Night UAV Vehicle Detection Based on Infrared-Visible Fusion and Improved YOLO11
Hongxia Su
Journal of Computer Science and Digital Technology ›› 2026, Vol. 2 ›› Issue (1) : 37 -50.
To address the challenges of vehicle detection accuracy in nighttime UAV aerial photography scenarios caused by low light conditions, strong noise, and dense small object distribution, this paper proposes an improved YOLO11 vehicle detection method based on multi-modal fusion. First, a collaborative enhancement strategy combining CLAHE and Gamma correction is applied to preprocessing nighttime visible light images, effectively restoring vehicle texture details in dark areas. Second, the enhanced visible light images are fused with infrared images through fixed-weight weighted fusion (infrared weight α=0.7), fully leveraging the complementary advantages of both modalities. Finally, the MANet module is introduced on top of YOLO11n to enhance backbone multi-scale feature extraction capabilities, while the ADFPN-DASI module is designed to collaboratively optimize neck multi-scale feature representation. Experiments conducted on the DroneVehicle dataset demonstrate that the proposed method achieves an mAP50 of 80.53%, representing an improvement of 11.45 percentage points over the YOLO11n baseline. The method maintains lightweight advantages in parameter and computational resources, outperforming mainstream comparison methods such as YOLOv5n.
Nighttime Vehicle Detection / Drone Aerial Photography / Multimodal Fusion / YOLO11 / Feature Pyramid Network
| [1] |
|
| [2] |
|
| [3] |
|
| [4] |
|
| [5] |
|
| [6] |
|
| [7] |
|
| [8] |
|
| [9] |
|
| [10] |
|
| [11] |
|
| [12] |
|
| [13] |
|
| [14] |
|
| [15] |
|
| [16] |
|
/
| 〈 |
|
〉 |