Unmanned Systems Technology 005 | Selex ES Falco UAV | Sense and avoid systems | RCV Engines DF70 | DSEI show report | Fuel cells | CUAV Expo, InterDrone and CUAV Show reports | SLAM

7 Platform one Unmanned Systems Technology | Dec 2015/Jan 2016 Dutch sensor and camera developer 3D-One has launched an OEM camera system that is optimised for unmanned aircraft in agricultural applications. The system uses two hyperspectral sensors from research laboratory IMEC in Belgium that cover visible and near- infrared light, with embedded processing. A field-programmable gate array (FPGA) processes the images in real time to build a hypercube of data that is then managed by an embedded computer running the Linux operating system. “Rather than line scanning, the IMEC sensors in the 3D-One system use a snapshot which is required on a UAV because the UAV is not stable enough for line scanning,” said Marco van Hout, business development manager at 3D-One. “That makes it a lot easier to get good data – typically you make 30 frames per second, and in each image acquisition you capture all 40 spectral bands.” The sensors use a CMV2000 2K x 1K CMOS array from CMOSIS that is usually used for HDTV cameras. IMEC adds a hyperspectral sensing layer over the top, creating macropixels onto a matrix of 5 x 5 pixels. This allows each macropixel to sense up to 25 bands. “That’s delivering a stunning amount of data, so we developed an acquisition and processing engine that pre-processes the data coming from the sensor and presents it to an embedded computer,” said van Hout. “It’s a two-stage processing engine where the first stage is an FPGA for real- time processing that reconstructs all the information into one hypercube. We do all sorts of corrections as recommended by IMEC so that the data is ready for analysis, and we add information from the GPS navigation system with a timestamp on each image and data from the inertial measurement unit. “We can also do a reflectance calculation from a spectral meter,” van Hout said. ‘This takes the reflectance of the incoming light, for example when flying a UAV over a field of crops; at certain points the sun is shining directly on the field, then you fly into a shadow and you don’t want this to influence the measurement. All these go into the FPGA to prepare the data for spectral analysis on the UAV.” The processing algorithms have agricultural applications in mind. The total system of the two cameras, acquisition and processing engine weighs less than 400 g, including the computer, which has an Intel quad-core processor using the PC-104 board format and the Qseven embedded module standard. “We have used an Altera Cyclone 5 FPGA on the acquisition board and a mass storage SSD up to 1 Tbyte with a SATA interface,” van Hout said. “That is more than enough to store the images produced during the flight. “Another thing we have to consider is power consumption, since every Watt you use is taken from the same battery that is used for flying,” he said. The system uses between 12 and 15 W depending on the processing requirements. Camera has crops in its sight Image processing Schematic of the 3D-One camera and image processing system