Visible-infrared object detection for Unmanned Aerial Vehicles (UAV) can exploit the texture details of visible images and the thermal radiation information of infrared images, thereby improving target perception capability in complex environments. However, existing methods mainly focus on cross-modal feature interaction and feature fusion, while ignoring high-frequency interference in single-modal features before fusion, such as thermal noise, pseudo-hotspots, texture redundancy, and background edges. As a result, high-frequency noise is further amplified during cross-modal propagation, reducing detection accuracy. To address these problems, this paper proposes a UAV visible-infrared object detection method based on frequency purification and interactive fusion. Through a “purification-before-fusion” strategy, the proposed method suppresses single-modal high-frequency interference and enhances cross-modal complementary information. Specifically, a Low-frequency Aggregation Frequency Purification (LAFP) module is designed to use cross-modal low-frequency structural priors to guide the selection and purification of high-frequency detail features in each modality, suppressing invalid high-frequency responses while preserving target edges and local discriminative features. In addition, a Bidirectional Interactive Guidance Fusion (BIGF) module is introduced. This module employs bidirectional cross-attention to mine the complementary relationships between purified visible and infrared features, and combines a dynamic gating mechanism to achieve adaptive cross-modal feature fusion. Experimental results on three visible-infrared object detection datasets, namely DroneVehicle, VEDAI, and LLVIP, show that the proposed method achieves good performance in both complex aerial vehicle detection and low-light pedestrian detection scenarios. Compared with existing state-of-the-art methods, the proposed method achieves mean Average Precision (mAP) values of 82.4% and 80.5% on the DroneVehicle and VEDAI datasets, respectively, outperforming the second-best methods by 3.0% and 2.7%. On the LLVIP dataset, it obtains an average precision of 96.1%, which is 0.7% and 0.4% lower than the best and second-best results, respectively, but still demonstrates strong competitiveness. The experimental results demonstrate that the proposed algorithm has high effectiveness and robustness in UAV object detection scenarios.

To address the problems of catastrophic forgetting, overfitting, and difficulty in adapting dense detector characteristics leading to limited detection accuracy faced by the current time-sensitive target class incremental object detection of UAV images, this paper proposes a time-sensitive target incremental detection algorithm based on multistage distillation. The proposed method includes three key modules: the continuous Wasserstein distance-based Inter-Class Distillation (WICD) module, the Prototype-Guided Intra-class Consistency Distillation (PGICD) module, and the Cross-prediction Adaptive Distillation (CAD) module. The WICD module captures the inter-class feature differences from the feature graphs and semantic query vectors using Gaussian distributions with continuous Wasserstein distances to enhance the inter-class discriminability. The PGICD module enhances the inter-class discriminative properties by minimizing the high-level semantic query of the instances in both the teacher's network and student's network, and the low-level feature graphs with the prototype differences to achieve effective intra-class feature transfer and enhance intra-class consistency. The CAD module optimizes the cross-prediction distillation process by dynamically adjusting the distillation weights of the classification and regression branches, mitigating the problem of catastrophic forgetting in incremental learning, and improving the model's detection accuracy in complex scenarios. Experimental results on SIMD and MAR20 datasets show that the proposed method performs well in all types of one-step and multistep incremental scenarios, and the Average Precision (AP) is significantly improved compared with traditional methods. The AP is as high as 70.8% in the incremental scenario of SIMD dataset with 8+7 classes, with an absolute gap of 1.7% and a relative gap of 2.3% from the upper limit. The results also show that the AP is as high as 60.2% in the incremental scenario of MAR20 dataset with 10+10 classes, with an absolute gap of 2.3% and a relative gap of 3.6% from the upper limit. In addition, the effectiveness of each module is verified by ablation experiments, which effectively improves the incremental detection performance of time-sensitive targets in UAV images.