多阶微分环形模板匹配跟踪方法

基于环形模板结构提出一种新的模板匹配跟踪方法,在匹配准则函数中增加了图像的多阶微分匹配信息,在确保算法具有旋转不变性的前提下,可有效提高算法识别的准确性.匹配模板中的像素按环形结构排列,环形结构的中心对称性确保了算法具有旋转不变性.根据图像的灰度值可计算图像的各阶微分值,匹配准则函数由灰度匹配值和各阶微分匹配值共同构成.微分匹配信息中包含了图像的细节信息的匹配结果,因此可提高算法识别的准确性.仿真结果表明:该方法可应用于序列图像的目标识别,在背景复杂或目标发生较大旋转变化的情况下,该方法的识别性能明显优于传统方法.

Circle template matching and tracking algorithm based on multi-order differential information

A new template matching algorithm based on circle template is proposed.Multi-order image differential information is added into matching criterion function,which makes the algorithm have not only rotational invariance but also higher recognition precision.The pixels in the template are arranged in a ring structure.The center symmetry of the circle structure can make the algorithm have the rotational invariance.Differential images can be obtained according to the gray values,and the matching criterion function is composed of gray image matching result and differential image matching results.Differential image matching results contain the more detail information of object,which can enhance the identification accuracy of the algorithm.Simulation results prove that the method can be applied to moving objects recognition in sequential images,and under the complex background or rotary motion circumstances,the method has better recognition performance than conventional methods.