Planet Labs PBC (NYSE:PL) said it has successfully carried out AI-powered object detection directly aboard its Pelican-4 satellite, marking a step forward in real-time space-based data analysis. The test, conducted on March 25, involved capturing imagery of an airport located about 500 kilometers above Alice Springs, Australia, and identifying aircraft shortly after the image was taken.
According to the company, the satellite used an onboard NVIDIA Jetson Orin computing module to perform the detection process within moments of capturing the image.
The milestone represents one of the early demonstrations of an Earth observation satellite performing onboard AI inference and analysis beyond simply collecting data, the company said.
“By running AI at the edge on the NVIDIA Jetson platform, we can help reduce the time between ’seeing’ a change on Earth and a customer ’acting’ on it, while simultaneously minimizing downlink latency and cost,” said Kiruthika Devaraj, VP of Avionics & Spacecraft Technology.
Planet Labs CEO and co-founder Will Marshall said the technology has the potential to significantly shorten the time needed to turn satellite imagery into actionable insights. “This can be the critical difference-maker for our customers from disaster response to security and beyond,” Marshall said.
Planet Labs added that the technology could support the evolution of its Pelican satellites and the upcoming Owl constellation into a near real-time intelligence network. In the test, the full process—from data capture and object detection to geo-rectification—was carried out directly in orbit. The company said the system produced GeoTIFF and GeoJSON data outputs through Docker-based processing containers running in space.
The development follows a previously announced strategic collaboration between Planet Labs and NVIDIA. The company noted that the onboard AI models are still in early development and will continue to be improved over time.