室外场景CSIFT的关键点检测和特征匹配

Keypoint Detection and Feature Matching of CSIFT in Outdoor Scenes

在目标的识别和匹配中,SIFT已经被证明是鲁棒性最好的局部不变特征描述符。虽然它能够对各种几何变化保持不变性,但却忽略了目标的颜色信息,限制了它的性能。提出一种改进的彩色的尺度不变特征变换(CSIFT)方法,结合目标的颜色特征与几何特征,针对室外场景的特点,简化了颜色不变量形式,实现了室外场景中目标的关键点检测和特征匹配。实验结果表明,CSIFT方法比经典的SIFT能更好地适用于室外场景中的关键点检测和特征匹配。

SIFT has proved to be the most robust local invariant feature descriptor in object recognition and matching. It maintains invarianee to various geometrical changes. However, it ignores the color information of object which limits its performance. This paper presents an improved Colored Scale Invariant Feature Transform （CSIFT） method, which combines the color features with the geometrical features of the objects perfectly. The method simplifies the form of the color invariant according to the characteristics of outdoor scenes, and implements the keypoint detection and feature matching for the ourdoor scene objects. The experiment results show that the improved CSIFT method can be better applied to the keypoint detection and feature matching than the conventional SIlT in outdoor scenes.