针对遥感图像背景复杂、尺度多变等特点导致目标检测精度低的问题，提出一种基于改进YOLOv8的遥感图像目标检测算法。首先，基于YOLOv8网络框架，在骨干网络部分引入全局注意力机制，使模型更关注于重要区域，降低无用背景对模型的干扰；其次，采用一种明智的交并比作为新的边界框回归方式，减小低质量示例对模型的影响，提高模型的边界框回归水平；最后，在特征融合网络部分和头网络部分分别引入高效多尺度注意力模块和动态检测头模块，全面提升模型的尺度适应能力。实验结果表明，所提算法在公开遥感数据集DIOR和RSOD上均取得较高的检测精度，阈值为0.5的平均精度均值分别达到了72.0%和94.6%，并且改进后的算法具有较强的鲁棒性，当输入数据集改变时，仍能保持较好的检测性能。

Aiming at the problems of low detection accuracy caused by complex background and variable scale of remote sensing images， an improved YOLOv8 based remote sensing image target detection algorithm is proposed. First， based on the YOLOv8 network framework， the global attention mechanism is introduced in the backbone network to make the model pay more attention to important areas and reduce the interference of useless background on the model. Secondly， a wise crossover ratio is adopted as a new bounding box regression method to reduce the influence of low-quality examples on the model and improve the bounding box regression level of the model. Finally， an efficient multi-scale attention module and a dynamic detection head module are introduced into the feature fusion network and the head network respectively to improve the scale adaptability of the model. The experimental results show that the proposed algorithm achieves high detection accuracy on both DIOR and RSOD， and the average accuracy with a threshold value of 0.5 reaches 72.0% and 94.6% respectively. Moreover， the improved algorithm has strong robustness and can still maintain good detection performance when the input data set changes.