The arithmetic is brutal, almost absurd. Ukraine's contaminated territory keeps shifting—138,000 square kilometers one month, 156,000 the next—depending on whether you're counting technical surveys or merely suspected hazards. Either way, it's roughly the size of England, all of it seeded with explosives. Traditional demining runs €1 to €3 per square meter. The Ukrainian government's current estimate for clearing it all: $34.6 billion over a decade. At the pace crews managed in early 2025—about 41,000 hectares in six months—the UN's mine-action advisor Paul Heslop quietly projects decades of work. Five years, he suggests, if you're only talking about the most immediate threats. And that's if the funding holds.
Which it probably won't.
This is why a small Danish-Ukrainian company called Dropla Tech has spent the past two years assembling something different: a multi-sensor detection platform that layers RGB cameras, thermal imaging, multispectral sensors, magnetometry, and LiDAR onto a fleet of drones, then feeds it all through machine learning models trained on what the company describes—with the confidence typical of startups—as Europe's most comprehensive landmine dataset. The pitch is straightforward, almost too clean: a six-drone swarm can survey half a square kilometer per day and cut costs to around €0.10 per square meter. That's not incremental. That's a 10-to-30-fold reduction, the kind of number that either changes an entire industry or collapses under scrutiny.
In August 2025, Dropla raised €2.4 million in pre-seed funding from Maj Invest, Denmark's Export and Investment Fund, and Final Frontier. Modest by Silicon Valley standards. Significant for a sector that has, for decades, relied on manual metal detectors and ground-penetrating radar operated by people inching across suspect terrain, one careful step at a time.
The Scale of the Problem
The casualty figures keep climbing. In 2023, landmines and explosive remnants killed or injured at least 5,757 people globally, 84% of them civilians, according to Landmine Monitor 2024. By 2024, that number rose to 6,279, with children accounting for nearly half of civilian casualties. Ukraine now sits among the top four countries by mine-related deaths and injuries—joining Afghanistan, Myanmar, and Syria in a category no nation wants to occupy.
The World Bank, European Commission, and UN released their fourth Rapid Damage and Needs Assessment in February 2025, pegging Ukraine's total reconstruction cost at $524 billion over ten years. Debris management alone: $13 billion. Demining threads through nearly every line item. The challenge isn't just humanitarian, though it's certainly that. It's economic. You can't rebuild power infrastructure or return farmland to cultivation when the ground beneath might explode. You can't reopen railways. You can't really do much of anything.
By mid-2025, Ukraine had certified 107 demining operators, scaling capacity faster than most observers expected. The Ministry of Defence reported clearing 41,243 hectares in the first half of the year—a pace reflecting both crisis urgency and the rapid deployment of mechanized equipment, trained personnel, and increasingly, technology that keeps humans out of the kill zone.
What's Actually New Here
The breakthrough isn't any single piece of hardware. Drone-mounted magnetometers have existed for years, mostly for surveys of old military ranges littered with unexploded ordnance. Thermal imaging from UAVs? Standard in border surveillance and precision agriculture. What's changed in the past 18 months is convergence: multiple sensors on the same platform, data processed in real time by neural networks trained to distinguish a buried anti-tank mine from a rusted tractor part.
HALO Trust, one of the largest humanitarian demining organizations operating in Ukraine, began testing drone-mounted magnetometers in Mykolaiv in 2025, identifying 17 explosives in early trials. Ukraine's Ministry of Defence showcased a UAV-magnetometer system in May that could scan more than 10 hectares daily, generating GIS-referenced anomaly maps for ground teams. In June, UNDP equipped the State Emergency Service with a locally developed "Mines Eye" system combining magnetometry, visual and infrared imaging, and AI processing.
Dropla's approach sits at the sophisticated end of this emerging spectrum. The company's "Dropla Vision" platform ingests data from five sensor types simultaneously: RGB cameras for surface features, multispectral imaging to detect disturbed soil, thermal sensors for heat-signature anomalies (mines sometimes retain or release heat differently than surrounding ground), magnetometry for ferrous objects, and LiDAR for high-resolution terrain mapping. The company claims 95% detection accuracy across object categories, though that figure hasn't appeared in peer-reviewed contexts or been validated by independent third parties.
The harder problem—perhaps the hardest—is georeferencing. GPS signals jam or fail near active combat zones, yet demining requires sub-15-centimeter precision. If your hazard map is off by a meter, someone dies. Dropla spent much of 2024 and 2025 working on GNSS-denied positioning systems at Denmark's Odense Robotics incubator, using visual odometry and other techniques to maintain accuracy when satellites aren't cooperating. This isn't a minor engineering challenge. It's the difference between a useful tool and a liability.
