Elmo Ranolo, Archival Sebial, Anthony Ilano, Angie Ceniza Canillo
Underwater ecosystems, comprising seagrass, is one of the integral components of marine biodiversity, requiring meticulous observation and analysis to ensure their preservation and ecological balance. However, the complex and challenging visual conditions in underwater environments demand innovative approaches to object recognition and detection. This research investigates the efficacy of Contrast-Limited Adaptive Histogram Equalization (CLAHE) and the You Only Look Once (YOLO) algorithm in enhancing object recognition within blurred fluorescence images of underwater ecosystems. The study specifically focuses on seagrass as this organism offers vital insights into marine habitat health. By conducting a comparative analysis of YOLO models and YOLO with CLAHE, this research aimed to identify the most effective and accurate methodology for detecting and recognizing underwater organisms in blurred fluorescence images. These insights contributed significantly to marine ecology research and facilitated practical applications in environmental monitoring and conservation efforts. The comparative analysis of these results sheds light on the impact of using blurred sample images versus authentic samples for training the YOLO-V3 model. In this analysis, YOLO-V3 was found to be the preferred algorithm for the task at hand, although it occasionally fell short in detecting all the required objects in test images. In contrast, YOLO-V5 exhibited impressive proficiency in identifying nearly all blurred images. Additionally, applying the CLAHE algorithm to CLAHE-equalized images had the potential to substantially boost the accuracy of YOLO-V5, leading to improvements of 8% to 10% in the results. © 2023 IEEE.
University of San Carlos, Department of Computer and Information Sciences and Mathematics, Cebu City, Philippines; University Carmen Campus, Cebu Technological, Carmen, Cebu, Philippines
