Event-Based Vision: Bringing More Performance and Efficiency to Improve Machine Vision Applications

01/02/2024 - 12 PM - 1 PM ET

ACERCA DE ESTE WEBINAR

The growing complexity and challenging operating conditions in industrial machine vision require innovative approaches to capturing and processing the necessary visual information that achieves the objectives of the system. These systems include high-speed inspection cameras that must deliver fast motion capture, robotic grinding or welding systems that need high dynamic scene capture, and smart Edge IoT cameras that must have both speed and accuracy as well as robust object tracking capabilities.

But for the most part, such systems have relied on an increasingly inadequate and incomplete vision method: frame-based vision. This method struggles to address many important challenges, such as capturing data in a high-speed, continuous fashion in dynamic scenes; working effectively in challenging lighting conditions; and functioning in operating environments where compute and power resources are limited.

Most important is a lack of completeness: Frame-based vision was never meant to address the challenges introduced by today’s industrial use cases. In frame-based image capture, an entire image (i.e. the light intensity at each pixel) is recorded at a pre-assigned interval, known as the frame rate. While this works well in representing the ‘real world’ when displayed on a screen, recording the entire image at every time increment oversamples all the parts of the image that have not changed.

Event-Based Vision introduces a new approach that, like our eyes and brains, uses independent receptors collecting all the essential information, and nothing else. With 10-1000x less data generated, >120dB dynamic range and microsecond time resolution (over 10k images per second equivalent), event-based vision opens vast new potential in areas such as industrial automation, robotics, security and surveillance, mobile, IoT and AR/VR.

Over the past several years, event cameras that leverage neuromorphic techniques have gained a strong foothold in machine vision applications in industrial automation, robotics, automotive and other areas. These are all applications where better performance in dynamic scenes, capturing fast-moving subjects and operating in low-light conditions are critical. In addition to benefits in terms of power consumption, data efficiency and dynamic range, event-based vision addresses a fundamental limitation of traditional camera techniques: how light is captured.

Among the advantages of event cameras are:

• Blur-free images as there is no exposure time
• High-speed data capture: 10K fps resolution equivalent
• High resolution of up to 1280x720px
• High dynamic range (>120dB)
• Shutter-free operation: no global or rolling shutter needed

Uses case discussed include:

• Object tracking that leverages the low data rate and sparse information provided by event-based sensors to track objects with low compute power. Well suited for pick and place, robot guidance and trajectory monitoring;
• Fluid monitoring that uses event-based optical flows to perform fluid monitoring in real time and analyze unwanted dynamics due to residue build-up, spot contaminants or unwanted air or gas bubbles. Well-suited for continuous liquid flow monitoring in food and beverage, oil and gas, and biological processes;
• Vibration monitoring that enables the monitoring of vibration frequencies continuously, remotely, with pixel precision by tracking the temporal evolution of every pixel in a scene. For each event, the pixel coordinates, the polarity of the change and the exact timestamp are recorded, thus providing a global, continuous understanding of vibration patterns. Used for predictive maintenance tasks such as motion monitoring, vibration monitoring and frequency analysis;
• Particle and object sizes monitoring: Event cameras can better control, count and measure the size of objects moving at very high speed in a channel or a conveyor. Implemented in systems for high-speed counting, batch homogeneity and gauging.
• Viewers will be able to learn about development tools, algorithms and open-source resources that can accelerate the understanding, experimentation and implementation of embedded event-based vision capabilities in machine vision systems
• Prophesee is the inventor of the world’s most advanced neuromorphic vision systems. Composed of patented Metavision® sensors and algorithms, its event-based vision technology enables machines to process visual information more efficiently, thoroughly and in challenging conditions

Key Takeaways:

• Introduction to the concept of event-based vision and how it compares to traditional frame-based methods
• Understanding of key benefits of event-based vision in terms of power, performance dynamic range, integration with other data acquisition technologies
• Overview of development tools and methods that can be used to integrate event-based vision into MV systems
• Examples of common use cases that leverage the advantages of event-based vision in industrial applications