Effectiveness of Clahe Preprocessing in Image Detection of Juvenile Ayu Using Deep Learning

Key Points

• The main aim is to efficiently and accurately determine the ascent count of juvenile ayu using underwater camera images.
• Applied object detection techniques using deep learning on underwater camera images.
• Introduced CLAHE preprocessing to enhance image features.
• Compared processed and unprocessed images to evaluate detection accuracy.
• Proposed a multimodel detection approach combining both image types.
• CLAHE preprocessing highlighted unclear features of juvenile ayu, reducing omissions.
• Similar enhancements were observed for other aquatic organisms like shellfish.
• Reduction in misdetections due to bubbles and water surface shadows was limited but present.
• The multimodel detection approach indicated potential for further improvement in detection accuracy.