一种基于SMOG模型的红外目标跟踪新算法

NEW INFRARED OBJECT TRACKING ALGORITHM BASED ON SMOG MODEL

空间颜色混合高斯模型(SMOG)在基于颜色特征的目标跟踪中显示出比颜色直方图更强的目标鉴别能力,因为它不仅考虑了区域的颜色信息,而且也考虑了颜色的空间分布信息.本文将SMOG模型合理地引入到了红外目标的建模中,改进了原SMOG模型中的相似性度量函数,进一步提高了其对目标的鉴别能力.在粒子滤波框架内,使用简单的二阶自回归模型作为系统状态转换方程,将改进后的相似度函数作为各粒子的状态观测,给出了有效的模式更新方法以适应目标外观的变化,并设计出了一种有效的红外目标跟踪算法.实验证明,SMOG模型能有效刻画红外目标,本文提出的算法对红外目标的跟踪是稳健的.

In color based object tracking, spatial-color mixture of Gaussians （ SMOG model） outperforms the popular color histogram in object discriminative power since it considers not only the colors in a region but also the spatial layout of these colors. In this study, SMOG model was applied to represent infrared objects, and the original similarity function in SMOG was revised in order to further improve its discriminative power. In particle filter framework, the two order AR model was utilized to describe the state transition equation, the revised similarity function was used as measuring to adapt the change of object appearance, and a valid model update method under this model was designed, then based on these, an efficient infrared object tracking algorithm was proposed. Experimental results show that the SMOG model can well represent the infrared object, and our proposed algorithm is effective and steady for tracking infrared objects.