Detecting small objects in unmanned aerial vehicle (UAV) imagery is a challenging and crucial task in computer vision. Most current methods struggle to address the challenges of small objects: fine-grained feature mining, multiple-layer feature fusion, and mismatches in scale between anchors and feature maps. To alleviate the aforementioned issues, we present FGHDet, which focuses on delving into fine-grained features in low-level features with a head selection mechanism. First, our approach introduces a detail-preserving semantic information enhancement module (DSIEM) to retain fine-grained information while excavating coarse-grained semantic details relevant to fine-grained information. Then, we devise a coarse-to-fine feature guidance module (CFGM) that leverages coarse-grained semantic information and fine-grained information to co-guide feature enhancement, further improving the model's classification ability. Finally, we introduce a multiscale detection strategy based on anchor-head matching, ensuring scale-level matching between anchors and feature maps to prevent overfitting due to overly fine anchor divisions. Extensive experiments on the VisDrone, CARPK, and Drone-vs.-Bird datasets demonstrate that FGHDet achieves notable improvements in $mAP$ (IoU range [0.5: 0.95]) of 4.9, 4.1, and 2.2, respectively. The code is available at https://github.com/b-yanchao/UAVDetection.git.

Despite the recent accomplishments in joint infrared-visible imaging, the bimodal defocus blur (BDB) phenomenon received scant attention. Our analysis reveals that BDB is predominantly attributable to disparities in optical parameters between cameras, resulting in two primary challenges: incomplete single-modal information and difficulty in cross-modal information interaction. With regard to the former, the infrared modality is the primary victim, as the deblurring networks’ bias toward high-frequency results in erroneous low-frequency reconstruction (e.g., over-sharpening). In the latter case, the relative nature of the blur effect can lead to ambiguity in determining which modality’s information should be prioritized for guidance, and conflicts may arise between the clear components of the blurred image and the blurry components of the clear image. To address these issues, we propose the first de-bimodal defocus blur (DBDB) method, which consists of a low-frequency semantic hold (LSH) module with a pre-trained infrared model and a cross-modal complementary feature induction (CCFI) module driven by a max-min blur entropy loss. LSH is designed to ensure that the low-frequency information captured by the infrared modality does not contain any misleading data, while CCFI facilitates the acquisition of accurate information by means of adaptive adjustment and the loss function. The experimental results of deblurring and downstream tasks on two synthetic datasets demonstrate the superiority of our method.