Efficient approaches to detect and classify Marine fish species using deep learning

Abstract

The goal of this research is to improve the performance of the newlinealgorithms for marine fish species detection and classification using newlinedeep learning frameworks. Marine Ecological Management systems newlinepreserve the ocean resources such as different species of fishes for the newlinebenefit of humankind, driving sustainability in the long-term. The newlinemarine observatory systems utilize underwater vehicles with highdefinition newlinecameras and sensors, Artificial Intelligence (AI) to monitor newlinefish migration and to classify fish species. Such a system influences newlineremote sensing approaches to track specific species, identify different newlinetypes and count the number of fishes within a given species in each newlinedepth range of the ocean. In this research, the performance of the fish newlinespecies detection and classification models is improved using the Mask newlineR-CNN based frameworks. It includes: (i) Developing Dynamic newlineclassifying algorithm for achieving effective fish species detection and newlineclassification, (ii) Introducing Optimal Deep Kernel Extreme Learning newlineMachine (ODKELM) classifying model with adjusting convolution newlinelayer structure to support the marine fish species detection and newlineclassification, and (iii) Architecting a dual-stage deep learning strategy newlinefor recognising and categorising a moderate fish model for newlineVI newlineautomatically fish species detection and classification. The first phase of newlinethe work proposes a unique dynamic classifying algorithm to identify newlinefish species and supervise fish activities to better understand newlinesynapomorphies. Mask Region Based Convolution Neural Network newline(Mask-R-CNN) enhances feature vectors from video samples to newlineimprove fish detection and tracking. In the second phase, Intelligent newlineDeep Learning for Marine Fish Species Classification (IDL-MFSC) is newlineproposed to efficiently detect and categorise marine fish species without newlinedisturbing their marine ecosystem habitat. Weiner filters remove noise newlineartefacts during pre-processing, followed by Mask-R-CNN marine fish newlinedetection. Water Wave Optimization (WWO) uses an Optimal Deep newlineKernel Extreme