The object detection job is widely employed in many different applications in our daily lives, including pedestrian location and medical imaging systems. As undersea observations and exploitation progress rapidly, a wide variety of marine organisms can be found in this fundamental marine ecosystem, which also provides significant normal and economic benefits. Due to the problems caused by lowered conditions, such as light maintenance, distribution, and assortment reshaping, advanced picture-taking care procedures are required to examine and proportion coral reefs. In this study, the underwater photos are preprocessed using a Gaussian channel. The proposed method reduces the noise in the input by preprocessing it with a Gaussian filter. The level set approach will then be used for training, dividing the preprocessed data into multiple classes. Finally, we used a fuzzy-based SVM classifier to classify the submerged images. Assessment on the dataset of EUVP shows that projected fuzzy logic-based hybrid SVM maintains a surprising small parameter size while achieving good segmentation results. Given that it achieves efficient coral reef segmentation with a notably reduced computational complexity, this demonstrates that FHKSVM is well-suited for real-world applications. With an average sensitivity of 97.11%, specificity of 99.44%, and segmentation accuracy of 98.95% on the underwater Coral Reef image dataset, the proposed model can outperform existing models currently in use.