Improved Military Camouflaged Target Detection via Multimodal Image Fusion

Effective detection of camouflaged targets remains a key challenge in military surveillance, where adversarial concealment strategies reduce visibility across individual spectral bands. This paper investigates the performance of multispectral image fusion using long-wave infrared (LWIR), near-infrared (NIR), and visible (VIS) sensors for enhanced camouflaged target detection. Four fused image configurations: LWIR + VIS, LWIR + NIR, NIR + VIS, and LWIR + NIR + VIS, are evaluated against single-sensor inputs using MUDCAD-X benchmark dataset. The detection performance is analyzed across different target types and color groups (green, gray, and yellow). Results show that the detection within the three fused modalities (LWIR + VIS, NIR + VIS, and LWIR + NIR + VIS) consistently outperform detection using individual sensors. Among all, LWIR + VIS yields the highest detection accuracy in terms of mAP@ 0. 5: 0. 95, followed by NIR + VIS. Among target groups, yellow targets remain the hardest to detect, while green targets are the most detectable. These findings demonstrate the capability of multimodal fusion to enhance detection systems for military surveillance and reconnaissance in complex natural scenes and provide color-specific detection insights for predicting camouflage effectiveness.