基于SNN核的景象匹配算法

An Image Matching Algorithm Based on SNN Kernel Method

提出一种基于核方法的下视等分辨率景象匹配算法.通过模拟电荷吸引模型,提出了计算不等维高维数据相似度的SNN核函数.将图像中的特征点映射到径向基向量(Radial basis vector,RBV)空间,利用SNN核函数计算两个特征点集的相似度及过渡矩阵.利用置换测试模块来增强SNN核的稳定性,以确保输出解的可靠性.实验证明,基于SNN核的景象匹配算法对图象畸变、噪声干扰与信号缺失具有很强的鲁棒性,并可保证高精度与高实时性.

In this paper, an isotropic resolution down-looking scene matching algorithm based on SNN kernel method has been proposed. By simulating the charge attract model, the SNN kernel function is presented to compute similarity of high dimensional data with unequal dimensions. By mapping image character points to radial basis vector （RBV） space, the transition matrix and similarity between two character point sets are constructed by SNN kernel function. Finally, we use a permutation testing module to enforce the stability of SNN kernel to ensure the output reliability. The test results show that the proposed algorithm has high accuracy, high real-time performance, and strong robustness against the image aberrance, noise, and signal concealment.