Potential applications include: fixed and mobile installations, including covert CCTV monitoring and surveillance, and unmanned aerial vehicles (UAVs). Such systems face common problems, including poor raw sensor imagery and very challenging lighting or atmospheric conditions. These challenges have led to an increasing demand for imaging in the infrared (IR), and the need for adaptive contrast enhancement to combat poor imagery in challenging lighting conditions.
The emergence of extremely high performance computing platforms with low size, weight and power, coupled with high performance and compact visible and IR sensors, promises to revolutionise video surveillance. RFEL's high performance video processing blocks are highly optimised and enable HALO to be compact and low power for portable use. HALO delivers high quality, high frame rate, HD video for human or machine interpretation, even in the most challenging of visual environments. HALO's modular, flexible design makes it ideal for retrofitting/upgrading existing infrastructure, as well as for providing a complete, innovative solution for new designs.
Alex Kuhrt, RFEL's CEO, said, RFEL has built its reputation by designing and licencing state-of-the-art digital signal processing solutions to run on the latest FPGAs. Increasingly, customers are asking us to provide complete final products as they know that we can optimise every aspect of the design to provide the very best performance. HALO draws on this experience and is the first product that we have designed for sale directly to customers who want an off-the-shelf solution to enhance video for surveillance in a wide range of security applications.
The video processing capabilities provided by HALO are:
•Intelligent fusion of multi-modal imagery, such as from a visible and IR sensor.
•Image stabilisation, even when the platform is subject to severe vibration, and when imagery is sparse in features or of low contrast.
•Contrast enhancement to maintain high performance operation in marginal lighting conditions -- visible and IR.
•Noise reduction for optimising operation in low ambient light and for ensuring robust image fusion.
•Digital zoom, lens distortion correction, image overlay and support for compression standards.
HALO uses the latest Field Programmable Gate Array technology to achieve the high processing bandwidths needed while maintaining low power consumption from a power supply of between 4.7 to 27V DC. The small form factor of the HALO processor board is housed in a rugged enclosure. The full system measures 105mm x 105mm x 80mm and weighs less than 400g.
HALO offers a range of physical interfaces and supports all major video standards from analogue video (e.g. PAL/NTSC/S-Video) through to digital standards such as Channel Link standards and GigE, both with full 1080p. The required interfaces can be configured in the factory to tailor the system to bespoke applications.
At the heart of HALO is a proprietary image fusion engine. Different imaging wavebands, such as daylight and IR, provide complementary information, especially at certain times of the day and when the atmosphere reduces the quality of daylight images. The purpose of image fusion is not to simply blend the images by crude averaging, or by an overlay technique commonly available within video analytics toolkits, but is to form a composite image based on selecting the best features from both images on a per pixel basis. If one of the modalities has no discernible image features in a particular region of the field-of-view then this image information should not be used, as it would degrade the composite image. If features are present in both modalities, then this information should be combined in an optimum sense.
An additional advantage of image fusion is that the key image features of both modalities are retained, whereas poor quality and replicated information is discarded. Consequently, the processing bandwidth, storage requirements and any downlink requirements can be reduced due to the potential halving of the retained image data (with the separate images still available if required).
As explained, Image fusion ensures that maximum available image information is presented to the operator in a single video stream. However, this powerful capability is largely wasted if the viewing platform is subject to random movement making the video unwatchable. For example, this can occur when performing surveillance from a UAV, moving vehicle, or from an installation subject to wind-induced movement.
The purpose of image stabilisation is to remove the random movement, while ensuring that any ''desired' movement captured by the sensor is retained in the scene. Stabilisation also improves the efficiency of standard compression algorithms significantly, and furthermore, monitoring the stabilised video, rather than the raw input stream, will considerably reduce operator fatigue and is likely to improve the detection rate of events of interest.
The HALO image stabiliser is highly sophisticated; it processes and extracts all useful image information and therefore works in the most challenging of circumstances, unlike the computationally simpler approach of feature tracking. This approach inevitably fails when a scene has poorly defined features or experiences extreme movements. HALO has been stress-tested using imagery that represents the most extreme operating conditions and is robust in the presence of frame-to-frame translations as large as 40 pixels (in both x and y) and frame-to-frame rotations of up to 5˚.
Modular solution can be extended with new features
HALO is highly configurable to satisfy the diverse range of customer requirements whilst maximising size weight and power efficiency. This modular design approach also minimises market entry costs, while facilitating future upgrades with the addition of new or enhanced firmware or software blocks. All functional IP blocks have been rigorously tested and therefore technical risk is reduced and down-time for in-service upgrades is minimised. HALO's integrated software processing capability provides the flexibility for implementing a host of other video post-processing functions, such as automated target tracking, feature identification, or change detection.
In addition to the HALO rugged enclosure product, RFEL offers HALO as board-only OEM variant and as a system-on-module variant to support highly integrated system developers.
RFEL is taking orders now; HALO units will be shipping from autumn 2013.