Visualization of humpback whale tracking on edge device using space-borne remote sensing data for Indian Ocean

Abstract

The conservation of humpback whale populations faces ongoing challenges, including human-induced mortality, despite the ban on commercial whaling. Recent advancements in high-resolution satellite imagery offer promise for estimating whale populations, particularly in remote and inaccessible regions. However, significant research gaps persist, necessitating innovative approaches for effective monitoring and conservation efforts. This paper presents a novel methodology that integrates high- resolution satellite imagery with state-of-the-art deep learning techniques to monitor and conserve humpback whale populations, with a focus on the Indian Ocean region. Specifically, application of cutting-edge deep learning models such as YOLO for object detection and EfficientNet for classification to automate the detection, classification, and tracking of humpback whales in satellite images is explored. By leveraging deep convolutional neural networks (CNNs), the proposed ensemble system offers a robust and generalizable approach for automatically detecting, classifying, and tracking whales in space-borne satellite imagery, thereby addressing the challenge of uncertain whale populations in the world’s oceans. The results demonstrate promising accuracy and performance metrics: the Segment Anything Model(SAM) achieves an accuracy of 89.2%, YOLO achieves an accuracy of 99.2%, EfficientNet achieves an accuracy of 99% across various tasks.

Keywords

• Space-Borne Remote Sensing
• Semantic segmentation
• Humpback Whale detection
• Whale classification
• Whale Tracking
• Deep Learning