ArmSoM CM5 Powers World's First AI Early-Warning Buoy to Safeguard Pilot Whales in New Zealand
- longji luo
- 11 hours ago
- 3 min read
From Harbor Monitoring to Stranding alarm: ArmSoM CM5 Accelerates Global Marine Conservation Science with Open-Source Compute
Following a successful 13-month deployment in Whangārei Harbour, New Zealand, where it reliably monitored dolphins, orcas, and other marine mammals in real-time, the edge intelligence solution powered by the ArmSoM CM5 System-on-Module has now been applied to a more pressing global challenge: early warning for pilot whale strandings. In December 2025, the world's first AI-powered stranding-alert buoy, centered on the CM5, was deployed off Farewell Spit in Golden Bay, New Zealand. This milestone underscores that low-cost, open-source, high-performance edge computing is becoming a key enabler in solving critical marine ecological monitoring challenges.
The Challenge: From Data Deluge to Real-Time Insight
Traditional Passive Acoustic Monitoring (PAM) relies on retrieving underwater recorders for data analysis months after deployment, carrying risks of data loss and significant delays. The complex acoustic environment of the ocean, filled with noise from vessels and other biological sources, poses a major challenge for real-time, accurate AI detection. Researchers urgently needed a high-performance, low-power, and cost-effective edge computing solution capable of processing complex acoustic streams on-device in real-time and supporting continuous over-the-air model updates.
The Solution: CM5 – An Efficient “Ear” and “Brain” for Real-Time Marine Acoustics
Both the “Pilot Pulse” project and the preceding Whangārei Harbour monitoring system chose the ArmSoM CM5 as their computational heart. Based on the Rockchip RK3576 SoC, the CM5’s integrated, high-efficiency NPU and LPDDR5 RAM provide the essential low-latency, high-parallel processing capability required to run sophisticated deep-learning acoustic classification models. This enables a single hydrophone to simultaneously detect multiple cetacean species while effectively filtering out ambient noise.
Balancing Performance & Cost: With an idle power consumption of only 0.5-1W while running Armbian, the CM5 is perfectly suited for long-term, solar-powered buoy operations. Its open-source hardware nature significantly reduces the build cost of the entire real-time acoustic monitoring system, allowing research institutes and non-profits to deploy the technology more widely.
Continuous Evolution: Via cellular networks (LTE Cat M1), researchers can perform remote, over-the-air updates to the AI models on the buoy, enabling a “human-in-the-loop” pipeline for continuous optimization, ensuring long-term accuracy and adaptability in the dynamic marine environment.
Outcomes: From Proof-of-Concept to Scalable Application
In Whangārei Harbour, the CM5-powered buoy system has operated flawlessly for over 13 months, successfully transmitting acoustic data from 219 marine mammal encounters in real-time and accurately logging activities of various species, including 24 orca encounters. Now, in Golden Bay, the same CM5 core is powering AI models aimed at detecting the previously unrecorded “pre-stranding behaviour” of pilot whales for the world's first early-warning system, striving to buy rescuers the vital “golden hour.”
Future Vision:
From port biodiversity surveys to endangered species stranding alerts, ArmSoM CM5 is consistently advancing marine conservation science from a traditional post-hoc analysis model towards a new era of proactive, real-time, and intelligent management. With its robust edge AI performance, exceptional power efficiency, and open-source cost advantage, ArmSoM remains committed to providing a solid and innovative hardware foundation for environmental scientists and engineers worldwide, working together to safeguard the vitality of our blue planet.
News source: