About NRV
NRV develops event-based vision technology for machines that need to follow fast motion and changing scenes. Our goal is to help engineers capture rapid changes and use that information in applications such as industrial inspection, robotics, and motion analysis.
We bring together sensing technology, camera hardware, and software tools to help customers move from exploring event-based vision to building with it. Our cameras, software development kit (SDK), and documentation support that process, from the first data capture to application development.
Why event-based vision?
For a machine inspecting a moving part or tracking a fast object, what happens between images can matter. Conventional cameras record complete images at set intervals. Capturing more images can reveal finer motion, but also increases the amount of data to transfer and process.
A Dynamic Vision Sensor (DVS) takes a different approach: each pixel detects changes in brightness. These changes are called events, and they indicate where the scene has become brighter or darker. This allows a vision system to focus on changes in the scene, making DVS useful for observing fast motion without repeatedly recording an entire unchanged background.
NRV's approach
NRV combines three complementary capabilities: frame-based event capture, compact sensor-side encoding, and synchronized measurement across all pixels.
1. Event frames: less timing overhead, ready for image processing
An individual event is usually described by its position, timestamp, and polarity. Event-processing software often stores that timestamp as a 64-bit value for every event. NRV instead captures pixel event states as event frames with shared frame-level timing, so the sensor stream does not need to attach a separate timestamp to every pixel event.
Removing repeated timing information makes the data lighter and leaves more of the available bandwidth for actual events. This supports high-speed capture of dense event activity, especially when many pixels change together.
Event frames are also easier to use. Once decoded, they can be displayed and processed as images with familiar image-processing tools, without first assembling individually timestamped pixel events into a time window. The result is an image of brightness changes, rather than a conventional intensity photograph.
2. Group AER: up to 16 pixel events in one 32-bit packet
The sensor outputs events directly in Group AER, a compact RAW format that can carry up to 16 pixel events in one 32-bit packet. Grouping events reduces repeated information and makes event data more efficient to transfer.
3. Global Reset and Global Hold: synchronized pixel measurements
Global Reset updates each pixel's brightness reference simultaneously across the sensor. All pixels therefore begin measuring brightness changes from the same reset instant, each relative to its own updated reference.
Global Hold latches the measured event states of all pixels at the same instant. The sensor preserves those results during sequential readout, rather than letting each pixel's readout time determine the end of its measurement interval.
Together, these sensor-internal functions synchronize the brightness-reference update and the brightness-change measurement interval across all pixels. This makes brief flashes and fast motion comparable across the entire image, without different parts of the sensor representing different measurement intervals.
Start building with NRV
Explore our cameras to find a starting point for your application, then use the SDK and guides to connect a device and work with event data.
To discuss camera selection or your application requirements, contact us at contact@nrv.kr