一种基于曲线拟合预测的红外目标的跟踪算法

A Tracking Method Based on Curve Fitting Prediction of IR Object

给出了一种将曲线拟合和相关跟踪相结合的红外目标跟踪算法。该算法以目标质心作为匹配模板的中心 ,采用SSDA进行相关运算 ,以图像相关匹配的输出值更新所拟合的曲线 ,并以曲线的预测值作为下帧图像匹配搜索的初始值 ,使相关运算仅在很小的窗口内进行。该算法还定义了跟踪的帧内置信度和帧间置信度信号 ,实时智能的更新模板。实验表明 ,该算法能有效减少相关匹配的计算量 ,保证跟踪精度。

A novel IR object tracking method which combines curve fitting and correlation matching is presented. The method chooses the centroid of the object as the center of the matching template. Using SSDA as tracking algorithm, the method updates the curve fitting according to the tracking result, and the curve fitting prediction serves as the beginning point of the matching search, which makes the correlation operate only in a small window. The method also defines the new confidence measure, which can updates the template automatically in real-time. The experimental result shows that the method can effectively reduce the calculating quantum and improve tracking precision.