基于方向约束的改进SIFT匹配算法

Improved SIFT Matching Algorithm Based on Orientation Constraint

SIFT算法对图像的旋转、尺度变换、亮度变化等情况具有较好的不变性,广泛应用于图像匹配中,但SIFT特征向量生成过程复杂,导致算法实时性不理想,同时匹配结果还存在一定的误匹配点,影响了算法的精确性。为此,对SIFT算法进行改进,提出采用栅格形式选取种子点简化特征向量的生成过程,并利用关键点的方向约束性进一步剔除具有方向差异的误匹配点,从而简化计算量,提高匹配率。实验结果表明,改进后的算法能在保持原有SIFT算法稳定性的基础上提高近一倍的特征向量描述速度,初匹配结果经方向约束后能够有效地剔除具有方向差异的误匹配点,提高匹配率,大大增强了算法的精确性。

SIFT algorithm has good invariance in rotation,scaling the image,brightness change,etc.,and is widely used in image matching.But the generation process of SIFT feature vector is complicated,and results in that the real-time of the algorithms is not ideal.In the same time,there are still some matching mistakes in the matching result,affecting the accuracy of the algorithm.Therefore,we improved the SIFT algorithm,and proposed to use the grid in the selection of seed points to simplify the feature vector generation process,and used the orientation constraint of the key points to reject matching points with directional differences,thus simplifying computation and improving the matching rate.Experimental results show that the improved algorithm can maintain the basic stability of the original SIFT algorithm and improves the speed of feature vectors described nearly doubled.Through orientation constraint of the key points,the false match point with direction differences is rejected,and the matching rate is improved,greatly enhancing the accuracy of the algorithm.