Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{208194,
        author = {Anupriya Gupta and Sakthivel Murugan S},
        title = {Underwater Coral Morphology Classification and Tracking: Spatio-Temporal and IoU},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {4},
        pages = {916-923},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=208194},
        abstract = {Automated analysis of underwater coral imagery is challenging due to variations in illumination, complex backgrounds, partial visibility, and the small size of some coral regions. This paper presents a coral-focused spatio-temporal pipeline for detecting and classifying coral morphology in underwater video. The approach uses an existing YOLO-based detector to identify coral regions, followed by overlapping tiled inference to improve spatial coverage. Detected coral regions are classified using a MobileNetV3-Small morphology classifier trained on 10,605 coral image patches across nine morphology categories. To improve temporal consistency, detections across consecutive frames are associated using Intersection over Union (IoU)-based tracking, and a stable morphology label is assigned at the track level. Additional filtering is applied to remove unrealistic, short-lived, and highly overlapping detections. The system was evaluated on a 911-frame underwater video and produced 2,867 candidate detections during tiled processing. After temporal association and subsequent cleanup, 1,443 detection records across 137 tracks were retained, with eight morphology categories represented in the final output. The best validation accuracy of the morphology classifier was 63.22%. The work demonstrates a practical adaptation of existing underwater computer-vision methods for coral morphology analysis while identifying the need for controlled baseline experiments to quantitatively evaluate the effect of spatial tiling and temporal stabilization.},
        keywords = {Coral morphology, underwater computer vision, YOLO, MobileNetV3-Small, tiled inference, object tracking, temporal consistency.},
        month = {September},
        }