Detecting these hidden threats is a first essential step to counter them. Since they are by design difficult to detect for humans, automatic detection technologies can play an important role. However, the task is intrinsically difficult, and the performance of existing technologies is still far from answering the needs. Scenarios classically encountered by armed forces in past missions such as route clearance already represent a challenge. In addition, IEDs are increasingly used in urban scenarios
where the detection is even more difficult, especially if multiple IEDs emplacements are used. There is a need to enhance detection technologies, especially for scenarios where single detection devices are not sufficient and the use of distributed sensors is deemed useful. There is also a need to determine their type (e.g., how they are triggered), in particular to ease their neutralisation (rendering safe, disabling or destroying).

While the above issues have been the subject of much research over many years, progress is hindered by the lack of standardised benchmarks, and there is a need to evaluate the performances of integrated functional demonstrators in an objective and
comparable manner, using representative testing environments and well-defined metrics.

Overall progress in IED and landmine detection and characterisation can be driven by progress along several lines:

• Physics-based sensors enhancement;
• Collection of representative data, combined with various artificial intelligence (AI) techniques, e.g., computer vision for object detection and localisation;
• Use of various sensors borne by a fleet of unmanned ground and aerial systems, combined with information fusion techniques;
• Better exploitation of limited amounts of data and use of models that are easier to adapt to new environments (through innovative AI techniques such as learning methods requiring less supervision from expert developers,
  transfer learning…);
• Multidisciplinary cooperation between the hardware sensors and AI communities.