CAJ | 학술논문

针对倾斜的遥感影像配准困难问题，提出一种基于集成最大极值稳定区域(MSER)和尺度不变特征转换(SIFT)的互补不变特征的自动影像配准算法。该算法首先应用目前公认的具有最佳仿射不变性的MSER特征区域进行影像的粗匹配，初步校正影像的空间形变。然后在粗匹配基础上采用匹配能力较强的 SIFT 描述子与仿射不变矩描述子相结合，进行精匹配。通过以上两步匹配，可以提高遥感影像配准精度，尤其对倾斜影像效果更明显。最后采用倾斜的无人机(UAV)影像进行试验，并与SIFT配准算法比较。结果表明，本文算法在仿射不变性和匹配正确率方面均优于SIFT配准方法。
침대경사적요감영상배준곤난문제，제출일충기우집성최대겁치은정구역(MSER)화척도불변특정전환(SIFT)적호보불변특정적자동영상배준산법。해산법수선응용목전공인적구유최가방사불변성적MSER특정구역진행영상적조필배，초보교정영상적공간형변。연후재조필배기출상채용필배능력교강적 SIFT 묘술자여방사불변구묘술자상결합，진행정필배。통과이상량보필배，가이제고요감영상배준정도，우기대경사영상효과경명현。최후채용경사적무인궤(UAV)영상진행시험，병여SIFT배준산법비교。결과표명，본문산법재방사불변성화필배정학솔방면균우우SIFT배준방법。
An image matching approach which integrates Maximally Stable Extremal Regions (MSER, Maximally Stable Extremal Regions) and Scale Invariant Feature Transformation (SIFT, Scale Invariant Feature Transformation) complementary invariant feature automatically is proposed for the tilt Remote Sensing image registration. Firstly, the images are coarsely matched by applying currently recognized as the best affine invariant MSER features, and the large deformation images are corrected initially. Then the images are fine matched by the matching ability of the SIFT descriptor joint the moments based on the coarse matching. The remote sensing image matching accuracy is improved through the above two steps, especially, the more pronounced effect on the large tilt images. Finally, the UAV(Unmanned Aerial Vehicle) image experiments show that this algorithm is more effective than SIFT algorithm in the affine invariant and matching the correct rate.