CAJ | 학술논문

针对线阵CCD 推扫式传感器的每条线阵影像为中心投影，扫描行影像为平行投影，各扫描行影像具有自身的外方位元素。由于飞行方向阵列影像间发生变形，难以建立严密核线模型。提出基于马氏距离和视差约束的线阵影像多级匹配算法。首先对线阵影像进行金字塔分层匹配，然后采用马氏距离和视差约束进行特征点筛选，剔除误匹配点，最后进行最小二乘子像素匹配。实验结果表明，本算法获取匹配点分布均匀、可靠性强。
침대선진CCD 추소식전감기적매조선진영상위중심투영，소묘행영상위평행투영，각소묘행영상구유자신적외방위원소。유우비행방향진렬영상간발생변형，난이건립엄밀핵선모형。제출기우마씨거리화시차약속적선진영상다급필배산법。수선대선진영상진행금자탑분층필배，연후채용마씨거리화시차약속진행특정점사선，척제오필배점，최후진행최소이승자상소필배。실험결과표명，본산법획취필배점분포균균、가고성강。
Aiming at every line array image of the push-broom sensor of linear CCD as center projection, line scan image is the parallel projection, and each scan line image has its exterior orientation elements. Due to the flight direction between array image distortion, it is difficult to establish strict nuclear line model. Based on Mahalanobis distance and parallactic restriction the multi-stage matching algorithm of linear array image is proposed. Firstly, the line array images do pyramid hierarchical matching, then use Mahalanobis distance and parallax constraints to select feature point, and eliminate false matching points. At last, the least squares sub pixel matching is carried out. The results show that the algorithm can make points distribution even and have high reliability.