A thin cable can be obvious to a person and almost disappear to a drone. A window can look like open space. Still water can return the wrong visual cues. At dusk, a feature that appeared dependable at noon may be limited or unavailable.

That is not a contradiction in the idea of obstacle sensing. It is the consequence of how these systems work. A camera, infrared sensor or rangefinder does not understand an “obstacle” in the human sense. It measures light, texture, reflection, distance or motion within a defined field of view, then software decides what those measurements mean.

The practical lesson is simple: obstacle sensing is a useful layer, but it is not permission to fly closer to hazards than you could safely manage without it.

The feature name hides the conditions

“Obstacle avoidance” sounds like an outcome. A manual usually describes something narrower: particular sensors, directions, speed ranges and environmental conditions in which detection or braking may work. Those details vary by aircraft and flight mode, so a claim about one model should not be turned into a rule about every drone.

DJI’s official introduction to aircraft obstacle-avoidance systems says vision sensing is intended for well-lit environments and that performance depends heavily on the reflective properties of the object being observed. It identifies dim conditions and smooth surfaces such as mirrors or tiles as difficult cases.

A detailed example appears in DJI’s Matrice 30 Series Dock Bundle manual. For that system, the manufacturer lists wires, cables, branches, mirrored or transparent surfaces, water, ice, low-visibility weather, dirty lenses, rapidly changing illumination and surfaces without clear texture among the situations that can reduce or defeat sensing. This is evidence about a named product family, not a universal specification for every DJI aircraft or every sensor design.

Skydio publishes a similar warning for a different platform. Its current Safety and Operating Guide says the X10 may not see thin branches, lines, ropes, netting, wires and chain-link fencing, and that obstacle avoidance can be degraded by transparent or reflective surfaces. It also warns that the system is intended to avoid stationary objects, not every moving vehicle, person, animal or aircraft.

The overlap between two manufacturers is useful because it identifies recurring categories of difficulty. It still does not prove that the two systems behave identically, or that another model has the same limits.

Why thin objects are a hard target

A power line or bare branch may occupy only a tiny part of the sensor image. Its apparent thickness shrinks with distance, and its contrast can collapse against trees, cloud, buildings or bright sky. If the aircraft or background is moving, the available observation time can become shorter still.

That does not mean every wire will be missed. It means a pilot cannot safely assume that a line will be detected early enough for the aircraft to stop or steer around it. The important variables include the sensor’s direction, resolution and field of view; the object’s size and contrast; aircraft speed; range; lighting; and the avoidance behaviour selected for that mode.

Operational inference: where wires, netting, small branches or guy lines may cross the route, route design and physical separation are more dependable controls than waiting for an alert.

Glass and water can present the wrong scene

Transparent and reflective surfaces cause a different problem. A window may reveal the textured scene behind it, reflect the sky or combine both. Still water can act like a mirror. Shiny metal and polished surfaces can change dramatically with viewing angle.

To the sensing system, the measured scene may not correspond neatly to the solid boundary the aircraft must avoid. Manufacturer guidance therefore treats windows, mirrors and water as specific caution areas rather than ordinary walls or ground.

The consequence reaches beyond collision avoidance. Some aircraft also use downward-facing vision to help hold position. A visually ambiguous surface beneath the drone can affect positioning as well as forward obstacle detection. The exact behaviour is model-specific; pilots should check the manual for the aircraft they are actually flying.

Bad light is more than “too dark”

Low light reduces the detail available to vision systems, but darkness is not the only difficult condition. Direct glare, a low sun, strong reflections, rapid movement from shade into bright light and featureless surfaces can also reduce useful contrast or temporarily overwhelm a camera.

DJI’s Matrice 30 documentation lists both extremely dark and extremely bright surfaces, frequent lighting changes and strong reflected light among its limitations. Skydio warns X10 pilots that a low sun can temporarily blind navigation cameras at some flight angles. These are manufacturer warnings, not independent comparative tests.

Practical implication: a short site check at midday does not fully describe the same route near sunset. If lighting is part of the safety case, assess it for the planned time and direction of flight.

Moving hazards do not wait for the algorithm

Avoidance systems may treat stationary geometry better than objects whose position changes unpredictably. Skydio explicitly says its obstacle avoidance applies to stationary objects and cautions against relying on it around moving aircraft, vehicles, people and animals.

This is where a reassuring feature label can create the wrong habit. The system may help with a tree and still be the wrong control for a cyclist entering the route, a bird crossing the frame or a crewed aircraft appearing nearby. Separation, visual scanning and an immediate escape or landing decision remain human tasks.

The pilot’s responsibility does not transfer to the sensor

For UK open-category flying, the CAA Drone Code says the remote pilot is responsible for flying safely, must keep the aircraft in direct sight and must maintain a full view of the surrounding airspace so hazards can be spotted and collisions avoided.

The underlying UAS.OPEN.060 responsibilities also require the pilot to observe the operating environment, check for obstacles, maintain a visual scan during flight and operate the aircraft in accordance with the manufacturer’s instructions and limitations.

Those UK requirements do not turn a product manual into law for every country. They show why, in this jurisdiction, a fitted avoidance system does not replace visual line of sight, environmental checks or pilot control.

A five-part plan for routes with difficult obstacles

1. Read the model-specific manual. Confirm which directions are sensed, which flight modes change the feature, and what the manufacturer says about light, speed, surfaces and small objects.
2. Walk or observe the route before launch. Look specifically for wires, netting, bare branches, glass, water, repetitive or blank surfaces, overhangs and moving traffic. A hazard that is hard to see from the launch point may be harder to judge once airborne.
3. Create clearance, not a detection test. Plan a route that stays well away from uncertain objects. Do not fly toward a cable to discover whether the aircraft will notice it.
4. Reduce speed and complexity. Extra distance and time can improve the pilot’s ability to recognise a bad scene and stop. The appropriate limits remain aircraft- and site-specific.
5. Set an abort condition. If light deteriorates, glare blocks the view, a sensor warning appears, people or vehicles enter the area, or the route becomes visually ambiguous, move to clear space and land or re-plan.

What to ask before buying or deploying the feature

A useful specification is not simply “360-degree obstacle avoidance”. Ask which sensors provide that coverage, whether it is continuous in the intended mode, what gaps remain above or below the aircraft, what speed and light limits apply, whether braking distance is published, and how the system behaves when a sensor becomes unavailable.

For professional work, add the site conditions that matter to the mission. Inspection near lattice structures, cables or reflective cladding presents a different problem from slow flight in an open field. An impressive demonstration in a textured, well-lit environment may say little about the route you need to fly.

Treat sensing as a layer, not a shield

Obstacle sensing can prevent mistakes and reduce workload. The safest interpretation is neither to dismiss it nor to trust it without conditions.

The feature is one layer in a larger system: an appropriate route, adequate clearance, direct observation, conservative speed, current manufacturer guidance and a pilot ready to take control. Thin wires, glass, water and bad light matter because they expose the gap between recognising a scene and merely measuring it.

Scope note: This article compares publicly available manufacturer safety guidance with current UK pilot responsibilities, rechecked on 13 October 2026. It does not report independent flight testing and does not claim that all aircraft or obstacle-sensing technologies share identical limitations.