| 基于生成对抗网络的动水驱动型滑坡状态识别方法 |
| |
| 引用本文： | 徐庆杰,刘勇,詹伟文,郭敬楷,李星瑞.基于生成对抗网络的动水驱动型滑坡状态识别方法[J].地质科技通报,2022,41(6):129-136. |
| |
| 作者姓名： | 徐庆杰 刘勇 詹伟文 郭敬楷 李星瑞 |
| |
| 作者单位： | 中国地质大学(武汉)机械与电子信息学院, 武汉 430074 |
| |
| 基金项目： | 国家自然科学基金重大项目42090054国家自然科学基金项目41772376自然资源部地质灾害自动化监测技术创新中心开放基金项目2022058014 |
| |
| 摘 要： | 动水驱动型滑坡状态识别能更有效地辅助分析滑坡形变规律, 实现滑坡状态的准确识别对深入展开动水驱动型滑坡状态研究具有重要意义。针对目前动水驱动型滑坡突变状态研究较少, 难以获得相关特征, 从而导致状态识别性能不佳的问题, 提出了一种应用于动水驱动型滑坡状态识别的生成对抗网络学习方法。本方法通过构建滑坡状态监测数据矩阵, 依据少量数据样本设计合理的生成器网络完成对滑坡状态的数据扩增并设计判别器网络实现扩增数据的筛选, 通过对抗生成网络实现对滑坡状态的分类, 达到滑坡状态识别的目的。以三峡库区白水河滑坡为研究对象, 将降雨、库水位、深部位移和地表位移等多源监测数据进行了规范化处理, 设计生成器网络和对抗器网络完成了对滑坡状态数据的扩增, 设计滑坡状态识别生成对抗网络完成了对滑坡状态的分类和识别。结果表明, 生成对抗网络对滑坡状态识别具有较高的准确率。研究结果证实本方法能够对目标区域内的动水驱动型滑坡状态进行准确识别和分类, 可直接应用于工程实际。
|
| 关 键 词： | 动水驱动型滑坡 生成对抗网络 滑坡状态 白水河滑坡 |
| 收稿时间： | 2022-05-19 |
State recognition method of hydrodynamic pressure-driven landslides based on a generative adversarial network |
| |
| Institution: | School of Mechanical Engineering and Electronic Information, China University of Geosciences(Wuhan), Wuhan 430074, China |
| |
| Abstract: | Identifying the states of hydrodynamic pressure-driven landslides can more effectively assist in the analysis of the landslide deformation law, and accurately identifying the landslide state is of great significance for the in-depth study of the hydrodynamic pressure-driven landslide state. Aiming at the problem that there are few abrupt states of hydrodynamic pressure-driven landslides, it is difficult to obtain relevant features, which leads to poor state recognition performance, and a generative adversarial network learning method for landslide state recognition is proposed.In this method, the landslide state monitoring data matrix is constructed, a reasonable generator network is designed based on a small number of data samples to complete the data amplification of the landslide states, and the discriminator network is designed to realize the screening of the amplified data, and the classifying landslide states being realized through the confrontation generation network to achieve the purpose of landslide status identification. Taking the Baishuihe landslide in the Three Gorges Reservoir area as the research object, multisource monitoring data such as rainfall, reservoir water level, deep displacement, and surface displacement are normalized. The states recognition generative adversarial network completes the classification and identification of the landslide state. The results show that the generative adversarial network has high accuracy for landslide state recognition. The research method in this paper can accurately identify and classify the landslide state in the target area. |
| |
| Keywords: | |
|
| 点击此处可从《地质科技通报》浏览原始摘要信息 |
| 点击此处可从《地质科技通报》下载免费的PDF全文 |
|
