Space
How AI innovations are revolutionising Earth observation
By Thomas Porchez, Teledyne e2v T
hrough access to Earth observation imagery, it is possible for a broad array of different potential use cases to be addressed. These include the recording of glacial movements, species migration and deforestation, as well as the monitoring of land erosion, flooding, urban sprawl, etc. With the image sensors employed for such tasks continuously evolving, the resolution and dynamic range that can be delivered by these devices have both improved markedly. Consequently, however, the amount of data associated with each image has likewise increased - and this brings a whole new set of challenges. As this article will explain, it is now necessary for decisions on acquired images to be made at source via artificial intelligence (AI).
Previously the data generated by image acquisition would have normally been relatively easy to manage, meaning that satellites could transmit it back to the Earth’s surface, where it could subsequently be analysed. Due to the much greater data quantities that each captured image now represents, the strain which is starting to be placed on downlinks is intensifying. A new approach is needed which respects these satellites’ bandwidth limitations.
Data processing at source Taking care of the image classification at the point of capture allows several major operational benefits to be realised. It means that a far smaller number of images will need transmitting (with only the ones deemed to be of actual interest being sent back). Bandwidth usage is thereby reduced substantially, and so is the power budget that must be assigned to such activities. Alongside all this, system reaction times will be much quicker - as there will be no latency issues to worry about. This is something that will be vital in circumstances where actions need to be taken quickly (such as dealing with natural disasters, like tsunamis, earthquakes or wildfires). If, during on-board processing, images
20 October2022
Figure 1: The Lynx processing single board computer from Beyond Gravity
are found to feature items that are likely to be worth further examination or suggest signs of a situation emerging that will need to be rapidly addressed, then they can be transmitted. On the other hand, if nothing of interest is seen in them, they can be discarded, with both power and bandwidth being saved.
Defining appropriate processor technology
It is paramount that the processing devices specified for space applications are compact and power efficient, due to the constraints faced in terms of available room and power reserves. They must also exhibit qualities that go way beyond those employed in a terrestrial context. The mission-critical nature of such applications means that exceptional levels of reliability are mandated. As with any space-deployed hardware, they will need to withstand shocks, vibrations and large temperature fluctuations too. Resilience to radiation is also essential, so that the risk of single event latch-ups (SELs) and single event upsets (SEUs) can be mitigated. In addition, these devices must be able to cope with the prospect of heavy total
Components in Electronics
ionizing dosage (TID) figures - otherwise their working lifespan will be curtailed.
Space-targeted processors in action
When Swiss space systems integrator Beyond Gravity started working on a high-performance image data processing platform for AI-based Earth observation tasks, it chose Teledyne e2v to provide the necessary semiconductor technology. This allowed the company to ensure that the processing devices incorporated into their platform not only met the performance benchmarks required, but were also rugged and supported prolonged operation, while still only drawing minimal power. The Teledyne e2v LS1046-Space processors would be utilised in Beyond Gravity’s Lynx platform. Running speeds of up to 1.8GHz, these application-optimised space-qualified processors are highly suited to real-time data classification/ categorisation tasks, with capabilities that rival solutions are unable to match. Each device
relies on a multi- core processor architecture, which
Figure 2: Teledyne e2v’s highly popular LS1046- Space processor
consists of four 64-bit Arm Cortex A72 cores. Through these cores, a 30k DMIPs combined processing output can be delivered. Also included are a cutting-edge DDR4 memory controller with embedded 8-bit error corrected code (ECC) for maintaining ongoing data integrity, a 2MB L2 cache, plus a multitude of interfaces (10Gbit Ethernet, PCIe Gen 3.0, SPI and I2C). With a 23mm x 23mm footprint, this processor model takes up very little board area.
LS1046-Space processors meet both NASA Level 1 and ESA ECSS Class 1 qualification guidelines, and their operating temperature range covers -55⁰C to +125⁰C. These devices have been tested to 100krad TID levels and they exhibit 60
MeV.cm²/mg SEL and SEU radiation tolerance.
Conclusion
By applying sophisticated AI algorithms to processing activities undertaken on satellites, the inconvenience of having to transfer large quantities of imaging data back to Earth may be avoided. Thanks to the Teledyne e2v processor devices leveraged, Beyond Gravity’s Lynx single board computer has the capacity to give satellites and other forms of spacecraft the intelligence needed for them to make decisions quicker and without so much resources being used up.
https://www.teledyne-e2v.com/en/
www.cieonline.co.uk
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62
