摘要
描述了一种局部不变特征点的方法,该方法能够用于在物体、场景的不同视图之间完成可靠的匹配。其具体方法是通过使用快速最近邻算法将来自已知对象的特征数据库的各个特征匹配,随后进行霍夫变换以识别属于单个对象的群集,最后通过用一致姿态参数的最小二乘解决方案进行验证。这种识别方法可以在实现接近实时性能的同时,较好地识别匹配图像中的对象。
This paper describes a method of locally invariant feature points that can be used to aeilieve a reliable match between objects,different views of the scene. Specifically,each feature of a feature database from a known object is matched by using the fast nearest neighbor algorithm,followed by a Hough transform to identify a cluster belonging to a single object,and finally verified by a least squares solution for uniform attitude parameters.This method of identification can achieve near real-time performance, better identification of matching objects in the image.
出处
《山西科技》
2017年第5期123-126,共4页
Shanxi Science and Technology
关键词
特征点
目标识别
尺度不变性
图像匹配
feature point
target recognition
scale invariance
image matching