The Sensor Fusion Puzzle
"Sensor fusion" sounds like jargon until you understand the problem. A magnetometer triggers on any ferrous metal: a buried anti-personnel mine, yes, but also fence posts, soda cans, agricultural debris. Thermal imaging works under certain conditions but struggles with surface clutter or when soil temperature equalizes. Multispectral cameras can spot disturbed earth, though they might also flag recently plowed furrows. Each sensor, alone, generates a high rate of false positives.
Stack them together with a trained neural network, and the noise collapses. A 2023 study in Sensors using YOLO-based detection on multispectral data showed improved recall compared to RGB alone. A December 2025 study using YOLOv11 achieved mean average precision of 86.8% on fused RGB and long-wave infrared datasets, though anti-personnel mines—smaller, less metallic—proved harder to detect than anti-tank types.
Dropla's models train on data from HUB-194, a 15.26-hectare testing ground in Ukraine scattered with more than 140 inert mines and 30 live ordnance items. The company claims to have integrated more than 400,000 real-world datasets into its training corpus. That's substantial, though still smaller than Safe Pro Group's dataset, which exceeded 36,000 labeled landmine detections from over 2 million analyzed drone images by October 2025, according to company statements.
The research community is catching up, albeit unevenly. In December 2025, a team released AMLID, the first open UAV dataset combining RGB and LWIR imagery with 12,078 labeled images covering 21 mine types. Another dataset, MineInsight, published in June 2025, integrated RGB, visible-to-short-wave infrared, LWIR, and dual LiDAR from an unmanned ground vehicle. The Geneva International Centre for Humanitarian Demining held an innovation session in October 2024 specifically calling for a global reference database of explosive ordnance imagery to accelerate algorithm development. The fact that they had to call for it tells you something about the current state of data sharing in this field.
Why Speed Matters (and Costs Even More)
Traditional handheld demining with dual-sensor metal detectors and ground-penetrating radar—devices like the VALLON VMR3 "Minehound" widely used in UN peacekeeping missions—costs roughly €1 to €3 per square meter when you factor in personnel, equipment, insurance, and operational overhead. A trained deminer might cover a few hundred square meters per day in favorable conditions. Less in dense vegetation or rocky terrain.
Dropla's claimed €0.10 per square meter comes from slashing labor time. A six-drone swarm surveying 0.5 square kilometers daily covers 500,000 square meters—more than 2,000 times a single manual operator's output, according to profiles from Invest in Odense and the company's own materials. Of course, not all that surveyed area requires intrusive clearance. The drones identify anomalies; ground teams still investigate and neutralize threats. But if you can eliminate 90% of the search area through remote sensing, you've just made the job ten times faster. And proportionally cheaper.
AILAND Systems, a Ukrainian company, claims its "Spinner EOD" drone-mounted detection system is up to 20 times faster than human visual inspection for surface ordnance. A 2026 field study published in MDPI Drones documented 100% detection of ferromagnetic unexploded ordnance buried less than 60 centimeters deep, covering 0.53 hectares in under an hour. These aren't apples-to-apples comparisons—different terrain, different target types, different verification protocols. But the trend is unmistakable.
The market is responding. Mordor Intelligence projects the UXO detection sector growing from $3.15 billion in 2025 to $6.17 billion by 2030, a 14.4% compound annual growth rate. Another firm, Reanin, forecasts a more conservative trajectory: $3.55 billion in 2025 to $5.81 billion by 2032, implying 7.3% CAGR. The discrepancies likely reflect category definitions, but both point the same direction—expansion, driven largely by Ukraine's needs and accelerating adoption of automated systems.
Building an Ecosystem
Dropla's €2.4 million raise is part of a broader pattern. The company operates from Odense, Denmark—a city that has cultivated a deliberate robotics cluster—and maintains R&D operations in Ukraine, where founder V'yacheslav Shvaydak holds a PhD and has spent years navigating the technical and regulatory landscape. The European Space Agency's ARTES program provided early-stage support. Final Frontier, one of the lead investors, cited battlefield validation and edge AI as core investment theses.
