复杂情形下目标跟踪的自适应粒子滤波算法
doi: 10.3724/SP.J.1146.2007.01272
Visual Target Tracking Based on the Adaptive Particle Filter in the Complex Situation
-
摘要: 该文提出一种自适应粒子滤波算法。首先建立目标的颜色模型，提出基于加权颜色分布图的目标颜色模型。采用该模型计算目标模板与粒子区域的相似程度，以此作为对目标物体定位的依据，使目标定位更加合理有效；进而在滤波过程中，针对粒子退化问题，提出基于mean-shift迭代的粒子重抽样方法，形成对抽样粒子集的自适应调节，提高了粒子质量，有效降低了粒子数量。最后，进行了对大机动快速运动的小目标和被严重遮挡目标的跟踪实验，结果表明该算法具有较强的鲁棒性。Abstract: This paper presents an adaptive particle filtering algorithm for image tracking based on weighted color probability contribution. First, a weighted color contribution graph is proposed. Taking use of the graph, similarity between target template and particles area is calculated, that makes the target located more reasonable and efficient. During the filtering, aiming at particles degeneration, a resampling method is proposed, forming adaptive adjustment to sampled particle-set. This promotes quality of particle and reduces its quantity while cost of calculation is reduced greatly. Experiment result shows, for some complex tracking conditions such as overall occultation and object tracking with fast moving and great maneuverability, the proposed algorithm has a better robust.
-
Kwok C, Fox D, and Meila M. Real-time particle filters [J].Proce. IEEE.2004, 92(3):469-484[2]Arulampalam S, Maskell S, Gordon N J, and Clapp T. Atutorial on particle filters for on-line non-linear/non-gaussianbayesian tracking[J].IEEE Trans. on Signal Processing.2002,50(2):174-188[3]Yoshinori Satoh, Takayuki Okatani, and Koichiro Deguchi. Acolor-based tracking by kalman particle filter [C]. IEEEProceedings of the 17th International Conference on PatternRecognition. Cambridge, United Kingdom. 2004: 502-505.[4]Zhou S K, Chellappa R, and Moghaddam B. Visual trackingand Recognition using appearance-adaptive models inparticle filters [J].IEEE Trans. on Image Processing.2004, 13(11):1491-1506[5]Koichiro Deguchi, Oki Kawanaka, and Takayuki Okatani.Object tracking by the mean-shift of regional colordistribution combined with the particle-filter algorithm [C].Proceedings of the 17th International Conference on PatternRecognition. Cambridge, United Kingdom, 2004: 506-509.[6]Jia J P, Wang Q, and Chai Y M. Object tracking by multidegreesof freedom mean shift procedure combined with theKalman particle filter algorithm [C]. Proceedings of the 2006International Conference on Machine Learning andCybernetics. Dalian, China, 2006: 3793-3797.[7]Kailath T. The Divergence, Bhattacharyya distance measuresin signal selection [J]. IEEE Trans. on Comm. Technology,1999, 15(2): 253-259.[8]Comaniciu D and Meer P. Mean shift: A robust applicationtoward feature space analysis [J].IEEE Trans. on PatternAnalysis and Machine Intelligence.2002, 24(5):603-619[9]Comaniciu D and Meer P. Kernel-based object tracking [J].IEEE Trans. on Pattern Analysis and Machine Intelligence.2003, 25(5):564-577 -
计量
- 文章访问数: 3563
- HTML全文浏览量: 92
- PDF下载量: 1588
- 被引次数: 0
下载:
