一种改进的基于SIFT特征的快速匹配算法

对SIFT特征匹配算法进行改进,采用D2OG金字塔的过零点检测代替DOG金字塔的局部极值点检测,建立DOG金字塔后,利用DOG金字塔相邻层相减得到D2OG金字塔并在其上进行过零点检测;采用改进RANSAC算法二次消除错配,匹配点对经过RANSAC算法筛选后,再次利用RANSAC算法对匹配点对做进一步筛选.实验表明,改进的SIFT特征匹配算法在保证了较高精度的同时提高了算法的速度,能适应于实时性要求较高的领域.

An Improved Fast Maching Algorithm Based on SIFT Features

The SIFT feature matching algorithm was improved. The improved algorithm builds a D2OG pyramid and extreme detection in the DOG pyramid is replaced by zero crossing detection in the D2OG pyramid. After the establishment of DOG pyramid, the D2OG pyramid is got by the subtraction of the DOG pyramid adjacent layers and in which extremes are detected. The algorithm uses improved RANSAC twice to eliminate mismatch. The matching double points utilizes the RANSAC algorithm which had been used once for further screening. The experiment shows that the improved SIFT feature matching algorithm ensures high precision and improves the speed of the algorithm at the same time. It can adapt to the fields of higher real-time request.