lswcap
Intel announced the details of its contribution to the small satellite PhiSat-1 launched on October 21 (local time) in solar synchronous orbit on September 2.
The PhiSat-1 is equipped with a hyper spectrum thermal camera and a Movidius Myriad Vision Processing Unit (VPU). These VPUs are also installed in many home appliances on the ground, but this time, they are launched into space for the first time, processing large amounts of local data, saving the research team time and satellite downlink bandwidth.
Specifically, the AI onboard PhiSat-1 automatically identifies clouds among images of Earth that scientists actually want to see. By removing these clouds before transmitting the image, the satellite can realize up to 30% bandwidth savings and transmit more useful data when in a position to communicate with the ground station.
PhiSat-1 AI software, powered by Intel Myriad 2, was developed by Ubotica, a startup that works with hardware companies dealing with hyperspectral cameras. Adjustments are also required to compensate for excessive radiation exposure, but CERN tests have shown that no hardware changes are required to meet the mission-specific criteria.
In the sense that edge computing is applied to satellites in orbit, it is clear that the new local AI makes sense. The reason companies want to process data processing and analysis near sensors is the same in space, but due to the difficulty of accessing the network and the quality of the connection. Related information can be found here .
Add comment