Abstract:The process of clearing areas, namely demining, starts by assessing and prioritizing potential hazardous areas (i.e., desk assessment) to go under thorough investigation of experts, who confirm the risk and proceed with the mines clearance operations. This paper presents Desk-AId that supports the desk assessment phase by estimating landmine risks using geospatial data and socioeconomic information. Desk-AId uses a Geospatial AI approach specialized to landmines. The approach includes mixed data sampling strategies and context-enrichment by historical conflicts and key multi-domain facilities (e.g., buildings, roads, health sites). The proposed system addresses the issue of having only ground-truth for confirmed hazardous areas by implementing a new hard-negative data sampling strategy, where negative points are sampled in the vicinity of hazardous areas. Experiments validate Desk-Aid in two domains for landmine risk assessment: 1) country-wide, and 2) uncharted study areas). The proposed approach increases the estimation accuracies up to 92%, for different classification models such as RandomForest (RF), Feedforward Neural Networks (FNN), and Graph Neural Networks (GNN).

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to support landmine clearance in humanitarian aid by developing a system named Desk - AId, using Geospatial AI to predict potential landmine areas. Specifically, the paper attempts to solve the following key problems: 1. **Improve the accuracy of landmine risk assessment**: - The first step in the landmine clearance process is "desk assessment", which is to identify and prioritize areas where landmines may exist by analyzing data such as reports, maps, and historical records. The paper proposes a Geospatial - AI - based method, combining geographical and socio - economic information, to estimate landmine risks more accurately. 2. **Solve the problem of insufficient negative sample data**: - In landmine risk assessment, usually only Confirmed Hazardous Areas (CHAs) have ground - truth data, and the data in safe areas are very limited. To solve this problem, the paper introduces a new hard - negative data sampling strategy, sampling negative sample points near known hazardous areas to increase data diversity and improve the generalization ability of the model. 3. **Enhance risk prediction in unexplored areas**: - The paper not only focuses on risk assessment on a national scale but also specifically tests the effect of risk prediction in completely unexplored areas. Through this method, resource allocation can be better guided to ensure effective landmine clearance work can also be carried out in unknown areas. 4. **Improve the detection accuracy of geographical granularity**: - Existing landmine risk assessment methods usually can only provide an overview of larger areas and cannot be precise to smaller geographical units. The paper improves the risk prediction accuracy of small - scale areas by introducing advanced machine - learning techniques such as Graph Neural Networks (GNNs), with a minimum granularity of 50 meters. 5. **Accelerate landmine clearance operations**: - By providing more accurate risk assessment tools, it can help experts determine areas that need to be intensively investigated more quickly, thereby reducing unnecessary waste of resources and accelerating the overall process of landmine clearance. ### Main contributions - **Constructed a comprehensive landmine detection dataset**: Combining data from multiple fields such as geography, socio - economics, and war remnants, a comprehensive landmine - polluted dataset was constructed, and the relationship analysis between different features was provided. - **Designed an intelligent balanced data sampling method**: Aiming at the problem of insufficient negative sample data, a strategy of sampling near positive samples was proposed, effectively alleviating the data imbalance problem. - **Explored the application of graph neural networks**: Using graph neural networks to model geographical proximity relationships significantly improved the accuracy of risk prediction. - **Conducted extensive experimental verification**: Multiple experiments were carried out across Afghanistan and in unexplored areas to verify the effectiveness of the proposed method. In conclusion, this paper is committed to improving the accuracy and efficiency of landmine risk assessment through advanced Geospatial AI technology, thereby accelerating the progress of landmine clearance and ultimately improving the quality of life of people in areas affected by landmines.

Desk-AId: Humanitarian Aid Desk Assessment with Geospatial AI for Predicting Landmine Areas