|
| 基于无人机正射影像采集滑坡表面位移数据的技术研究与应用 |
| Research and Application of Technology for Collecting Surface Displacement Data of Landslides Based on Orthophoto Images of Unmanned Aerial Vehicles |
| 投稿时间：2025-04-27 修订日期：2025-05-09 |
| DOI： |
| 中文关键词: 滑坡监测 无人机正射影像 改进YOLOv8 房屋目标检测 深度学习 |
| 英文关键词:Landslide Monitoring UAV Orthophoto Improved YOLOv8 House target detection Deep Learning |
| 基金项目: |
|
| 摘要点击次数: 15 |
| 全文下载次数: 0 |
| 中文摘要: |
| 滑坡作为典型的地质灾害，其早期识别与精准监测对于灾害预警和风险防控至关重要。针对传统GNSS点式监测在空间覆盖的局限，本文基于无人机正射影像采集风险斜坡单元，提出一种改进YOLOv8模型的滑坡房屋位移监测方法，并以湖南吕洞村滑坡区为案例进行实证研究。通过搭载RTK模块的无人机获取高精度正射影像，构建房屋监测数据集，利用引入Swin Transformer与BiFPN模块改进的YOLOv8模型，实现对房屋中心像素点的精准识别与提取，并结合不同时间影像开展变形量分析。监测结果表明，该方法能够识别滑坡变形区并揭示降雨诱发的变形规律，识别结果与GNSS监测数据最大差值不超过8.77 mm，误差率在10%以内，验证了本方法的可靠性与实用性。研究结果可为滑坡灾害的高效识别、快速响应和早期预警提供新思路与技术支撑。 |
| 英文摘要: |
| As a typical geological disaster, the early identification and precise monitoring of landslides are of vital importance for disaster early warning and risk prevention and control. In view of the limitations of traditional GNSS point monitoring in spatial coverage, based on the orthophoto image acquisition of risk slope units by unmanned aerial vehicles (UAVs), this paper proposes an improved YOLOv8 model for monitoring the displacement of landslide houses, and conducts an empirical study taking the landslide area of Ludong Village, Hunan Province as a case. High-precision orthophoto images are obtained through unmanned aerial vehicles equipped with RTK modules to construct a house monitoring dataset. The YOLOv8 model improved by introducing Swin Transformer and BiFPN module is utilized to achieve accurate identification and extraction of the central pixel points of the house, and deformation analysis is carried out in combination with images at different times. The monitoring results show that this method can identify the landslide deformation area and reveal the deformation law induced by rainfall. The maximum difference between the identification result and the GNSS monitoring data does not exceed 8.77mm, and the error rate is within 10%, verifying the reliability and practicability of this method. The research results can provide new ideas and technical support for the efficient identification, rapid response and early warning of landslide disasters. |
| 查看/发表评论 下载PDF阅读器 |
| 关闭 |