A drone can look perfectly still on a phone screen while its flight controller is doing a steady stream of small corrections. Satellite positioning is one way to estimate where the aircraft is. Close to the ground, some drones also use cameras and other sensors to measure how the surface appears to move beneath them. One term for that process is optical flow.

The phrase sounds more complicated than the idea. A downward-facing camera captures a sequence of images. The system tracks how visible features shift from one frame to the next and uses that apparent movement to estimate the aircraft’s motion relative to the surface. It is useful, but it is not a tiny GPS receiver and it does not make every indoor flight easy.

What optical flow actually measures

Optical flow is about apparent motion in an image. If a drone moves sideways over a floor with visible texture, points in the camera image move in the opposite direction. By comparing successive frames, software can estimate how the aircraft is moving relative to the scene.

That estimate is local. It can help a flight controller reduce drift or hold a position over a suitable surface, especially when satellite signals are weak or unavailable. It does not, by itself, tell the aircraft its global latitude and longitude. Nor does it provide a permanent map of the room.

Manufacturers may combine optical flow with an inertial measurement unit, a distance sensor or other visual-positioning methods. The exact combination matters. Parrot’s ANAFI Ai documentation, for example, describes its vertical camera using optical flow for velocity estimation and keypoint detection and matching for position estimation. Those are related visual methods, but they are not identical measurements.

Why the surface below the drone matters

A camera needs something it can recognise. A patterned floor, concrete with visible detail or another textured surface gives the system features to follow. A plain, glossy or repetitive surface may provide too little useful information. A change in lighting can also make features harder to track.

Parrot’s ANAFI UKR white paper explains that optical-flow tracking depends on visible details in the scene; when the scene lacks texture, changes between frames become harder to measure. That is a general explanation of the algorithm, not a guarantee about every aircraft or sensor.

This is why a drone that holds position well over a textured floor may behave differently over a blank white surface, dark carpet or reflective tile. The sensor has not suddenly become a different device; the scene has become harder for its software to interpret.

The operating limits are model-specific. DJI’s Avata 2 specifications describe downward and backward visual positioning and a downward time-of-flight sensor. They also specify an operating environment for the visual system: discernible surface patterns, diffuse reflectivity above 20 percent and adequate illumination above 15 lux. Those figures apply to that product’s stated system, not to every drone with a downward camera.

Optical flow is not the same as obstacle avoidance

The names can be confusing because the same aircraft may have several cameras and sensors. A downward camera used for positioning is not automatically an obstacle-detection camera. Its job may be to estimate movement relative to the ground, not to identify a wall or stop the aircraft before it reaches one.

DJI’s developer documentation describes downward-facing cameras as a source of relative position and velocity information, including in environments where GNSS is unavailable. It discusses obstacle sensing separately. That distinction is important: a drone may use visual positioning and still have limited obstacle sensing, or no obstacle avoidance in a particular direction.

Do not assume that a positioning system will protect the aircraft from furniture, people, cables or a low ceiling. The pilot still has to maintain visual awareness and leave enough room to manoeuvre.

What happens when GPS is weak?

Optical flow can help with relative movement when satellite positioning is unavailable, but it is not a universal indoor-navigation solution. The aircraft still needs suitable visual information and must remain within the operating limits stated by its manufacturer. Other sensors may contribute to altitude and attitude estimates, and the flight controller has to combine those inputs.

Parrot’s ANAFI Ai documentation describes its indoor positioning approach as a combination of vertical-camera measurements: optical flow for velocity estimation and keypoint detection and matching for position estimation. It also describes LED illumination beside the camera to support operation in low light, with the stated indoor positioning use below five metres above the ground. That is a product-specific design, not a general promise that any drone can hold position in a dark room.

If a positioning warning appears, the sensible response is not to assume that the drone will compensate indefinitely. Follow the aircraft’s manual, move to a suitable environment if necessary, and land if stable control cannot be maintained.

What to check before buying

A product listing that says “optical flow” is only a starting point. Before relying on the feature, check the exact model’s manual and technical specifications:

• Sensor type: Is the system described as optical flow, visual positioning, downward vision, time-of-flight, or a combination?
• Working range: At what height and under what conditions does the manufacturer say it can assist with hovering?
• Surface and light requirements: Does the manual specify texture, reflectivity or minimum illumination?
• Other sensors: Does the aircraft combine camera measurements with an IMU, infrared or time-of-flight sensor?
• Failure behaviour: What warning appears if the positioning system cannot obtain a reliable estimate, and what does the manufacturer tell the pilot to do?

A downward camera is a useful part of a positioning system, not a guarantee. The important question is not whether a drone has a sensor on its underside, but what that sensor is designed to estimate and the conditions in which the manufacturer says it can do so.

Sources

• Parrot, ANAFI Ai technical documentation — indoor flight and vertical-camera measurements
• Parrot, ANAFI UKR white paper — optical-flow feature tracking and scene texture
• DJI, Avata 2 specifications — visual-positioning sensors and operating environment
• DJI Developer, Flight Controller — vision-system and positioning assistance

Editorial disclosure: ProDrone operates a drone retail business. This article explains a technical term using manufacturer documentation; it is not a hands-on test, a comparative product review or a claim that any specific model is suitable for every indoor environment.