Object tracking under low signal-to-noise-ratio with the instantaneous-possible-moving-position model

详细信息    查看全文

• 作者：Cheng-Liang Wang^a ; ^b ; ^{wangcl55@gmail.com} ; Xiaoming Huo^b
• 关键词：Object tracking ; Motion model ; EM algorithm ; Probabilistic data association ; Hidden Markov models
• 刊名：Signal Processing
• 出版年：2013
• 出版时间：May, 2013
• 年：2013
• 卷：93
• 期：5
• 页码：1044-1055
• 全文大小：1501 K

文摘

Combing image processing technique and the probabilistic data association (PDA) motion model, we develop a novel framework to solve the problem of object tracking for non-electromechanical system with overwhelming noise background. The new model has two advantages: (1) By integrating the statistical motion model, the movement of object in many non-electromechanical systems could be more precisely simulated than existing ones. (2) Because of the adoption of a global search for optimal model parameters, the proposed model is better to track objects in high noise environment, comparing with other methods that rely on consecutive frames differentiating. We derive the expectation-maximization (EM) algorithm within the proposed model. Its usefulness is demonstrated with both synthesized data and image data set. Model Stability is introduced to quantify the usefulness of the model.