The key attributes of a Sky Hub system are:
Data quality - ensuring accuracy and consistency of data collected at each hub
Affordability - making this as low cost as possible
Scalability - ability to create a large network of trackers around the globe
Longevity - robustness of trackers for long deployment durations
Accessibility - generated data is available for scientists and the public
Versatility - detecting all kinds of unknown aerial phenomenon, perhaps outside of our perception
Some of the above attributes conflict with each other (e.g. data quality, affordability) so much discussion and thought has been put into the decisions of the design of the Sky Hub system by our Hardware and Science Teams. Join the discussion on future hardware decisions in our chat.
The core platform for a Sky Hub Tracker is the Nvidia Jetson platform. Nvidia provide a range of small cutting edge embedded boards that are intended for edge AI i.e. where algorithms are run locally on incoming data before being uploaded to the cloud. The Jetson boards are the Nano, Xavier NX and AGX.
The Jetson boards are designed with specialized hardware including a powerful GPU unit that can be leveraged to rapidly run sophisticated machine learning algorithms on video streams and other sensor data. The price of these devices is also relatively low opening up a range of possible applications that were previously out of reach for the general population until relatively recently.
This cutting edge platform allows us the tracker to run the best machine learning models available in the world today to intelligently identify known objects in our skies and determine anomalous events. Other microcomputer platforms were consider early in the development of this project but were rejected in favour of Jetson due to the low cost and high capability reasons outlined above.
The fisheye camera is used to provide a wide view of the sky and detect events occurring within the vicinity. The goal is that it will eventually be used in combination with other cameras like a Pan Tilt Zoom (PTZ) camera which can be steered and zoomed to provide a closer look at objects.
The quality of the fisheye is an important factor which is why we recommend the Dahua Fisheye cameras, which are very capable for their cost. They are designed for security applications, however the IMX335 CMOS 1/1.7” sensor they use allows for high sensitivity in the visible light and near infrared (NIR) spectrums (300nm UV(A) to 1100nm NIR) making it well suited to our application providing good video captures at day and night. The cameras can provide 25-30fps video stream at 12 Megapixels. In addition, they are IP67 rated which means rain shouldn't be an issue when they are mounted in the inverted position.
The system is made to be deployed outdoors. Initially our design was focused on using POE (Power over Ethernet), but POE and POE+ didn't provide a power source that was stable enough. We recommend not using PoE due to the reasons that for low voltages, the amps can go high to compensate for the power needed. This increases the danger of cables overheating and can lead to fire.
In the end, we opted for a PSU (as typically found in desktop PCs) as it solves a number of issues. Firstly the PSU is a tried and tested source of power that is flexible and stable. It also has a built in fan that we can use to exhaust air from the Sky Hub enclosure and help with temperature regulation.
We recommend not to use the small form factor PSUs - Mini PSUs are meant for small compartments so are therefore not suited to the Sky Hub enclosure. They are not usually cheaper either, and there is a potential risk of overheating inside the enclosure due to the small fan that is employed.
The base level tracker consumes about 100W, so we are recommending a PSU of 350W or greater to allow for future system expansions that are planned e.g. additional of further sensors and cameras.
If you want to place a hub in a remote location to collect data you could potentially use a solar setup. For example, an array of 100w panels plus battery management system, batteries and converters for 5v and 12v.
We recommend a hard-wired ethernet cable directly to the Sky Hub from your Router. We do not recommend wifi to connect to a Tracker as many issues can be introduced with wifi. For example, collisions with "noisy neighbour" wifi channels. Packet loss and latency issues, increased jitter and so forth. We are also aiming to reduce RF around sensors.
We understand that not using wifi may present some complications to some who have a router that is a long distance from where the Tracker will be placed. So one means of extending the range of your router without a long cable through you home is to use a powerline extender such as a TP Link.
The GPS antenna is probably one of the most important aspects of the system. Whilst the Sky Hub Tracker isn't moving, including a GPS antenna addresses multiple issues. First and foremost it provides an accurate location. GPS devices also provide an accurate and stable GPS time so we can tag all of the data with GPS timestamps and GPS location. This is useful for helping ensure the provenance and accuracy of the data. Also if there happens to be more than one Sky Hub Tracker that is observing the same object and you know the GPS Time/Location of all of the Sky Hub Trackers, you can start calculating positions of the object in the sky.
Alternative approaches were of course considered, such using the IP address for getting a location. However, this is not accurate, even if you pay for the "Enterprise" GeoIP database. So, for a relatively low sum we can get an accurate GPS location and time that isn't prone to error like IP GeoIP look ups, or error introduced by humans in manually configuring this information.
First and foremost the enclosure is designed to protect your tracker hardware. While you can use any type of enclosure, the most important aspects are weatherproofness and resistance to ultraviolet light and airflow. The choice of materials and design of this enclosure addresses these issues (5mm white ABS sheet with extra UV layer).
The enclosure also plays an key role in correctly and consistently positioning the tracker hardware components in relation to each other. For example, we plan to support other cameras that will allow for distance and speed calculations and zoomed videos and in order to do this we require the known position of the cameras.