Computer Vision for Embedded Systems
SRI offers a robust portfolio of real-time video processing hardware and software intellectual property (IP) for embedded system design. Built on more than 30 years of SRI experience developing advanced computer vision and machine learning systems for government and commercial applications, SRI’s IP can be licensed and customized to fit individual requirements.
As the demand for smart systems increases, developers are challenged to incorporate more decision-making intelligence into every new system. Embedded systems designers can leverage SRI’s technology, domain knowledge, and decades of experience to deliver advanced, state-of-the-art computer vision and machine learning solutions.
Partnering for Success
By licensing SRI’s embedded processing IP, developers of real-time systems could decrease time to market, reduce program risk, and stretch development resources, while meeting unique customer requirements.
SRI’s Center for Vision Technologies team works directly with developers to understand and address specific needs. From image enhancement to object detection and tracking, SRI’s customized vision processing solutions maximize the practical use of video and imagery in numerous application domains.
Hardware Cores and Software Libraries
SRI technology on FPGA and smartphone processor for drone applications.
SRI has a number of technology offerings, from software libraries to hardware cores in field-programmable gate arrays (FPGAs). SRI’s embedded IP is specifically tuned for applications requiring low video processing latencies. Plus, it is optimized for the low-power performance required of portable, battery-operated systems. Our software libraries are designed for various embedded operating systems including Linux and Windows. SRI also provides software capabilities on smartphone platforms, including Android and iOS.
For example, the SRI FPGA design portfolio consists of register-transfer level (RTL) netlist implementations and companion embedded processor software code that can be integrated by SRI or the developer to provide a complete custom solution. SRI provides portable IP modules with well-defined hardware interfaces and software APIs to ease customer integration—that support scaling and extensibility throughout the product lifecycle.
SRI Vision Processing Features Mapped to Typical Embedded Vision Applications
SRI has delivered stabilization and registration IP solutions for performing geo-registration of aerial images as well as tracking of objects from large standoff distances. These embedded video processing solutions included the registration of multimodal ground imagery, mosaicking for live updating of terrain maps, and feature-based target tracking.
Application Example: Handheld Imaging Devices
SRI has worked with a number of leading U.S. suppliers of military handheld imaging devices for range finders, target locating devices, and spotter scopes. For these devices we provided vision processing hardware acceleration modules—RTL—and control software that the client integrated into their hardware/software framework. Functions which have been incorporated into handheld imaging devices include electronic stabilization, local contrast enhancement, rolling shutter correction, object tracking, and image sensor pipeline (ISP) functions such as non-uniformity correction (NUC), histogramming, Bayer filtering, and noise filtering.
In a rapidly expanding commercial UAV landscape, system manufacturers and software developers can differentiate their offerings with SRI’s processing technologies. Developed under ongoing government funding, our licensable technologies enable rapid go-to-market solutions. Below are some possible use scenarios.
Live Events – See It All
Sports – Follow the Action
Inspection – Measure Precisely