--- title: "Pest control is easy over rice. A Chinese drone maker just rebuilt its flagship for trees" date: 2026-11-09 category: AI in Action site: NeuroAI canonical: https://neuroai.site/a/na-app-eavision-orchard-spray-vision language: en --- # Pest control is easy over rice. A Chinese drone maker just rebuilt its flagship for trees > Suzhou's EAVision has fitted its 2026 flagship with 80 litres of spray and a 180-metre lidar, betting orchards are where farm AI earns its keep. A pomelo orchard in Fujian looks, from above, like a green sea with no lanes. The canopy is four metres deep, the branches interlock, and the insects that cost the grower money sit on the undersides of leaves in the middle of the tree. A drone hovering three metres above the crown sprays into the top and misses the rest. That geometry is the reason a Suzhou company has spent a decade pointing cameras and lidar at fruit trees instead of rice paddies. ## Key takeaways - EAVision (极目机器人), founded in Suzhou in 2016, has launched its 2026 line around the J160: 80 litres of spray payload, 100 litres of spreading payload, and up to 200 mu (about 13 hectares) per sortie. - The pitch is perception rather than payload — a 256-line lidar with phased-array radar above and below, 180-metre detection and a 20-millisecond emergency hover, all aimed at canopies, wires and slopes. - The company says its equipment is in use in more than 30 countries and that it added six overseas markets in 2025, with overseas revenue up more than fivefold. ## Why trees break the drone playbook Field crops are a solved-ish problem. The ground is flat, the plants are uniform, the flight line is a grid, and the spray lands where it is aimed. Rice, wheat and corn are why agricultural drones became a Chinese export category in the first place. Orchards invert every one of those assumptions. Tree height varies by several metres within a single plot. Branches stick out at angles that a downward-facing nozzle cannot reach. Terraces on hillsides turn a straight line into a contour. And the pest pressure that matters — mites, thrips, citrus canker — often sits on the underside of a leaf, which is exactly where gravity does not take a droplet. The industry's answer has been to fly lower, spray finer, and add airflow. EAVision's answer is to make the aircraft understand the shape of the tree first. ## The sensor stack is the product The J160 carries a 360-degree perception system the company calls HiDAS: a 256-line lidar at the front, phased-array radar above and below, and a processor that handles roughly 920,000 point-cloud points per second. The claimed detection range is 180 metres, and the aircraft can stop and hover within 20 milliseconds when something appears in its path. For a pilot, that translates into not clipping a power line at dusk. For the business, it translates into being able to sell into terrain where a cheaper drone is a liability — hillside orchards, tea terraces, the narrow plots of southern China where a tractor cannot go and a person with a backpack sprayer is the incumbent. The company's lineage helps explain the emphasis. EAVision started as a binocular-vision company, and its founders spent years on the unglamorous problems of stereo cameras outdoors: glare, ghosting, calibration drift, and the fact that a tree moves in the wind. Vision that survives a farm is a different engineering problem from vision that survives a demo. ## Droplets, not litres, are the real specification The J160's spray system covers a droplet range from 1 to 500 microns, which sounds like a spec-sheet flourish until you map it to crops. Fine droplets penetrate a canopy but drift; coarse droplets land but roll off. Mites want one setting, a soil drench wants another. The more interesting feature is spot spraying. Using the same perception stack, the aircraft can treat a defined radius around a detected canopy rather than a continuous swath, or follow a prescription map generated from an earlier survey. That is where the "AI" claim actually attaches to something measurable: less chemical applied to the gaps between trees. The company says field work in Fujian's pomelo belt cut pesticide use by about 30 percent while improving control in the lower canopy to 95 percent, and that in a Sichuan pepper trial on sloping ground, a lighter model cut water use sharply and raised yield. Those are the numbers that matter to a grower — cost per mu, and whether the pest comes back. ## The export story is the growth story China's agricultural drone market is crowded and price-sensitive, with two very large players setting the floor. EAVision's strategy has been to go where the geometry is hard and the crop is expensive: durian in Malaysia, citrus in Southeast Asia, row crops in Brazil. The company says it now sells into more than 30 countries and works with major seed and chemical firms including Bayer and BASF on trials — partnerships that matter less for revenue than for credibility, because a chemical company will not put its product in a tank it does not trust. Its 2025 expansion claim is the one to watch: six new overseas markets, and overseas revenue more than five times the previous year. That is a company figure from a bank profile of the firm, not an audited segment disclosure, but it points at the real shift. For Chinese agri-drone makers, the domestic market is a proving ground; the margin is abroad, where a machine that saves labour is worth more than a machine that is cheap. ## What readers can do now - **Judge an agri-drone by crop, not by payload.** Litres matter less than whether the aircraft can hold a canopy-following line on a slope with wires overhead. - **Ask for the trial data behind the chemical-saving claim**, including the control plot and the pest-counting method. "Thirty percent less pesticide" means little without knowing what the untreated trees looked like. - **If you are buying services rather than hardware**, price per mu including refills, battery swaps and the operator — the drone is rarely the expensive part. ## Honest limitations Most technical figures here — payload, detection range, point-cloud throughput, droplet range, hover response — come from EAVision's own product pages and launch materials and were not independently measured. The pesticide-reduction and yield claims come from company field summaries; we could not obtain the underlying trial designs or untreated-plot data. The company's cumulative treated-area figures are inconsistent across its own and third-party accounts published months apart, so we deliberately did not use them. The "more than 30 countries" and "six new overseas markets, revenue up more than fivefold" claims come from a bank profile quoting company managers, not from audited financials; EAVision is privately held and does not publish revenue. Partner names such as Bayer and BASF indicate trials, not commercial endorsement. Nothing in this article is a recommendation of the hardware, and prices vary too widely by configuration and region to quote reliably. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-app-eavision-orchard-spray-vision Free to quote with attribution and a link to the original.