一种新的任意角度旋转的景象匹配方法

为了解决任意角度旋转的景象匹配问题，在分析了传统的模板匹配算法对旋转敏感的本质原因后。提出了圆投影匹配算法，并对其进行了改进。实验结果表明，文中提出的算法不仅具有旋转不变性。而且对灰度变化、噪声、光照以及对比度变化等也具有很好的鲁棒性，同时匹配速度比归一化积相关匹配算法(NProd)提高了近一倍。

New Scene Matching Method for Arbitrary Rotation

Scene matching is used for image matching in military fi elds, right matching is carried out when there is an arbitrary angle between ref erence and template images. But traditional template matching algorithms can onl y work under the condition of small angle rotation, and cannot use for practical purposes. By analyzing the intrinsical matter that why the traditional pixel-t o-pixel template matching algorithms are sensitive to the rotation, an algorith m for the rotation invariance scene matching method based on circular projection is developed. Experimental results show that the developed algorithm is rotatio n invariance and robust to gray change, noises, illumination, and contrast chang e. Computational cost of the developed algorithm is only a half of the normalize d product correlation (NProd) method.