部分遮挡目标的稳健局部特征点提取方法

摘要

Feature extraction in partially-occluded object is a challenging problem in the field of object detection. For the partially-occluded object,a robust local feature points extraction method based on the statistical analysis of robust measurement is proposed to solve the problem. It’s built on the basis of gradual purification of local feature points in object training set. First, SIFT algorithm is used to get the corresponding local feature points in the object set which are obtained under various condi-tions. Second,the local feature points are extracted roughly according to the statistical analysis. Then,the relatively robust local feature points are screened out by calculating comprehensive confidence of all points. Finally,the spatial distribution and other factors of local feature points are analyzed and the most robust local feature point set is extracted. The experimental results of ob-ject detection show that this method not only ensures the accuracy,but also increases the speed of detection. The time-consump-tion is only 30% of SIFT detection algorithm.%部分遮挡目标的特征提取是目标检测中的难点问题。针对这一问题，提出一种基于稳健性度量统计分析的部分遮挡目标稳健局部特征点提取方法，该方法建立在目标图像训练集的局部特征点逐步提纯基础上。它首先对目标在不同条件下获取的图像集分别应用SIFT提取相应局部特征点；然后基于统计分析对局部特征点进行粗提取，再通过计算置信度进一步筛选出较稳健的局部特征点；最后分析局部特征点的空间分布等因素提取出最稳健的局部特征点集。实验结果表明，该方法耗时仅为SIFT检测算法的30%，并且保证了检测精度。