Visualization of Fin Whale Tracking on Edge Device using Space-Borne Remote Sensing Data of Indian Ocean

Abstract

The conservation of fin whale populations faces formidable challenges exacerbated by human-induced mortality despite the ban on commercial whaling. Recent advancements have highlighted the potential of high-resolution satellite imagery for estimating whale populations, especially in remote and inaccessible areas. Several research gaps persist, offering opportunities for further exploration and innovation in this field. This paper proposes the application of high-resolution satellite imagery and deep learning models to detect and track fin whales in the Indian Ocean, focusing on specific spectral bands crucial for underwater visibil- ity. This study investigated the application of cutting-edge deep learning models, including U-Net, enhanced YOLO and ResNet101, to automate the detection, classification, and tracking of fin whales in satellite and infra-red images. It pro- vides a scalable and adaptable solution for marine conservation efforts, addressing the challenges posed by remote and inaccessible regions. Outputs from deep learn- ing models are displayed on user interfaces connected to the edge devices such as mobile devices, providing real-time information for informed decision-making. The results demonstrate promising accuracy and performance metrics: the U-Net model achieves an accuracy of 92.21, YOLO achieves 0.82 mean average Precision, and ResNet101 achieves 99 percent accuracy across various tasks.