Video Surveillance Weighs Performance with Power



In the quest for smaller pieces of equipment on the battlefield, embedded systems are having to produce more performance in reduced form factors. This in itself is not a stumbling block, but connectivity to ensure the video data is reliably and securely received at command center adds complications.

At this year’s Embedded World, Connect Tech introduced the Astro Carrier for Jetson TXI, NVIDIA’s developer kit for visual computing. The lightweight baseboard is for use with the Jetson TX module is compact enough at 57 x 87mm, to be used for computing-intensive mil/aero applications such as drones and autonomous robotic systems. At the time, the company said that it was the first, commercially available, deployment-ready carrier board for the Jetson TX1 supercomputing module, without having to design and manufacture custom boards.

Figure 1: The Jetson TX1 from NVIDIA is used in the Astro Carrier board and ROSIE embedded module, continuing the company’s relationship with Connect Tech.

Figure 1: The Jetson TX1 from NVIDIA is used in the Astro Carrier board and ROSIE embedded module, continuing the company’s relationship with Connect Tech.

The board connects with off-the-shelf or custom breakout boards, and connectors can be customized to meet a particular design’s requirements and reduce the need for cabling. It is also able to operate across a range of environments, with a temperature range of -40 to +85 ºC.

There are two Gigabit ports, one from Jetson TX1 and one from an on-board controller, together with one High Definition Multimedia Interface (HDMI) port, up to three camera serial interface channels and support for Mini PCIe expansion.

Embedded Module

This was the first entry into the Jetson TX1 ecosystem and the announcement in February was swiftly followed by one in April, at the GPU Technology Conference (GTC). The two companies consolidated their working relationship with the introduction of ROSIE, an embedded system (Figure 2) based on the Jetson TX1.

Figure 2: ROSIE is a rugged, compact, military-qualified embedded system, from Connect Tech, based on NVIDIA’s Jetson TX1.

Figure 2: ROSIE is a rugged, compact, military-qualified embedded system, from Connect Tech, based on NVIDIA’s Jetson TX1.

The compact enclosure has optional mounting brackets and uses NVIDIA Maxwell architecture with 256 CUDA cores to deliver over 1-Tera Floating Point Operations (1TFLOP) performance, with 64-bit ARM-A57 Central Processing Unit (CPU). Dimensions are 163.3 x 146.1 x 99.4mm with the mounting bracket and 163.6 x 108 x 96.3mm without. It weighs just 1.43kg, so can be fitted where space is constrained and where weight tolerances are limited.

The embedded system is designed to MIL-STD 810g and DO-160G for shock and vibration and is rated to IP67/68 ingress protection. It also operates in the extended temperature range of -40 to +85°C.

Connectivity options are varied, with two Gigabit Ethernet ports, a serial RS-232 with modem, four parallel video inputs or Camera Sensor Inputs/ Mobile Industry Processor Interface (CSI-2/MIPI) via coaxial input. As well as two USB 2.0 ports, there is Wi-Fi and Bluetooth 4.0, operating at 24-Mbit/second.

The Jetson development platform can be used to emulate applications, such as guidance systems or terrain analysis, and move them across to other systems, says Eddie Seymour, European Technical Director, NVIDIA. He identifies analytics as significant for the defense market, where a multimedia engine may be used to perform video encode and decode or improve video streams as they are received. In applications such as cartography or reconnaissance, higher resolution, larger images are preferred but these have to be streamed without consuming too much bandwidth. Full High Definition (HD), 4K touch displays have to be compact and consume little power, as the weight and power budget of every component on-board has to be balanced with fuel capacity, power and electricity. Jetson consumes around 5 to 15W, says Seymour, while discrete GPUs can consume 50, 100W, or more.

An Aerial Network

He likens an aircraft (or military tank) to an endpoint on the Internet of Things (IoT). “It passes data back to the command center, which needs a high level of compression. Tegra and Jetson offer 4K encode and decode to pass [graphical data, such as terrain views] to the command center over wireless networks.”

CUDA is used by BAE and Boeing, explains Seymour, to access both the Graphic Processing Unit (GPU)’s virtual instruction set and parallel computing, allowing for deep learning. It allows for discrete GPUs and graphics cards to be plugged into the architecture, so that the footprint can be reduced, and power demands lowered, to suit small, mobile devices for military vehicles and autonomous vehicles, including drones.

Figure 3: The Black Hornet micro drone uses NVIDIA Tegra K1 to provide video images from its three cameras–all in a body that is just 160mm long.

Figure 3: The Black Hornet micro drone uses NVIDIA Tegra K1 to provide video images from its three cameras–all in a body that is just 160mm long.

A drone model that was announced at the GTC was the Black Hornet (Figure 3). The 160mm long, 18g drone has a range of one mile and a flight time of up to 25 minutes. It is quiet in operation, has three cameras and sensors. It has been developed and produced by Norwegian company, Prox Dynamics. It uses a reconfigured NVIDIA Tegra K1 embedded system to provide the 4K-quality video and to maintain connectivity with the hand-held remote control.

The micro drone is quiet in operation, but noise was not one of the biggest challenges faced by the company. John Lund, Prox Dynamics’ research and development software engineer revealed at the conference in Silicon Valley, CA, that it was weight and size that preoccupied the design team. “Weight really guides our entire design process. Every milligram is modeled and accounted for,” he told delegates.

Future challenges are to improve the autopilot capabilities, at present the drone can auto navigate to avoid obstacles. This will place extra demands on the computing performance, added to which, Lund also wants to reduce the size again. “We want to be able to go into tighter spaces and have more autonomy,” he added.  The drone is part of the UK Ministry of Defence (MoD) equipment program and is also used by Norwegian armed forces.

NVIDIA supplies the Jetson TXI Developer Kit, for visual computing. It uses Linux and supports many common APIs. (Figure 4).

Figure 4: The Jetson TXI Developer Kit enables an extensible platform for high computation at low power.

Figure 4: The Jetson TXI Developer Kit enables an extensible platform for high computation at low power.


Caroline_Hayes_ThumbCaroline Hayes has been a journalist, covering the electronics sector for over 20 years. She has edited UK and pan-European titles, covering design and technology for established and emerging applications.

Share and Enjoy:
  • Digg
  • Sphinn
  • del.icio.us
  • Facebook
  • Mixx
  • Google
  • TwitThis

Tags: