Low-power Ship Detection in Satellite Images Using Neuromorphic Hardware

Transmitting Earth observation image data from satellites to ground stations is costly in terms of power and bandwidth. To address this, a low-power, two-stage system for maritime ship detection that optimizes performance by processing data on-board was designed. The first stage employs a lightweight binary classifier on Brainchip’s Akida 1.0 to detect the presence of ships, significantly reducing power consumption. The second stage uses a YOLOv5 object detection model to accurately identify ship locations and sizes. This approach not only improves detection accuracy but also drastically cuts energy consumption, demonstrating the efficiency of heterogeneous computing systems. Read the full paper here: https://arxiv.org/html/2406.11319v1