Contrast Enhanced Object Recognition (CLAHE_YOLOv3) Technique for Clear and Medium Level Turbidity Lake Underwater Images

Abstract

This study discusses object recognition based on the underwater image that has to cope with physical particles, especially in lake underwater environments, making it difficult to achieve high-quality underwater images. In this study, we have developed a controlled condition image database specifically for lake underwater images in different turbidity levels. The developed database can be accessed at the following link: https://bit.ly/3thcM2w. It is based on 5 different object classes, which are Fish, Aeroplane, Helicopter, Luggage and Submarine. Each set of objects contains 1152 images. The test images are selected from 2 categories of water conditions, which are the clear water and medium turbidity water classes. The object recognition technique of the YOLO version 3, YOLOv3, is used as an algorithm to recognize the object. The proposed method introduced in this study is the combination of the image enhancement technique, the CLAHE and object recognition of YOLOv3, CLAHE_YOLOv3. The proposed method has improved the average accuracy of object detection by using the YOLOv3 alone by 11.83% for both clear and medium turbidity conditions of lake underwater images.