Skip to content
Blog

Visual navigation at night: low-light and thermal cameras

Published:

Visual odometry, VIO and SLAM all depend on seeing distinct features in camera images. At night, in fog or under heavy cloud, a standard camera sees too little, and those methods degrade. Low-light sensors and thermal cameras are the two main ways to keep visual navigation working when light is scarce.

Why darkness hurts visual navigation

In low light a standard camera has two options: raise the gain, which adds noise, or lengthen the exposure, which blurs moving images. Both make it harder to detect and track the same points between frames, so feature tracking becomes sparse and unreliable.

The result is more drift, or a visual system that stops contributing altogether and leaves the navigation filter to rely on the IMU.

Low-light cameras

Low-light sensors use larger pixels and more sensitive electronics to capture usable images with very little light. They keep the familiar visible-light appearance of the scene, so the same algorithms that work in daylight can often run with modest changes.

Their limit is that they still need some light: on moonless, overcast nights or in deep shadow, image quality falls off.

Thermal cameras

Thermal (long-wave infrared) cameras measure heat radiated by surfaces rather than reflected light, so they work in complete darkness and see through some smoke and haze. They bring their own challenges for navigation:

  • Thermal crossover: around dawn and dusk, different surfaces can reach similar temperatures and the image loses contrast.
  • Lower resolution and fewer sharp features than visible-light cameras of similar size.
  • Periodic recalibration of the sensor can briefly interrupt the image stream.
  • Feature detectors tuned for visible images may need adjustment for thermal imagery.

Combining sensors for all-hours navigation

Because each camera type has gaps, robust systems treat vision as one input among several. The IMU carries the solution through short periods with poor imagery, a barometer bounds altitude, and the fusion filter weights visual updates by how confident it is in them at that moment. Some systems use both a visible and a thermal camera and switch or blend between them as conditions change.

Sensor fusion and the Kalman filter in drone navigation →

Frequently asked questions

With a standard camera it degrades quickly as light falls. Low-light or thermal cameras extend it into darker conditions, and fusing vision with an IMU keeps navigation going when image quality drops.

Each has strengths. Thermal cameras work in total darkness but have lower resolution and lose contrast during thermal crossover. Low-light cameras keep visible detail but still need some ambient light. The right choice depends on the conditions the aircraft will fly in.

Thermal crossover is the period, typically around dawn and dusk, when different surfaces reach similar temperatures. The thermal image loses contrast, which makes features harder to track.