Color Information-Based Automated Mask Generation for Detecting Underwater Atypical Glare Areas

TL;DR

This paper tackles detecting underwater breath bubbles for diver monitoring by eschewing data-hungry supervised methods in favor of an unsupervised K-means clustering approach. It enhances clustering by fusing multiple color spaces (RGB, Lab, HSV) and augmenting each pixel with relative spatial coordinates, while CLAHE and image resizing manage underwater noise and computational load. A three-stage pipeline (preprocessing, pixel clustering, post-processing) yields glare masks that isolate breath-bubble regions, with sub-mask refinement via erosion and contour-based filtering. Experiments in an indoor water tank show that color-space fusion and coordinate information significantly improve detection, achieving a mean IoU around 0.6562 for the best configurations and highlighting practical potential for underwater diver safety robots, though robustness across datasets remains a challenge and domain adaptation is proposed for future work.

Abstract

Underwater diving assistance and safety support robots acquire real-time diver information through onboard underwater cameras. This study introduces a breath bubble detection algorithm that utilizes unsupervised K-means clustering, thereby addressing the high accuracy demands of deep learning models as well as the challenges associated with constructing supervised datasets. The proposed method fuses color data and relative spatial coordinates from underwater images, employs CLAHE to mitigate noise, and subsequently performs pixel clustering to isolate reflective regions. Experimental results demonstrate that the algorithm can effectively detect regions corresponding to breath bubbles in underwater images, and that the combined use of RGB, LAB, and HSV color spaces significantly enhances detection accuracy. Overall, this research establishes a foundation for monitoring diver conditions and identifying potential equipment malfunctions in underwater environments.

Figures (7)

• Figure 1: Flow of the proposed glare detection algorithm and input/output examples of each phase: (a) Input image; (b) Image after pre-processing (brightness enhancement, image color space combination, coordinate information channel supplementation, and image resizing); (c) Pixel-clustering results showing clusters per channel; (d) Binary region obtained using cluster information; (e) Output image.
• Figure 2: The original RGB image input into the preprocessing phase and the phase output, which includes the GBL image converted to a color space with three channels, along with the X and Y coordinate information channels.
• Figure 3: Clustered images for each channel generated by the pixel clustering phase.
• Figure 4: Example of the sub-mask generation process in the post-processing phase: (a) ensembled result, (b) result after applying erosion-based filtering, instance assignment using the marching squares technique, and area-based filtering, (c) result overlaid on the original input image.
• Figure 5: The images of the respiration bubble formation, expansion, and dissolution stages: 10 consecutive frames with results shown at 2-frame intervals. (a) Results of the respiration bubble formation and expansion stages, (b) Results of the respiration bubble dissolution stage.