针对复杂背景下视频目标跟踪的实时性和可靠性问题,提出了基于Rao-Blackwellized粒子滤波(RBPF)的颜色矩形特征和方向边缘信息融合的自适应跟踪算法。该算法采用Rao-Blackwellized粒子滤波提高滤波算法性能,采用积分图像快速计算颜色特征和方向边缘信息。根据跟踪实际情形,利用模糊逻辑自适应调节各特征权值,提高算法的跟踪速度和精度。视频跟踪仿真试验表明,该算法稳健、跟踪精度高,实时性好,能够在复杂背景下对可见光及红外等运动目标进行有效、可靠的跟踪。

Aiming at the problem of real time and reliability of visual object tracking in a clutter background,an adaptive tracking algorithm by fusion of the color rectangle feature with the Edge Orientation Histogram is presented based on the Rao-Blackwellized Particle Filter(RBPF). To improve the tracking speed and precision, the RBPF is adopted to improve the filter performance,and an integral image is used to quickly compute the color rectangle feature and Edge Orientation Histogram.Finally,a fuzzy logic is applied to adaptively adjust each feature weight and to improve tracking accuracy in real time. The simulation result from a video tracking illustrates that the algorithm is robust and effective for visible and infrared object tracking under complex backgrounds.