基于改进分块匹配和AKF的抗遮挡目标跟踪

Anti-occlusion target tracking based on improved block matching and AKF

在基于图像处理技术的机动目标跟踪研究中,为更好实现目标处于持续遮挡状态下的跟踪,提出了一种改进的分块匹配和抗野值卡尔曼滤波(AKF)相结合的新方法。在每一帧图像序列里,依据目标与模板的匹配置信度动态调节子块的大小和位置,提高目标运动矢量的测量精度;建立AKF模型对目标的位置参数滤波;在目标被严重遮挡时,依据莱特准则在线判定和修正卡尔曼(Kalman)新息野值序列,得到被跟踪目标位置的最优估计。多组实验结果表明:新方法在目标处于遮挡状态下跟踪结果较准确,具有更好的抗遮挡跟踪能力,且算法的速度达110 FPS,满足实时性要求。

In the research of maneuvering target tracking based on image processing technology,propose a new method combining a modified block matching algorithm and anti-outlier Kalman filtering（AKF） to achieve better tracking result of the target under continued occlusion. Dynamically adjust the size and position of sub-blocks according to the matching reliability of the target and template in each image sequence to achieve measurement precision of motion vectors of target. AKF filtering model is established for filtering of target position parameters.Judge and modify the Kalman innovation value online according to Wright criterion when target is occluded,so as to achieve optimal estimation of the target position. The experimental result shows that the tracking result of new method is more accurate and it has better anti occlusion tracking ability when the target is occluded,and the speed of the algorithm reaches 110 FPS,which meets the real-time requirements.