Camera data: Image correlation

참고문헌: 원격탐사개론, 동화기술

*화상 상관 (Image correlation)

화상 상관은 2개의 서로 다른 화상 중에 동일 대상물에 찍혀 있을 때, 두 개의 화상 상호 간에 대응하는 물체의 위치를 인식하는데 사용되는 기술 

활용: 스테레오 화상의 대응점 (tie point) 탐색

        다른 일시에 촬영된 화상 기준점의 자동 인식

-Area matching : 한쪽 화상에 n x n 윈도우를 만들고 그 중심점에 대응하는 점을 오른쪽 화상 내에서 탐색한다. 탐색에 있어서는 오른쪽 화상 상에  n x n의 윈도우를 움직여서 윈도우 화상 간의 유사성을 나타내는 상호 상관이나 상관계수 식이 최대로 되는 점을 구한다. 

-윈도우의 크기는 화상의 해석력이나 특징에 의해 선택된다. 

 항공사진 화상의 경우에는 9x9 로부터 21 x 21 윈도우가 자주 이용된다. 

*처리의 고속화를 위해 다양한 방법 제안

e.g.) Sequential Similarity Detection Algorithm (SSDA)

 

Metashape forum 

Nice questions and answer

https://www.agisoft.com/forum/index.php?topic=10730

https://www.agisoft.com/forum/index.php?topic=12781.msg56630#msg56630

Relationship between tie points and dense cloud from ChatGPT

Certainly! Here's a more detailed explanation of how tie points are used to generate a denser point cloud representation:

1. Tie Point Extraction: Initially, tie points are selected or extracted from multiple overlapping images. These tie points are typically distinctive features or identifiable points in the images, such as corners, edges, or other salient points. Feature detection and matching algorithms are commonly used to identify and match corresponding tie points across the images.

2. Initial Matching: The tie points extracted from the images are used as initial correspondences or matches. These matches indicate the likely correspondences between points in different images that represent the same location or feature in the scene.

3. Triangulation: Once the initial matches are established, triangulation techniques are employed to estimate the three-dimensional coordinates of additional points in the scene. Triangulation calculates the position of a point in 3D space by intersecting the lines of sight from multiple camera viewpoints.

4. Point Cloud Generation: By triangulating additional points based on the initial matches, a denser point cloud is gradually formed. The triangulated points, together with the initial tie points, contribute to the creation of the point cloud. The more tie points and matches available, the denser the resulting point cloud will be.

5. Point Cloud Refinement: After the initial point cloud generation, further refinement techniques may be applied to improve the accuracy and quality of the dense point cloud. This can involve filtering out outlier points, applying bundle adjustment to optimize the camera positions and tie point coordinates, or applying additional algorithms for noise reduction and smoothing.

By using tie points as the starting point for establishing initial correspondences and then triangulating additional points, a denser point cloud representation of the scene can be generated. This point cloud can then be further processed and utilized for various applications in remote sensing and photogrammetry.