Research on Camouflage Target Detection in Two-Branch Multi-Source Remote Sensing Images

In military operations, the accurate detection of enemy camouflaging targets is of great significance to improve our combat capability. Most of the existing camouflage target detection methods are aimed at a single kind of natural image, and their ability to detect military camouflage image in complex combat environment is limited. In order to solve this problem, this paper proposes a two-branch multi-source remote sensing image camouflage detection method. By combining the fine-grained features of the target in infrared and visible images, the method fuses the camouflage target information to improve the detection capability. Experiments show that the accuracy, recall rate and [email protected] of the proposed method are 0.88, 0.859 and 0.923, respectively, which are 0.054, 0.099 and 0.096 higher than the values of YOLOv5s, YOLOv7 and [email protected] based on IFCNN method, respectively. Therefore, the method proposed in this paper has outstanding performance in the detection of pedestrian and vehicle camouflage targets, and provides a new method in the field of multi-source information fusion camouflage targets detection.