What is an event-based vision camera?
Just as the human eye perceives light…
It is believed that the human brain "selects" information from vision and processes it efficiently.
| Playing with dogs in a large field | → | Dogs |
| Fishing | → | Rod tip or float |
| Lifesavers | → | Flow of people and abnormal movements on the sea |
| Driving a car | → | Cars or pedestrians in front of you |
| Inspection work | → | Product operation or appearance |
etc.
These are examples of how the brain's processing ability is thought to be streamlined by perceiving changes in light (objects) captured by vision as vivid at the point of interest and ambiguous around the point of interest.
An event-based vision camera is characterized by its high-speed output of information on changes in light (objects) in single-pixel units. This feature is similar to passing visual information (image data) to the brain (CPU) at high speed.
Conventional frame-based cameras output frame rate data (e.g., 30 images per second at 30 fps), so the image processing load performed by the CPU is enormous.
In contrast, event-based vision camera is expected to reduce CPU load because the output data itself is more efficient. In addition, it has the feature of high-speed detection because it transmits light change information at the microsecond level for each individual pixel.
Mechanism of event-based vision sensor
The event-based vision sensor detects the luminance changes in each pixel and extracts only those that exceed the preset threshold value then outputs the event "coordinate, polarity, and time". This operation is performed independently and asynchronously for each pixel.
Light incident on each pixel is converted into a voltage, the delta from the reference voltage is detected, and an event is output when the threshold value is exceeded in a comparator.
Differences from conventional camera
While conventional cameras output information for the entire screen according to a fixed frame rate, event-based vision cameras output asynchronously data only for pixels that have undergone a change in brightness or darkness. Thus, high temporal resolution (in 1 µsec increments), low data rate, and low power consumption are achieved simultaneously.
Conventional camera ▽
Event-based vision camera ▽
It outputs data of all pixels in every frame.
Areas with no luminance change (background or parked cars) are not be output.
The features of the event-based vision sensor are also explained on the Sony Semiconductor Solutions Corporation (SSS) website (external site).
Use cases
Industry
Motion detection, piece counting, vibration detection, weld monitoring, measurement, etc.
Security
Intruder detection, low-light area monitor, flow line monitor, etc.
Robotics
Obstacle detection, abnormality detection, data for AI, object measurement, automatic positioning, etc.
Chemistry, Research
Fluid Monitoring, Microscopic Detection, Research and Development, New Technology Development, etc.
Sample Video
Material Tracking
The length and color of the arrows indicate the speed and direction of movement of the small balls in the box. The background is visible in the normal camera image, but not in the event-based vision camera. This is one of the features of this camera.
Wide dynamic range
Even in intense light that would cause halation with a conventional camera, an event-based vision camera can reliably capture the subject.
Slow Motion
With an event-based vision camera that outputs per-pixel 'coordinate, polarity, and time' data asynchronously, you are free to generate video with a different frame rate after capture.
Click the right button for other videos.
Software toolkit
All SilkyEvCam products can use the "Metavision® SDK", a software toolkit provided by PROPHESEE. It supports everything from first-time image output and prototype evaluation to full-scale application development.
The "Metavision® SDK" provides basic libraries, algorithms, and practical sample programs for handling with event-based vision camera data.
For more information, please visit PROPHESEE website (external site).
CenturyArks event-based vision cameras
- Each pixel outputs data of motion asynchronously, independently from any other.
- Highly efficient data (1/10 to 1/1000 of data volume compared to frame-based camera)
- Wide dynamic range
- Frame rate can be changed after recording
- No motion blur
- Compact size can be installed anywhere (30mm x 30mm x 36mm)
- Free plug-in for PROPHESEE's SW development environment and algorithm library, METAVISION® SDK
For Inquiry
Inquiry by phone
+81-03-6804-5752
JST 10:00-17:00
(except Saturdays, Sundays, national holidays, and other company holidays)
Inquiry by e-mail
- SilkyEvCam® is a registered trademark of CenturyArks Co., Ltd
- USB Type-C™ is a registered trademark of USB Implementers Forum.
- The names of companies and products mentioned in this document are registered trademarks or trademarks of their respective companies or organizations.
- The specifications and other information in this document are subject to change without notice.