基于ASF-WIoU-YOLOv8的无人机航拍图像多目标检测算法

Multi-Object Detection Algorithm for UAV Aerial Images Based on ASF-WIoU-YOLOv8

摘要: 针对无人机航拍图像的多目标检测问题，本文提出了一种基于改进YOLOv8的目标检测算法ASF-WIoU-YOLOv8。首先，在YOLOv8的基础架构上，加入一种注意尺度序列融合机制(Attentional Scale Sequence Fusion-ASF)，该机制能够对不同尺度的特征图进行融合，从而获得更好的图像特征，提取出更丰富、更准确的特征信息。然后，对损失函数进行改进，引入Wise-IoU机制，该机制通过自适应地调整权重系数提高目标检测的灵活性和鲁棒性，从而进一步提高算法的检测精度。实验结果表明，在VisDrone数据集上，本文所提算法比YOLOv8算法的平均精度mAP50提升了2.0%，该算法在无人机航拍图像上具有更高的检测精度。

Abstract: For the problem of multi-object detection in UAV aerial images, this paper presents an improved YOLOv8 object detection algorithm named ASF-WIoU-YOLOv8. Firstly, on the infrastructure of YOLOv8, an attention-scale sequence fusion mechanism (Attentional Scale Sequence Fusion-ASF) is added, which can integrate feature maps at different scales, so as to obtain better image features and extract richer and more accurate feature information. Then, the loss function is improved by using the Wise-IoU mechanism, which improves the flexibility and robustness of target detection by adaptively adjusting the weight coefficients. Wise-IoU can further improve the detection accuracy of the algorithm. The experimental results show that the average accuracy of the proposed algorithm is 2.0% higher than the YOLOv8 algorithm, which has higher detection accuracy on aerial images of UAV.