【摘要】 针对实时跟踪过程中提高特征点匹配的准确性和稳定性问题,提出一种将SURF算法与RANSAC算法相结合实现图像的实时匹配与跟踪的方法。基于SURF算法进行特征点检测;利用欧氏距离进行特征点粗匹配;基于RANSAC算法剔除误匹配点,并进一步根据匹配特征点分布的几何特性进行优化筛选;利用获取的正确的匹配点对求解单应性矩阵,得到两幅待配准图像之间的映射关系,从而实现图像的实时匹配与跟踪。实验结果表明,在不同实验条件下,包括旋转、缩放、局部遮挡和局部光照增强或减弱等,改进算法均能够在目标区域内寻找到基准图像,并进行实时跟踪,且具有很好的准确性和稳定性。更多还原

【Abstract】 Aiming at obtaining high accuracy and stability of image matching,a method of real-time tracking and mapping combined with the SURF algorithms and the RANSAC algorithms was presented. SURF algorithm was utilized to extract image features. Euclidean Distance was used to find the matching points of two images. Then,RANSAC algorithm was utilized to get rid of error matching points,and the results were optimized according to geometric characteristic of feature points distribution. Finally,the homography matrix and the mapping relation between the two images can be obtained according to correct matching points. Experiments were carried out under different conditions,such as partial occlusion,different local illumination,and so on. Experimental results show that the foreground objects in the target area can be tracked and mapped accurately and reliably in real time,under these different conditions.更多还原