基于圆投影向量的景象匹配方法研究

景象匹配中经常会出现基准图和模板图之间存在任意角度旋转的情况,传统的结构化模板匹配算法通过逐个比较基准图和模板图对应像素灰度值之间的相关性进行匹配,所以只能工作在小角度旋转的情况下,对任意角度旋转的景象匹配是无能为力的。利用圆的各向同性和投影特征提出了圆投影匹配算法,并对其进行了改进。实验结果表明,两种算法都具有旋转不变性,但是改进后的圆投影匹配算法对灰度变化、噪声、光照以及对比度变化等具有更好的鲁棒性。

Research on scene matching method using circular projection

Scene matching of arbitrary angle rotation is a difficult problem,but important for missile guidance.The conventional structural pixel-to-pixel template matching algorithms are the process of determining the presence and the location of a real time image or an object inside a scene image under analysis by a spatial cross-correlation process.So the conventional algorithms are helpless to the problem,and cannot be used for practical purposes.An algorithm for a rotation invariant scene matching method based on isotropy of circular and projection feature is proposed.Experimental results show that the two algorithms are rotation-invariance,and the proposed algorithm is more robust to gray change,noises,illumination,and contrast change.Furthermore,the computational cost of the proposed algorithm is cheaper than conventional algorithms.