Sign in to save

Bookmark this page so you can find it later.

Sign in to save

Bookmark this page so you can find it later.

An event camera is a vision sensor designed for robots that need to react to fast changes instead of waiting for full video frames. Each pixel works independently and reports only when it detects a change in brightness. This makes the camera very fast, efficient, and useful for drones, robotic arms, self-driving systems, and other machines that must perceive motion quickly.

Event cameras matter because many robotic tasks fail when normal cameras blur, saturate, or respond too slowly.

Key Facts

  • An event is often written as e = (x, y, t, p), where x and y are pixel location, t is time, and p is polarity.
  • A positive event means brightness increased at a pixel, and a negative event means brightness decreased.
  • Event cameras can reach microsecond-scale latency, often around 1 us to 100 us depending on the sensor and system.
  • Brightness change is commonly detected using log intensity: ΔL = Δlog(I).
  • An event is triggered when |ΔL| >= C, where C is the contrast threshold.
  • High dynamic range allows event cameras to work in scenes with both very bright and very dark regions, often exceeding 120 dB.

Vocabulary

Event camera
A camera that outputs changes in brightness at individual pixels instead of recording full image frames at fixed times.
Polarity
The sign of an event that tells whether brightness increased or decreased at a pixel.
Latency
The time delay between a change in the scene and the camera or robot responding to it.
Dynamic range
The range of light intensities a sensor can handle without losing detail in dark or bright areas.
Neuromorphic sensor
A sensor inspired by biological vision that processes changes asynchronously rather than capturing fixed frames.

Common Mistakes to Avoid

  • Thinking an event camera records normal video frames, which is wrong because it outputs sparse brightness-change events instead of full images.
  • Ignoring event polarity, which is wrong because positive and negative events carry different information about how brightness is changing.
  • Treating microsecond latency as the same as frame rate, which is wrong because event timing is asynchronous and not limited to evenly spaced frames.
  • Assuming event cameras see stationary objects clearly by default, which is wrong because objects with no brightness change may produce few or no events.

Practice Questions

  1. 1 A pixel has log intensity L = 2.10, then changes to L = 2.18. If the contrast threshold is C = 0.05, will an event be triggered, and what is its polarity?
  2. 2 A robot moves at 5.0 m/s. A standard camera has 20 ms latency, while an event camera has 100 us latency. How far does the robot travel during each latency?
  3. 3 Explain why an event camera is useful for tracking a fast spinning drone propeller in a scene with both bright sunlight and dark shadows.