Image Courtesy: Planet Labs
Planet Labs has demonstrated a new capability in satellite technology by performing AI-based object detection directly in orbit. The company announced that its Pelican-4 satellite successfully identified aircraft within captured imagery in near real time, marking a shift toward onboard data processing in space.
The test took place on March 25 at an altitude of approximately 500 kilometers above Alice Springs. Using an onboard NVIDIA Jetson Orin system, the satellite analyzed images immediately after capture and detected objects without relying on ground-based processing, according to Defence Industry Europe.
Traditionally, Earth observation satellites transmit raw data to ground stations for processing, which can introduce delays ranging from minutes to hours. By moving AI inference directly onto the satellite, Planet Labs aims to reduce this latency significantly, enabling faster delivery of actionable insights.
The development is part of the company’s broader concept of “Planetary Intelligence,” which envisions a network of satellites capable of continuously observing and interpreting changes on Earth. By integrating AI processing at the edge, satellites can detect events such as aircraft movement, infrastructure changes, or environmental shifts almost immediately after imaging.
Company executives stated that the approach could reduce the time between observation and decision-making from hours to minutes. This has potential applications across multiple sectors, including disaster response, environmental monitoring, and security operations, where rapid situational awareness is critical.
The Pelican-4 satellite is part of a new generation of Earth observation platforms designed to support higher-resolution imaging and advanced onboard computing. Future constellations, including the planned Owl satellites, are expected to expand these capabilities further by enabling coordinated data processing and communication between satellites.
Another key aspect of the system is its ability to generate processed outputs, such as geospatial data formats, directly in orbit. This reduces the need for large data transfers to Earth and lowers operational costs associated with bandwidth and ground infrastructure.
While the onboard AI models are still in early stages of development, the successful demonstration highlights a broader shift in satellite design. Embedding computational capabilities in space allows satellites to function not only as sensors but also as processors of information.
The advancement reflects a growing trend toward decentralizing data processing in space systems, potentially transforming how Earth observation data is collected, analyzed, and delivered to users.