Beyond Drones and AI: Rethinking the Future of Humanitarian Demining

I have been working with drones since 2014, but the outbreak of the war in Ukraine represents a turning point in my career. Since 2022, my focus has turned to explore how drones can be used to automate human waves – what capabilities they need, and how technology can make these efforts safer and more efficient. As part of this work, I closely follow the Geneva International Center for Human Arming (Gichd), attend their events, and interact regularly with their experts.

Given the solutions -based drones associated with AI, they are actually useful only in the NTS scanning stage (NTS) of the human removal process. This means that drones wiped large areas and collects data. Then, the automatic learning model analyzes this data to the mark areas may It contains mines. It is not the exact places of mines.

Technical scanning (TS), which confirms and maps of polluted areas, is still dependent on employees with metal detectors, trained dogs, and mechanical removal machines. They go to the microscopic location area.

The process continues to be long, risks and expensive:

Mines are also a threat to civilians – there were at least 5,757 mines in 2023.

In this post, I explain why the current drones in the technical questionnaire (the most expensive stage and take a long time at the present time) and share what I see is the best way to fix this.

Discovering mines under the soil or vegetation is almost impossible

Done -made optical or thermal aircraft usually take pictures from an angle facing down. This approach works well to discover abnormal cases at the surface level, but it fails to discover buried or hidden mines. For this reason, drones are often used for non -technical surveys to remove human waves.

One of the Safe Pro Ai solutions is that they have a discovery rate of only 5 percent in areas with trees and shrubs.

Although it is less related to Ukraine, where most mines spread to the ground, rather than burying it, the situation is completely different (for example) to Cambodia:

• 4-6 million landmines are still conflicts in the 1970s to the 1990s
• 64000+ victims since 1979, with children as the main victims

It is difficult to discover non -mineral and ancient metal mines, even on the surface

Non -metal mines offer a large portion of landmines in current and previous conflict areas. It is intentionally designed to overcome the detection of traditional metal detectors.

Visually, it is difficult to discover non -mineral mines. They do not shine, or they rush in the pictures, or they appear well on thermal cameras. Minerals and magnetic standards either miss or make a lot of wrong warnings.

Therefore, the currently -based detection tools are often lacking completely non -mineral miners.

When it comes to ancient metal mines, corrosion changes what it looks and behaves, mixing with the ground and badly responding to detection tools. It is difficult to get to know those distorted in the pictures.

Because these mines are more difficult to locate, they take longer to find and remove them, or they remain hidden and put both Deminers and civilians at risk.

Weather and dependence during the day

If we are talking about drones with RGB and multi -spectrum cameras, they require daylight. In cloudy or low areas, or shaded areas (forests, antiquities), image quality decreases and detection of objects as well.

Thermal detection, in turn, works better at dawn or dusk, when the Earth and the mine differ in the temperature. During the middle of the day, the sun is evenly heated evenly, which reduces the contrast.

While rain and soil in the soil change the soil color and temperature, and can hide soil disorders or anomalous thermal conditions. Snow covers visual signs and equals the surface temperature, making mines uncovered.

The drones only fly at certain times to the NTS stage of removal of security, especially in unpredictable areas.

Technology is very expensive

In 7 countries, the anti -eclipse pollution area was estimated at more than 100 square kilometers.

According to the tests in Ukraine, the removal of the new technology can Cut Costs from 3000 to $ 5,000 to $ 600-800 per hectare, which still is $ 70,000 per square kilometer. In some areas, the price of the land may exceed itself.

The main cause of high costs is the multiple misconduct that treats real threats. On average, the team wipes more than 50 suspected mines to find only one actual ground games element.

Most of the areas are largely polluted in developing countries. They cannot afford to remove discrimination without financing from international organizations or governments.

The costs are very high for the companies they jump. Once the performance removal becomes cheap enough, companies may rent lands contaminated with mines, provided that they are clarified. On the other hand, they will get a long -term use of a symbolic price and some tax exemptions.

solution?

With my team, we explore the methods that collect more data, and we can see through the foliage and soil, and we still maintain enough accuracy.

An example of promising development is a project by researchers at Oviedo University. They are testing the GPR-SAR radar system installed on drones.

The validity of the journey in realistic scenarios has proven that technology solves the following problems:

1) The radar determines the mine’s location accurately, leaving only disarmament or destruction manually.

With all possible radar paths (full -time composition), they got high -resolution images where buried goals were brighter and more clear. She managed to discover difficult targets such as small, non -metallic and buried organisms, such as anti -plastic floor mine, wood pressure panels and PVC tubes.

2) The solution can work day or day, in diverse weather, and even with mild vegetation.

How to work:

• He sends radar pulses to the ground.
• It discovers reflections of sub -surface changes (for example, plastic, metal, voids).
• It builds three -dimensional images with a surface with accuracy at a centimeter by combining radar signals from several pairs and flight sites.

The solution still has its restrictions, but based on my background, it is the most related trend to research and development at the present time.

One of the main strengths of GPR is the amount of data you can collect. More data means that researchers can improve accuracy in the identification/classification phase with artificial intelligence. This leads to a more efficient and removal survey and reduces total costs by 50 % or more.