在深度学习领域，目标检测模型的性能通常依赖于充足且高质量的标注数据集。然而，在遥感考古应用中，大规模、多样化的数据集获取困难且成本高昂，这导致数据不足的问题，使模型在训练时容易过拟合。尤其在墓葬赋存环境差异较大的情况下，特定环境特征的过度代表会导致样本不均衡，从而限制模型的环境适应能力。为了解决这些问题，本文提出了一种基于扩散模型的遥感影像数据增强算法。通过设置不同的环境提示词，在不破坏原始数据集标签分布的前提下，丰富图像的视觉语义，缓解数据样本稀缺和样本不均衡的问题。本研究基于谷歌地球构建了阿勒泰地区的高分辨率墓葬影像数据集，并评估了该算法在不同目标检测模型中的性能。实验结果表明，该算法在测试集上的平均精度比基线方法提高了12.2%，在异源数据集上的平均精度提升了16.4%。本文提出的算法显著提高了模型的检测准确性、稳定性及跨数据集的适应性，为小样本遥感考古目标识别提供了有效的技术支撑，拓展了智能考古检测的研究思路。

n the field of deep learning, the performance of object detection models typically relies on abundant and high-quality annotated datasets. However, in remote sensing archaeology applications, obtaining large-scale and diverse datasets is challenging and costly, leading to data scarcity issues that cause models to overfit during training. Particularly in cases where the environmental conditions of burial sites vary significantly, the overrepresentation of specific environmental features can lead to sample imbalance, which restricts the model"s adaptability to different environments. To address these issues, this paper proposes a remote sensing image data augmentation algorithm based on diffusion models. By setting different environmental prompts, the algorithm enriches the visual semantics of images without altering the original dataset"s label distribution, thereby alleviating data scarcity and sample imbalance problems. This study constructed a high-resolution burial site image dataset for the Altai region using Google Earth and evaluated the performance of various object detection models with this algorithm. Experimental results demonstrate that the proposed algorithm improves the average precision on the test set by 12.2% compared to the baseline method and enhances average precision on heterogeneous datasets by 16.4%. The proposed algorithm significantly improves the model"s detection accuracy, stability, and adaptability across different datasets, providing effective technical support for small-sample remote sensing archaeological object recognition and expanding the research perspectives in intelligent archaeological detection.