Enhancing Underwater Object Detection through CNN-based Image Enhancement and Classification

This research focuses on object detection using Convolutional Neural Networks (CNN) applied to underwater image datasets. Underwater images often suffer from issues such as low clarity and quality, which pose challenges for accurate object identification. To address this, the research employs image enhancement techniques, including image illumination methods, to improve image quality and facilitate object detection algorithms. Subsequently, the study developed algorithms capable of detecting objects and accurately predicting their categories. The primary objective is to achieve optimal accuracy and efficiency in underwater recognition. This research utilizes Machine Learning techniques through Tensor Flow and Image Processing to accomplish underwater object detection. Deep learning techniques, particularly feature learning, object classification, and detection, have gained significant attention and momentum. In this research we implemented different image enhancement techniques on dataset and evaluated their performance. While one metric, IQI (Image Quality Index), slightly favoured histogram equalization (HE), the other three metrics strongly favoured the enhanced version of HE known as Contrast Limited Adaptive Histogram Equalization (CLAHE).