期刊
ENGINEERING GEOLOGY
卷 259, 期 -, 页码 -出版社
ELSEVIER
DOI: 10.1016/j.enggeo.2019.105166
关键词
Landslide; Image recognition; Superpixel segmentation; Continuous deformation
资金
- National Natural Science Foundation of China [4181101168, 51874268]
- Nanchong School Science and Technology Strategic Cooperation Special Project [NC17SY4016]
- Sichuan Bureau of Surveying, Mapping and Geoinformation Science and Technology Plan Project [2017ZC05]
Obtaining continuous landslide deformation information is important for analyses of landslide processes. This paper proposes a landslide deformation area image recognition method. Using this method, the landslide area Can be automatically identified, and continuous landslide deformation data can be obtained. This novel method is implemented using Python and the OpenCV open source libraries, and it is validated using video data of rainfall-induced landslide experiments. The results indicate that the image recognition method can identify the landslide features in video monitoring images with high accuracy. An analysis of the error in the recognition results shows that the environmental changes in light intensity and specular reflections are the main causes of the recognition errors. The new image recognition method is straightforward and can be applied to bare slopes, especially man-made slopes.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据