CAJ | 학술논문

针对 SIFT 特征向量检测效率不高、存在误匹配点对等问题，论文在结合其与 Harris 角点检测优点的基础上提出了一种新的特征点检测方法——— Harris＿SIFT 特征点检测。首先利用引导互匹配算法对特征点进行粗匹配，再引入改进的随机抽样一致性算法（RERANSAC）以及非线性最小化迭代算法（LM ）求取精确稳定的变换矩阵，然后结合视频追踪中KLT 算法对匹配点对进行偏移量纠正，最后利用小波变换融合图像。
침대 SIFT 특정향량검측효솔불고、존재오필배점대등문제，논문재결합기여 Harris 각점검측우점적기출상제출료일충신적특정점검측방법——— Harris＿SIFT 특정점검측。수선이용인도호필배산법대특정점진행조필배，재인입개진적수궤추양일치성산법（RERANSAC）이급비선성최소화질대산법（LM ）구취정학은정적변환구진，연후결합시빈추종중KLT 산법대필배점대진행편이량규정，최후이용소파변환융합도상。
In view of the SIFT feature vector detection efficiency is not high ,false matching points equivalence prob-lems ,in this paper ,in combination with the advantage of Harris corner detection ,a new feature point detection method —Harris_SIFT feature[1] point detection is proposed .Firstly ,lead cross matching algorithm is used to coarse matching of fea-ture points ,then improved random consistency algorithm[4] (RERANSAC[9] ) and nonlinear minimization iterative algorithm (LM ) are introducted to calculate the precise and stable transformation matrix ,then KLT algorithm for matching point in the video tracking is combined with to offset correction ,finally the image is mixed by using wavelet transform .