Next Generation CUDA/GPGPU Systems Is this email not displaying correctly? View it in your browser.

NVIDIA and Second Generation i7 Systems combine for COTS: GPGPU Processing, Video Tracking and Detection, DVR Next generation systems with a focus on the provision of electro-optical (EO) system management hardware, software and integration services for use in harsh environments. Our rugged GRIP PCs & DVRs utilize the latest Intel mobile processors, and our PCs can include powerful NVIDIA GPUs enabling high performance GPGPU processing. Complementing our rugged hardware is GRIP-VMS, a modular video management software suite. GRIP-VMS features DART an embedded software video tracker that is able to offer robust video detection and tracking of multiple objects. Grip Epsilion GPGPU The GRIP Epsilon is a true commercial off the shelf (COTS) high performance rugged computer system which is targeted at applications that require GPU/GPGPU processing in harsh environments with an extended operating temperature range.  The GRIP Epsilon combines the latest 2nd generation Intel Core i7 (Sandy Bridge) Mobile Processors with a range of mobile GPUs, all the way up to the CUDA enabled NVIDIA GTX580M.  The GRIP chassis provides an IP67 sealed enclosure for the internal COTS hardware technology. Within the chassis a combination of convection and conduction cooling ensures minimal heat stress of the components. If required, finning can be added to the enclosure to increase the level of external convection cooling.  A wide temperature qualified microcontroller is used internally for Health and Usage Monitoring (HUMS) and general supervisory control for hot/cold start conditions and other programmed startup conditions.  Application include security, automotive, transportation, oil and gas, nuclear, military and aerospace sectors. Data Capture A single conduction cooled PMC or XMC card can be integrated into the GRIP with I/O provided via a custom connector(s) on the GRIP connector panel. Typically this is used to support video interfaces such as: PAL/NTSC Analog Video (Up to 4 inputs) HD/SDI (SMPTE 292) up to 1080p resolution RGB up to WUXGA resolution CameraLink, Firewire, GigE Vision, 10GbE (FO) and others available GRIP DVR The GRIP DVR is a rugged Digital Video Recorder based on PC components. The use of high end COTS processing elements provides a very powerful and extremely flexible recording solution. The GRIP DVR is based around the GRIP Alpha hardware solution and can include a wide variety of COTS and custom video acquisition cards to record video from your sensor of choice. Digital video is recorded on either high capacity Hard Disk Drives (HDD) or alternatively Solid State Disk (SSD) technology can be used for increased robustness and to allow operation in unpressurised airborne environments. The GRIP DVR application software is built around the ‘GRIPworkx’ capture, processing and display library. The application provides a simple yet flexible graphical user interface which can be accessed through monitor and mouse ports, or through the Ethernet connections. The application can also be controlled through chassis mounted switches, which instigate programmable functions such as start and stop of the recording function. Chassis mounted LEDs also provide basic status information allowing the unit to be operated without the need for a monitor, keyboard or network connection. The GRIP DVR takes advantage of the GRIPworkx optimised GPGPU processing functions to offer storage of H.264 compressed video. The use of the GPGPU allows high fidelity compression and storage of multiple high resolution and/or high frame rate sensors. Key features Large solid state disk storage: Rugged chassis for harsh environments: Flexible video I/O Low power DART Video Tracking The Vision4ce DART (Detection, Acquisition, with Robust Tracking) target tracking product is an embedded software based tracker that can use either a Linux or Windows based system for video tracking and image processing applications. A video tracker analyzes video image sequences from a sensor system (one or more cameras), mounted on a servo controlled pedestal (platform) to keep the camera pointing at the nominated person or object. In this context a Tracker has two primary processing functions. Detecting and locating objects of interest in the video image (object location). Controlling the platform (Pan and Tilt) position and rate such that the camera follows the designated object (Pan and Tilt Control). DART is able to provide the user with these functions as components within the GRIP-VMS modular video management system or as a stand-alone application. The application is able to run on a commercial desktop or laptop PC, the improved speed and commercial availability of multicore CPUs make an off-the shelf, embedded software tracking platform increasingly powerful and achievable. For OEM customers we can also offer custom solutions tailored for specific or non-standard hardware platforms. Existing hardware based tracking systems are in general highly proprietary and expensive to adapt to meet a particular tracking requirement. In contrast Vision4ce's DART software based tracker allows for rapid implementation of new and different tracking algorithms offering advanced functionality and low research and development costs. The software's architecture can also take advantage of commercially available peripherals (e.g., touch screen monitors, joystick controllers, image capture cards, etc.). Key Features Multiple object detection and tracking Centroid & edge measurement Feature based correlation algorithm Moving object detection Adaptive background removal Automatic coast Grey level invariant algorithms Robust