Scanning-sonar Image Matching Algorithm Using Overlapping Region Detection and ICP
Gwonsoo Lee, Kihwan Choi, Phil-Yeob Lee, Sukmin Yoon, Juwon Son, Jihong Lee · Journal of Institute of Control Robotics and Systems · 2024
This paper describes a method for matching scanning-sonar images using the iterative closest point (ICP) algorithm. In the case of scanning-sonar, data cannot be taken while moving, resulting in only a partial overlap between the datasets. Furthermore, the ICP algorithm is based on point-cloud data, requiring three-dimensional data. However, scanning-sonar images are two-dimensional data. These issues make it challenging to match scanning-sonar images using the ICP algorithm. This study proposes a three-step method as a solution to address the challenge. First, the overlapping area between two scanning-sonar images is estimated using the localization result from a navigation algorithm. Second, a method is proposed for transforming two-dimensional image coordinates and pixel intensities into three-dimensional data. Lastly, feature points are extracted from noisy data using histograms. The proposed algorithm was validated using scanning-sonar images collected from a real-sea environment. The experimental results showed that the proposed algorithm successfully matches scanning-sonar images.