JMEE Home · Issue Contents · Forthcoming Papers

Video Stitching Method Based on Manned Deep Diving in Deep-Sea Hydrothermal Vent Areas
Zhongjun Ding, Guangyang Ma, Xingyu Wang and Chen Liu

In order to further explore the ecology, structure and topography of the deep-sea hydrothermal zone, a series of image data obtained by manned submersible for close observation of the hydrothermal zone are of high research value. This paper proposes a method for image stitching of deep-sea hydrothermal vent areas based on observations from manned submersibles. Firstly, it corrects color deviations in deep-sea images using red channel compensation and gray world algorithm. Then, it adjusts the brightness and contrast of the images globally and locally by combining improved gamma correction and the CLAHE algorithm, resulting in enhanced images. An adaptive weighted fusion algorithm is utilized to merge the images. Key frames are extracted from the video based on the structural similarity of inter-frame images, reducing the impact of manual operations such as hovering and varying speeds during the submersible exploration on the stitching quality. The SURF algorithm is employed to extract feature points, and then KD tree algorithm is used for coarse matching and KNN classification algorithm to refine matching points, from which motion vectors are calculated. Transformation matrices are derived from these motion vectors, and images are stitched iteratively to obtain the overall morphology of deep-sea hydrothermal vent areas. The experimental data used in this study were obtained from a certain voyage of the ‘Jiaolong’ manned submersible. Experimental results demonstrate the effectiveness of the proposed method.

Keywords: manned submersible, underwater image enhancement, keyframe extraction, image stitching