基于曲率特征与改进的RANSAC策略的图像匹配算法

Image matching algorithm based on curvature feature coupling improved RANSAC

为解决当前图像匹配算法主要利用图像梯度特征完成匹配,导致其存在较多误配率的不足,提出基于曲率特征与改进RANSAC的图像匹配算法。引入Harris算子,对图像特征点进行检测,利用这些特征点对应的尺度特征定义特征点优化模型;基于Haar小波,获取特征点对应的特征方向,构造同心圆,求取圆内指定方向的曲率特征,获取特征描述符;计算特征点对应的欧氏距离,完成图像特征点匹配;设计几何距离度量模型,改进RANSAC(random sample consensus)策略,剔除错误匹配特征点。实验结果表明,与当前图像匹配算法相比,所提算法具有更高的匹配正确度与鲁棒性。

To solve the defects of higher mismatch rate of image feature points induced by using the image gradient feature to complete the matching in current image matching algorithm,an image matching algorithm based on curvature feature coupling and improved RANSAC was proposed.The Harris operator was introduced to detect the feature points of the image.The pseudo feature points and redundant feature points were eliminated using the scale characteristics of these feature points to define the feature points optimization model.The feature direction corresponding to feature points was obtained based on Harris wavelet,and the feature descriptors were got by constructing the concentric circle to obtain the curvature characteristics of the specified direction in the circle.The Euclidean distance of feature points was calculated to match the feature points of the image.The RANSAC strategy was improved by designing the geometric distance measurement model to eliminate the error matching feature points.Experimental results show that the proposed algorithm has higher matching accuracy and better robustness compared with the current image matching algorithm.