Rapid array images stitching for underwater structures based on camera integrated calibration with multi-scale detectable features pattern

Abstract

Optical inspection has emerged as a mainstream method for underwater structure assessment but presents several limitations, such as restricted fields of view, suboptimal image stitching, and inefficiencies. To address these problems, this paper proposes an image stitching method based on camera integrated calibration, obviating the need for feature extraction. Laboratory experiments validate the efficacy of various stitching methods. Compared to feature extraction-based methods, the proposed method reduces processing time by approximately 60 %, achieving a stitching time of 42 s for high-resolution images. The proposed method also achieves superior image quality, with information entropy of 6.7, an average gradient of 1.8, and a spatial frequency of 5.0. Additionally, a quantitative evaluation reveals that the proposed method achieves a stitching quality score of 64, surpassing the feature extraction-based method's score of 63, with fewer distortions and improved clarity. This paper contributes to advancing underwater multi-camera optical detection methodologies.