Multi-target positioning method based on adaptive k-means clustering

The invention provides a multi-target positioning method based on adaptive k-means clustering, a density peak clustering algorithm (DPC) and a k-means clustering algorithm are combined, the adaptive k-means clustering algorithm is provided, the number of targets to be positioned is adaptively determined based on extracted feature points, and feature point sets of different targets are clustered; and rough matching is carried out through a nearest neighbor ratio algorithm, optimal geometric constraints are constructed by utilizing feature point voting to carry out fine matching, and multi-target accurate positioning is realized. According to the method, for different types and numbers of to-be-positioned targets, the targets can be accurately positioned in complex environments such as rotation, scale transformation, partial shielding and illumination transformation, and the robustness is good. 本发明提出一种基于自适应k均值聚类的多目标定位方法，密度峰值聚类算法(DPC)和k-means聚类算法相结合，提出自适应k-means聚类算法，基于提取的特征点，自适应确定待定位目标数量，对不同目标的特征点集聚类；通过最近邻比值算法进行