Unmanned Systems Technology 002 | Scion SA-400 | Commercial UAV Show report | Vision sensors | Danielson Trident I Security and safety systems | MIRA MACE | Additive manufacturing | Marine UUVs

42 Focus | Vision sensors the image, unlike the deeper infrared. This means images from an InGaAs camera can be comparable to visible images in resolution and detail, and are increasingly used as the payload camera for a UAV for both day and night use. The InGaAs sensors can be made extremely sensitive so when built as an array with millions of pixels, SWIR cameras will work in very dark conditions. Using SWIR at night has another major advantage. An atmospheric phenomenon called night sky radiance emits five to seven times more illumination than starlight, nearly all of it in the SWIR wavelengths. This means a SWIR camera can use this night radiance – often called nightglow – to ‘see’ objects with great clarity on moonless nights. Falling cost and weight Infrared vision has become a more significant capability for controlling autonomous systems as the cost and weight of these specialist sensors have come down. This has been a result of the combination of new sensor technologies and their integration into a camera, which is not simple. Previous IR cameras have had to be cooled, making them too heavy for most autonomous systems, but a new generation of uncooled cameras opens up fresh applications, and these – either as modules or finished units – are being used increasingly for control systems and as the payload. However, the post-processing of images and the integration of the camera into the autonomous control system varies depending on whether it is used for control or payload. This often depends on the sensor characteristics, making it a complex development that has to balance the sensor array and the optics. The future of small infrared cameras lies in increasing the number of pixels. The current generation has a small number of pixels, typically an 80 x 60 (4800) array with a horizontal field of view of 50 º , giving a field of view of each pixel of 0.625 º per pixel. That’s a wide angle per pixel and really limits the ability to see small targets at any distance – cameras with a small number of pixels and wide fields of view are essentially near-sighted. There’s nothing you can do to change that for a given sensor size unless you reduce the field of view by using optics with a longer focal length, but then if you have a really narrow field of view it’s like looking down a tube, limiting the ability to see what’s happening. That is driving the industry towards developing sensors with more pixels. Another way to increase system angular resolution is to stitch the images together from multiple cameras looking in different directions. There are systems such as the ‘thermal grenade’ camera, which is a sphere with cameras around it pointing in all directions, allowing a larger field of view to be synthesised using multiple sensors. Alternatively there are multi-spectral algorithms that blend visible-light images with IR images. This takes the visible image and throws out all the information except for the edges of objects, and this is added to the thermal image. IR-visible combination For example, some cameras combine the smaller IR sensor with a 640 x 480 visible camera using these algorithms. The visible camera produces an image with decent edges if there is enough visible light. The advantage is that the visible camera doesn’t have to be high resolution to work, and when those edges are blended with the 80 x 60 IR sensor it can be very effective for both payload and control applications. The pixels are made using micro- machined mechanical systems (MEMS) – tiny silicon structures a few microns in size that are grown on a substrate. These silicon micro-bridges are then Spring 2015 | Unmanned Systems Technology Infrared vision has become more significant for controlling autonomous systems as the cost and weight of the sensors have come down This intelligent vision acquisition board uses a field-programmable gate array (Courtesy of 3D-One)