking scenes with comparisons were carried out to validate the approach ,and results shown that the combination evidence strategy and objects tracking model based on DSmT are fast and efficient .Especially ,40 particles are used to handle the high conflict information between evidences for the tracking targets of crosses and occlusions .Further ,by analyzing the RMS errors of estimated position for every target ,the introduced ap‐proach exhibits an excellent tracking ability and accuracy for dealing with high conflict information between evidences in natu‐ral environment .The achievement of this work will provide new methods and new ideas for the application of vision tracking technique .