Small object detection remains a challenge in the remote sensing field due to feature loss during downsampling and interference from complex backgrounds. A novel network, termed SEMA-YOLO, is proposed in this paper as an enhanced YOLOv11-based framework incorporating three technical advancements. By fundamentally reducing information loss and incorporating a cross-scale feature fusion mechanism, the proposed framework significantly enhances small object detection performance. First, the Shallow Layer Enhancement (SLE) strategy reduces backbone depth and introduces small-object detection heads, thereby increasing feature map size and improving small object detection performance. Then, the Global Context Pooling-enhanced Adaptively Spatial Feature Fusion (GCP-ASFF) architecture is designed to optimize cross-scale feature interaction across four detection heads. Finally, the RFA-C3k2 module, which integrates Receptive Field Adaptation (RFA) with the C3k2 structure, is introduced to achieve more refined feature extraction. SEMA-YOLO demonstrates significant advantages in complex urban environments and dense target areas, while its generalization capability meets the detection requirements across diverse scenarios. The experimental results show that SEMA-YOLO achieves mAP₅₀ scores of 72.5% on the RS-STOD dataset and 61.5% on the AI-TOD dataset, surpassing state-of-the-art models.

Small object detection remains a challenge in remote sensing field due to feature loss during downsampling and interference from complex backgrounds. A novel network, termed SEMA-YOLO, is proposed in this paper as an enhanced YOLOv11-based framework incorporating three technical advancements. By fundamentally reducing information loss and incorporating a cross-scale feature fusion mechanism, the proposed framework significantly enhances small object detection performance. First, the Shallow Layer Enhancement (SLE) strategy reduces backbone depth and introduces small-object detection heads, thereby increasing feature map size and improving small object detection performance. Then, the Global Context Pooling-enhanced Adaptively Spatial Feature Fusion (GCP-ASFF) architecture is designed to optimize cross-scale feature interaction across four detection heads. Finally, the RFA-C3k2 module, which integrates Receptive Field Adaptation (RFA) with the C3k2 structure, is introduced to achieve more refined feature extraction. SEMA-YOLO demonstrates significant advantages in complex urban environments and dense target areas, while its generalization capability meets the detection requirements across diverse scenarios. Experimental results show that SEMA-YOLO achieves mAP50 scores of 72.5% on the RS-STOD dataset and 61.5% on the AI-TOD dataset, surpassing state-of-the-art models.

Traditional unmanned aerial vehicle (UAV) path planning methods for image-based 3D reconstruction often rely solely on geometric information from initial models, resulting in redundant data acquisition in non-architectural areas. This paper proposes a UAV path planning method via semantic segmentation of 3D reality mesh models to enhance efficiency and accuracy in complex scenarios. The scene is segmented into buildings, vegetation, ground, and water bodies. Lightweight polygonal surfaces are extracted for buildings, while planar segments in non-building regions are fitted and projected into simplified polygonal patches. These photography targets are further decomposed into point, line, and surface primitives. A multi-resolution image acquisition strategy is adopted, featuring high-resolution coverage for buildings and rapid scanning for non-building areas. To ensure flight safety, a Digital Surface Model (DSM)-based shell model is utilized for obstacle avoidance, and sky-view-based Real-Time Kinematic (RTK) signal evaluation is applied to guide viewpoint optimization. Finally, a complete weighted graph is constructed, and ant colony optimization is employed to generate a low-energy-cost flight path. Experimental results demonstrate that, compared with traditional oblique photogrammetry, the proposed method achieves higher reconstruction quality. Compared with the commercial software Metashape, it reduces the number of images by 30.5% and energy consumption by 37.7%, while significantly improving reconstruction results in both architectural and non-architectural areas.

Traditional image-based 3D reconstruction methods often rely solely on geometric information from initial models, resulting in redundant data acquisition in non-architectural areas. This paper proposes a UAV path planning method via semantic segmentation of 3D reality mesh models to enhance efficiency and accuracy in complex scenarios. The scene is segmented into buildings, vegetation, ground, and water bodies. Lightweight polygonal surfaces are extracted for buildings, while planar segments in non-building regions are fitted and projected into simplified polygonal patches. These photography targets are further decomposed into point, line, and surface primitives. A multi-resolution image acquisition strategy is adopted, featuring high-resolution coverage for buildings and rapid scanning for non-building areas. To ensure flight safety, a DSM-based shell model is utilized for obstacle avoidance, and sky-view-based RTK signal evaluation is applied to guide viewpoint optimization. Finally, a complete weighted graph is constructed, and ant colony optimization is employed to generate a low-energy-cost flight path. Experimental results demonstrate that, compared with traditional oblique photogrammetry, the proposed method achieves higher reconstruction quality. Compared with the commercial software Metashape, it reduces the number of images by 30.5% and energy consumption by 37.7%, while significantly improving reconstruction results in both architectural and non-architectural areas.