模板图像尺寸可变的相关跟踪方法

相关跟踪过程中，序列图像中的实时图像与模板图像间存在灰度差异和几何形变，当前工程中采用相关匹配算法中的模板图像为矩形结构，为了在模板中尽可能的包含目标，模板中就不可避免的含有背景数据，但传统相关匹配算法中模板图像中的每一个像素都要参与计算，而且模板尺寸是固定的，这样就使得算法跟踪不稳定、容易丢失目标。为此提出一种变模板图像尺寸的相关跟踪方法，该方法实现了对模板图像尺寸的实时修正和非矩形模板相关匹配，试验证明此方法具有较强的抗几何变形的能力以及对复杂背景环境的适应能力。

A New Pattern Matching Algorithm Based on Adjusting the Template Image Size

In the process of tracing, the presence of differences in gray values and the distortion between the object image and the template image is inevitable. As the template image in traditional pattern matching algorithm is of the rectangle structure, so it is inevitable that the background data exist in the template image when the object is made to contain in the template image. The size of the template image in traditional pattern matching algorithm is fixed. Based on these factors, the process of tracing is instable and easily loses the object. A new pattern matching algorithm based on variable template image sizes was proposed in this paper. By applying the algorithm, the size of template image can be modified real time and the shape of template image is non-rectangular in structure. The tests show that this algorithm possesses the ability to resist distortion and adapts to complex background improved greatly.