clutter rejection GRIP Hardware DART is fully compatible with the Vision4ce GRIP range of rugged PCs, enabling a complete, low cost, detection and tracking system to be implemented using COTS hardware. GRIP VMS GRIP-VMS is a comprehensive video management software tool enabling the complete control of electro-optical (EO) systems; applications include visual situational awareness, naval gun fire control and test range TSPI. Minimum operator workload: Maximum data quality GRIP-VMS provides a user interface and control methodology that minimises the operator involvement whilst maximising the quality of the information available. The Graphical User Interface (GUI) can be used to display multiple video streams in a flexible layout that can be customized to meet operational requirements. The display features on-screen presentation of ancillary data and semi-transparent video overlays for important user interface items. An operator can interact with GRIP-VMS and hence the EO system using a combination of touch screen, joystick and keyboard control. The system provides the operator with control and feedback of all aspects of the EO system, including Servo platform, position, velocity and operation mode Individual video sensor zoom, focus, offset and gain GRIP-VMS image processing functions Control of the display configuration Cost-effective custom configured solutions A GRIP-VMS application is comprised of a number of standard software modules that can be custom configured to meet end user requirements. This modular approach enables the rapid and cost-effective deployment of sophisticated video management solutions with customers only needing to purchase those modules required for their particular application. VMS Key Processing Options DART: Software Video Tracking Multiple object moving target detection amd tracking with recognition based acqusition modes. Highly robust single target tracking modes for edge, centroid, and correlation tracking. Effective combination modes using both detection and tracking algorithms operating in parallel to suit complex acqusition situations and cluttered background scenarios Panoramic Display:  High resolution 360⁰ or sector persistent situational awareness from panning sensors or multiple TV/IR cameras. Real-time stitching with blended overlap regions performed on live video feeds. Multiple object identification when integrated with DART Digital Video Recording (DVR): Recording and playback of multiple video channels SD & HD video supported. Parallel recording of sensor metadata. Video Streaming: Customized SD/HD video and metadata streams. Standards include DVI over Ethernet, Def Stan 00-82 and Stanag 4609.  Electronic Image Stabilization (eStab):  Compensates for sensor and platform movements using a combination of image processing algorithms and data from roll sensor inputs. Image Enhancement: Pixel level fusion of multiple video streams, including GPU accelerated rescaling and alignment algorithms. Contrast enhancement by high pass filtering and local contrast stretching GRIP Workx GRIPWorkx is Vision4ce video capture, processing and display software development kit. The SDK provides a framework for the development of real-time video acquisition, GPGPU processing and display software applications on Vision4ce GRIP hardware. Real-time video processing using the GPU. Video data is streamed from a frame grabber into main memory using a DMA process and an interrupt is generated when a complete frame is available for processing. The frame is then transferred to the GPU into CUDA memory space for processing with a CUDA kernel. After processing the data can be transferred back to the host or marshalled into a compatible memory space in OpenGL texture memory. This OpenGL texture memory can then be displayed on a monitor. GRIPWorkx simplifies all of these allocation and marshalling tasks, accelerating the development process. The main component of the GRIPWorkx SDK is the CoreLib ‘C’ API, which provides functions for image capture and display. The CoreLib API also links to the NVidia CUDA driver API and provides functions for the transfer of image data to and from CUDA enabled GPU hardware for processing. The CoreLib API also provides for rapid image transfer from CUDA device memory to OpenGL display buffers for visualization.The GRIPWorkx SDK includes source code for a number of sample applications. The AcqProcDemo sample application shows how to integrate the CoreLib API with CUDA runtime kernels for development of GPU processing algorithms. CoreLib Image Sources The Vision4ce GRIP hardware can encapsulate different types of video capture devices, each of which has a different vendor-specific API. In order to unify the capture process across all GRIP hardware configurations and protect end-users from changes to capture hardware, GRIPWorkx implements a software abstraction layer based on the concept of generic image sources Sample Application 'AcqDemo' Screenshot ReImage sources are loaded and controlled by the CoreLib. At run-time, image sources produce GRIPWorkx compatible images at the capture rate and notify the user application via a capture callback.  The Image sources themselves are libraries that wrap vendor specific APIs and are loaded under user control by the GRIPWorkx CoreLib. With GRIPWorkx the developer needs to learn only one capture API and is free to decide from application to application which platform and capture hardware is required. Vision4ce GRIP hardware supports a wide variety of acquisition hardware from different OEM vendors, but with GRIPWorkx, hardware from different vendors can be interchanged, effectively isolating the end user from future hardware changes.

For more information CLICK HERE