Deep Learning-Based Camouflage Detection for Anti-Personnel Mine Identification in Natural Environments
Abstract
The detection of anti-personnel mines in natural environments remains a critical humanitarian and technological challenge due to the high visual similarity between explosive devices and their surrounding backgrounds. This study presents a deep learning-based camouflage detection framework for the identification and segmentation of PFM-1 anti-personnel mine surrogates embedded in visually homogeneous outdoor scenes. The proposed approach employs a Deep Camouflage Detection Network (DCDN) designed to extract multi-scale contextual features and enhance boundary sensitivity under low-contrast conditions. A dedicated dataset was constructed using 3D-printed PFM-1 surrogates positioned in vegetated environments under varying illumination conditions, viewing angles, occlusion levels, and object scales. The network architecture integrates a pretrained convolutional backbone, multi-scale feature aggregation modules, and a composite loss function (consisting of Binary Cross-Entropy and Dice loss) to address class imbalance and weak edge contrast.
Experimental evaluation on an independent test set demonstrates robust segmentation performance, achieving a mean Intersection over Union (IoU) of 91.8% and a Dice coefficient of 95.7%. Precision and recall values of 96.3% and 94.9%, respectively, confirm a balanced detection capability with limited false positives. Stratified analysis indicates stable performance under illumination variability and partial occlusion, while ablation studies highlight the importance of multi-scale aggregation and region-aware loss optimization. The results confirm that deep camouflage-aware segmentation architectures provide reliable detection of low-contrast objects in complex natural environments.
Downloads
References
[2]. ***, Anti-Personnel Landmines Convention (Ottawa Convention), United Nations Office for Disarmament Affairs (UNODA), accessed Feb. 23, 2026.
[3]. ***, International Campaign to Ban Landmines (ICBL-CMC), Landmine Monitor 2023, Geneva, Switzerland, Nov. 2023.
[4]. Baur J., et al., Applying deep learning to automate UAV-based detection of scatterable landmines, Remote Sensing, 12(5), 859, 2020.
[5]. García-Fernández M., et al., Autonomous airborne 3D SAR imaging system for subsurface sensing: UWB-GPR on board a UAV for landmine and IED detection, Remote Sensing, 11(20), 2357, 2019.
[6]. Marin F.-B., Marin M., Drone detection using image processing based on deep learning, Annals of “Dunarea de Jos” University of Galati, Fascicle IX, 44(4), p. 36-39, 2021.
[7]. Ronneberger O., et al., U-Net: Convolutional networks for biomedical image segmentation, Proc. MICCAI, p. 234-241, 2015.
[8]. He K., et al., Deep residual learning for image recognition, Proc. IEEE CVPR, p. 770-778, 2016.
[9]. Liu Z., et al., Swin Transformer: Hierarchical vision transformer using shifted windows, Proc. IEEE ICCV, p. 10012–10022, 2021.
[10]. Woo S., et al., CBAM: Convolutional block attention module, Proc. ECCV, p. 3-19, 2018.
[11]. Lin T.-Y., et al., Feature pyramid networks for object detection, Proc. IEEE CVPR, p. 2117-2125, 2017.
[12]. Fan D.-P., et al., Concealed object detection, IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(10), p. 6024-6042, 2022.
[13]. Le T.-N., et al., A. Sugimoto, Anabranch network for camouflaged object segmentation, Computer Vision and Image Understanding, 184, p. 45-56, 2019.
[14]. Skurowski P., Kasprowski P., Evaluation of saliency maps in a hard case-images of camouflaged animals, Proc. IPAS, p. 244-249, 2018.
[15]. Abraham N., Khan N. M., A novel focal Tversky loss function with improved attention U-Net for lesion segmentation, Proc. IEEE ISBI, p. 683-687, 2019.
