드론으로 제작된 정사영상의 기하보정을 위한 특징점 추출기법 비교

초록

In the mid-2010s, research on photogrammetry methods using drones was actively conducted, and various sensors for unmanned aerial vehicles were developed, making it possible to quickly update in formation on small areas compared to traditional aerial photogrammetry methods. However, one of the disadvantages of drones is that it takes a lot of time to acquire and input the ground control point. To compensate for this, geometric correction and accuracy researches using real time kinematic drone have been conducted, but it is difficult to obtain accurate location coordinates when the network is dis connected or the server is unstable. In addition, SIFT or SURF techniques were mainly used in geo metric correction research, and there were only studies on geometric correction between single images. In this study, SIFT, SURF, AKAZE, and BRISK methods were compared and evaluated. In area A, SURF was better than AKAZE when time was considered, and when time was not considered, AKAZE extracted many matching pairs. In site B, it was confirmed that the AKAZE method was superior.

키워드