Discussion of Results
Weather is one of the main factors when it comes to detecting landmines. The primary weather conditions that affect landmine detectability are diurnal cycles, clouds, wind, and rain. When the temperature is hotter, the temperature difference between the landmine and the surrounding soil is greater. The clouds can block the solar heat flux, which reduces the differences in temperature. Increased winds can cool the landmine quicker than the soil, reducing the difference in their temperatures. Lastly, increased rain can saturate the soil, which increases its thermal inertia and makes it take longer to change temperatures. This effect would make landmines on the surface easier to detect when the soil is wet. Another observation made was that even a small coat of spray paint or other substance that has lower emissivity could make the landmine much harder to detect. If camo patterns were used to spray paint landmines, then landmines even on the surface could become almost invisible to infrared cameras that primarily detect based on reflection. Progression of the project was significantly affected by logistical and equipment-related setbacks, which limited the scope of experimental testing and prevented full implementation of the planned system. Delays associated with inoperable waterjet cutting equipment hindered the timely fabrication of simulated landmine targets. This resulted in a compressed testing schedule and reduced opportunity for iterative refinement of experimental setups.
Additionally, delays in the procurement and integration of the thermal imaging system further constrained the project timeline. As a result, testing efforts were primarily focused on establishing a proof-of-concept for ground-based thermal detection rather than progressing to more advanced stages of the project. The reduced timeline prevented the successful implementation of drone-based data collection, which was originally intended to evaluate detection performance at operational altitudes.
Furthermore, due to these time constraints, the project did not advance to the development and training of a machine learning model for automated landmine identification. While preliminary data collection was completed and could serve as a foundation for future work, insufficient time remained to curate, label, and train a dataset of adequate size and quality.
These limitations should be considered when interpreting the results of this study. The findings primarily reflect idealized, ground-level detection conditions and do not fully represent the performance of a deployed aerial system. Future work should prioritize early acquisition of critical components and fabrication resources to ensure sufficient time for system integration, aerial testing, and data-driven model development.
Conclusion and Future Work
This study evaluated the feasibility of using infrared imaging systems for the detection of simulated landmines under varying environmental and material conditions. The results indicate that thermal imaging is moderately effective for identifying surface targets when sufficient thermal contrast exists between the target and surrounding soil.
Detection performance was found to be strongly dependent on time of day. Contrary to the initial hypothesis that post-sunset conditions would provide optimal detection due to thermal release, experimental results showed that peak solar loading conditions (midday to late afternoon) produced the most distinguishable thermal signatures. During these periods, targets absorbed and retained heat more effectively than the surrounding soil, resulting in increased contrast in thermal imagery.
Material properties and surface conditions were also shown to significantly influence detectability. Targets composed of varied materials exhibited varying thermal responses, and the presence of surface coatings or paint altered emissivity and reflectivity, in some cases reducing contrast and making detection more difficult. Additionally, environmental factors such as soil composition, vegetation cover, and surface uniformity impacted the clarity of thermal signatures, with more heterogeneous surfaces reducing detection reliability.
The study further demonstrated that camera geometry plays a non-negligible role in detection performance. Variations in camera angle affected the amount of infrared radiation captured, likely due to changes in reflected versus emitted energy reaching the sensor. This suggests that consistent viewing angles are important for reliable detection, and that oblique viewing angles may reduce effectiveness.
Due to limitations in sensor resolution and thermal sensitivity, testing of buried targets was not pursued. It is highly likely that subsurface landmines would not produce sufficient thermal contrast to be detectable with the imaging system used in this study. As a result, all testing focused on surface-level targets, representing a best-case scenario for detection capability.
Limitations in sensor capability were also identified as a major factor. The resolution and sensitivity of the thermal camera used in this study restricted detection performance, particularly when considering deployment from a drone platform. It is expected that a higher-resolution, higher-sensitivity thermal imaging system would significantly improve detection capabilities, especially at increased altitudes. Furthermore, qualitative observations suggest that real-time video analysis may provide better target recognition compared to still imagery due to temporal changes in thermal contrast; however, it remains unclear whether video or still image datasets would be more effective for training machine learning models.
Overall, while infrared imaging shows promise as a non-contact method for landmine detection, its effectiveness is highly dependent on environmental conditions, material properties, sensor quality, and viewing geometry. As implemented in this study, the method is best suited for controlled, surface-level detection scenarios and would require significant enhancements for reliable real-world applications.
Future work should focus on expanding the system from a ground-based proof-of-concept to a fully integrated aerial detection platform. A primary objective will be the successful implementation of drone-based data collection, including testing at varying altitudes and flight speeds to determine operational limits for reliable detection. This will require the use of a higher-resolution and higher-sensitivity thermal imaging system, as the current sensor is unlikely to provide sufficient detail at typical drone flying altitudes.
Further investigation is also needed to determine the most effective data acquisition and processing approach. Comparative studies between real-time video feeds and still image datasets should be conducted to evaluate their suitability for both human detection and machine learning applications. Building a sufficiently large and well-labeled dataset will be critical for developing a robust machine learning model capable of distinguishing between landmine signatures and environmental false positives.
Additional testing should be conducted to better understand the influence of environmental conditions and surface characteristics on detection performance. This includes controlled studies on surface coatings, material emissivity, soil composition, and vegetation coverage, all of which were shown to significantly affect thermal contrast. While subsurface targets were not evaluated in this study, future work may explore whether improved sensors or alternative imaging techniques could enable limited detection of shallow buried objects.
Finally, improvements in experimental planning and resource availability should be prioritized. Timely access to fabrication equipment and imaging hardware will allow for more extensive testing, iterative design improvements, and full system integration. With these enhancements, future efforts can build upon the findings of this study to develop a more reliable and scalable infrared-based landmine detection system.