LiDAR for orchards and perennials: geometry, canopy and the field decisions that depend on it

For an orchard vehicle, the issue is not what LiDAR can measure in a showroom but whether its geometry holds up in a real canopy row. LiDAR builds a three-dimensional point cloud of the scene around the sensor, and that point cloud is the raw material for obstacle detection, row following and canopy measurement. An operator who trusts that cloud in dust, rain and bright sunlight without testing it is making a field decision on unverified hardware, because no failure-rate figure appears in the vendor materials or independent sources cited here.

A point cloud is a set of coordinates, not a list of objects. Segmentation groups points that belong to the same surface or object; only then can the system classify what it sees and tell a tree trunk from a person or a row from an open lane. Dust, rain and bright sunlight produce spurious returns, which makes segmentation and classification harder. Asaptic, an engineering firm working on LiDAR in harsh environments, describes rain producing backscatter that the sensor treats as range returns, creating a dense cloud of false positive points in the two-to-fifteen-metre band while real-surface pulses attenuate (Asaptic on LiDAR degradation in harsh environments).

Range, resolution and weather

Resolution is set by the number of points per second and the angular spacing between them. Detection range trades against resolution and eye-safety limits. Heavy dust or fog can scatter the beam and shorten effective range, and a reflective surface can saturate the receiver. Even published accuracy figures carry conditions: DJI specifies the Zenmuse L2 at four-centimetre vertical and five-centimetre horizontal accuracy under stated operating conditions and a defined calibration workflow (DJI Zenmuse L2 specifications). The figure is not a universal number. Because no single sensor works in every condition, vehicle designs often fuse LiDAR with cameras and inertial measurement rather than rely on one modality.

Trust and failure modes

The common failure in autonomy is not a missing point but a confident, wrong interpretation of the points that are there. A false positive treats a soft bush as an obstacle and stops the vehicle: cautious but unproductive. A false negative classifies a person as vegetation and continues: dangerous. The behaviour an operator must test is how the vehicle degrades when the cloud is corrupted, whether it slows, stops or hands control back to a human, because that behaviour decides whether a degraded sensor is an inconvenience or a hazard.

The spray case that pays for the sensor

The published case for canopy LiDAR rests on variable-rate spraying, not on autonomy. A 2025 peer-reviewed review in Sensors reports pesticide-use reductions of 15 to 40 percent from variable-rate orchard spraying, with typical return on investment within one to three growing seasons (Sensors 2025 review of canopy perception for variable-rate spraying). The review's payback range assumes the cost of the sensing system is recovered against the chemistry saved, so an orchard operator should cost the sensor against its own spray bill before believing any marketing payback figure. That separation matters: the spray-reduction case has published evidence, while the safe-autonomous-operation case for the same point cloud does not.

What matters in an orchard

For orchard use, an operator should test whether row detection holds up in variable light, whether obstacle detection works at walking and working speeds, and how the system behaves in dust. LiDAR performance is site-dependent and season-dependent, so the test must run in the conditions the machine will actually work in. No South African deployment data appears in the sources cited here, and no claim is made. Whether the spray-reduction evidence starts appearing in South African orchards with local crop geometry and canopy density is the follow-up worth watching, because that is the test that will decide whether LiDAR earns its place on the machine or stays a showroom capability.