Study on Elimination Algorithms for Line Segment Mismatches

Image matching is a key step for remotely sensed image registration and digital elevation model (DEM) generation. Compared with point matching, few studies have focused on line matching for images, especially elimination algorithm of mismatched line segments. Therefore, this work systematically stud...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Canadian journal of remote sensing Ročník 48; číslo 3; s. 400 - 410
Hlavní autoři: Li, Chang, Jia, Wenqi, Wei, Dong
Médium: Journal Article
Jazyk:angličtina
Vydáno: Taylor & Francis 04.05.2022
Taylor & Francis Group
ISSN:0703-8992, 1712-7971
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Image matching is a key step for remotely sensed image registration and digital elevation model (DEM) generation. Compared with point matching, few studies have focused on line matching for images, especially elimination algorithm of mismatched line segments. Therefore, this work systematically studies elimination algorithms of line segment mismatches by combining 2 transformation models (i.e., affine and homography) with 2 M-estimators or 2 sample consensus methods (i.e., random sample consensus, RANSAC, and least median of squares, LMedS). The main idea is as follows. After line segments are extracted and matched, the proposed algorithms can automatically remove mismatched line segments based on an error function of line segment. Aerial images with panchromatic bands and standard false color synthesis were selected for testing. Experiments were performed to compare different combinations of these models and methods and to quantitatively evaluate the performance of the algorithms in terms of accuracy and run time. The results show that the proposed algorithm can be effectively applied to automatically eliminate mismatched line segments, and among all combinations the homography model with LMedS performs the best. The algorithm can also ensure and control the quality of line segment matching from stereo pairs.
ISSN:0703-8992
1712-7971
DOI:10.1080/07038992.2022.2052032