伊朗西北部水体分布

水库在陆地水文系统中发挥着关键作用，但很少有中小型水库的记录，尤其是在发展中国家，水库建设速度很快。准确识别水库对于理解社会和经济发展很重要，但将其与其他水体区分开来是一个重大挑战。因此，我们提出了一种使用高分辨率卫星图像和深度学习算法识别水库的方法。我们用各种参数和网络结构训练模型，You Only Look Once版本7（YOLOv7）的性能优于其他算法，并被选中构建最终模型。该方法被应用于伊朗西北部的一个地区，那里有丰富的各种大小的水库。评估结果表明，我们的方法具有很高的精度（总体精度：95.43%，用户精度：99.23%，生产者精度：91.57%）。我们能够识别该地区393个水库，表明所提出的方法可以准确地检测水库，并有可能在更大甚至全球范围内调查水库。

Reservoirs play a critical role in terrestrial hydrological systems, yet records of small- and medium-sized reservoirs are scarce, especially in developing countries where reservoir construction is advancing rapidly. Accurately identifying reservoirs is crucial for understanding social and economic development, but distinguishing them from other water bodies presents a major challenge. To address this gap, we propose a reservoir identification method that utilizes high-resolution satellite imagery and deep learning algorithms. We trained models with diverse parameters and network architectures, and You Only Look Once version 7 (YOLOv7) outperformed all other algorithms, thus being selected to develop the final model. This method was applied to a region in northwest Iran, which abounds with reservoirs of varying sizes. Evaluation results demonstrate that our method achieves high accuracy (overall accuracy: 95.43%, user accuracy: 99.23%, producer accuracy: 91.57%). We successfully identified 393 reservoirs in this region, indicating that the proposed method can accurately detect reservoirs and has the potential to conduct reservoir surveys at larger scales or even globally.