But Dropla isn't working in a vacuum. The ecosystem includes established players like SENSYS, Geometrics, and GEM Systems, which manufacture UAV-mounted magnetometers used globally for UXO surveys. Safe Pro Group, a U.S.-based company trading on Nasdaq under ticker SPAI, was selected for the U.S. Army's 2026 Combined Federated Warfare Exercise and has secured integrations with major defense contractors. AILAND's Spinner EOD received NATO codification in December 2025. These are no longer garage experiments. They're systems entering institutional procurement pipelines.
The regulatory infrastructure is evolving alongside the technology. Ukraine introduced a formal specialty for UAV operators focused on explosive ordnance detection, with standard SSU 820-4:2025 taking effect May 1, 2025. The International Mine Action Standards (IMAS 07.11) updated land release guidelines to clarify how non-intrusive methods like camera-equipped drones fit into evidence-based clearance frameworks. UNDP, working with APOPO and Mines Advisory Group, is integrating Technical Survey Dogs—yes, trained rats and dogs—into national protocols under a €2 million EU grant.
Perhaps most critically, the sector is beginning to address its data problem. Demining algorithms are only as good as their training datasets, and those datasets have historically been fragmented, proprietary, or classified. Dropla has indicated intentions to contribute standardized data to global repositories. The release of open datasets like AMLID and MineInsight signals a shift toward collaborative benchmarking. The Geneva Centre's push for a centralized image database reflects institutional recognition that vendor claims alone won't scale the sector—it needs reproducible, peer-reviewed validation.
What Comes Next
The immediate future of mine clearance technology is less about inventing new sensors than integrating existing ones more intelligently. Magnetometry works. Multispectral imaging works. Thermal detection and LiDAR work. The question is how to fuse them in ways that minimize false positives, maximize coverage speed, and maintain the sub-meter georeferencing accuracy that life-or-death decisions require. Those aren't small qualifiers.
Edge AI processing will likely become standard. Dropla's "Blue Eyes" unit processes video at roughly 130 frames per second directly on the device, eliminating the latency and bandwidth constraints of cloud processing. That matters for military route clearance, where a convoy can't wait 20 minutes for a server farm in Poland to analyze footage. It matters equally for humanitarian demining in areas with intermittent internet access—which is to say, most places where mines are actually a problem.
Autonomy is the next frontier. Dropla's SEER system pairs an aerial detection platform with an unmanned ground vehicle designed for neutralization tasks. The concept: detect, map, navigate, and neutralize without continuous human intervention in the hazard zone. HALO Trust is piloting mechanized systems like the MineWolf flail in Ukraine—machines that churn soil to detonate or disable mines, though they're expensive and prone to breakdown. Smaller, modular UGVs that can handle vegetation clearance, logistics, and selective excavation represent a different approach. Whether they work at scale remains an open question.
The funding landscape remains... complicated. The European Union's Foreign Policy Instruments directorate has committed roughly €80 million to humanitarian mine action in Ukraine since 2022, with combined EU, member state, and EFTA contributions exceeding €370 million as of February 2025. Lithuania and Iceland lead a Demining Capability Coalition that has mobilized over €100 million and delivered 240 mine detectors, 39 mine rollers, and other equipment through August 2025. The United States remains the largest global donor at $198.1 million in 2024, though that figure is down 36% year-over-year amid shifting political priorities in Washington.
The sector will need sustained capital if it's to compress Ukraine's decades-long clearance timeline to something resembling the five-year target for priority areas that UN advisors consider feasible. TechCrunch argued in February 2025 that Europe will "look to Ukraine for the future of defense tech," and demining is a clear test case. The World Bank's reconstruction framework ties mine clearance directly to infrastructure and agricultural recovery, creating potential pathways for public-private partnerships and performance-based financing. Whether those pathways materialize is another matter.
What remains uncertain is how fast institutions can absorb innovation. Humanitarian demining organizations operate under strict protocols, for good reason—mistakes kill people. Integrating AI-driven detection into IMAS-compliant workflows requires documentation, quality assurance, and independent validation that takes time. Ukraine's adoption of national standards for UAV-based detection suggests one path forward, but extending those frameworks to other contaminated countries will require diplomacy, training, and likely, more open datasets. The bureaucracy isn't optional.
If Dropla and its competitors can deliver on their cost and speed claims—and critically, if they can prove those claims through transparent, third-party validation—the sector could be looking at a productivity revolution comparable to what precision agriculture experienced over the past decade. The alternative is another 70 years of manual metal detectors and patient crawls across suspect ground, one square meter at a time.
The technology to change that equation exists. Whether the institutions, regulations, and funding can keep pace is the real question. And on that front, the data isn't promising.
