崩岗不仅加剧了水土流失问题，而且治理难度大，严重影响农业生产，带来一系列的社会问题。传统崩岗监测方法，在时空分辨率、监测范围、效率、成本等方面各有不足，已无法满足当前崩岗自动识别和监测需求。近年来，遥感与摄影测量及无人机技术的飞速发展，为崩岗自动识别提供了新的解决思路。本项目拟开展基于无人机超高分辨率DSM和DOM的自动崩岗识别研究，首先提出一种适用于崩岗DSM的具有尺度不变性地学特征，能够从地形上区别于平缓地区，从而获得不同于外观特征的描述力；然后分析多种DOM特征对崩岗的辨识力，制定基于概率主题模型的崩岗DSM和DOM特征融合策略，更全面、独特地描述崩岗；最后设计面向崩岗识别的半监督支持向量机，实现高性能计算平台下基于DSM和DOM多特征潜在语义描述的崩岗自动识别。本研究从特征描述、特征融合和半监督分类三阶段最大程度提高崩岗识别精确度，为推动崩岗侵蚀机理和治理措施的研究提供基础支撑。

Benggang that exacerbates the problem of soil erosion is very difficult to be cured. It has a severe impact on agricultural production, and brings a series of social problems. Due to the limited spatial-temporal resolution, monitoring range, efficiency, cost and etc., traditional monitoring methods of Benggang are not able to satisfy the current demand of automatic Benggang recognition and monitoring. In recent years, rapid development in remote sensing, photogrammetry and unmanned aerial vehicles (UAV) technology provided us a new solution for automatic Benggang recognition..The research presented in this proposal mainly focus on the automatic Benggang recognition based on UAV Very High Resolution (VHR) DSM and DEM. .Firstly, a scale-invariant geology feature which is suitable for Benggang DSM is proposed, it can differentiate flat terrain, therefore obtain abilities to describe different forms of appearance..Secondly, we will analyze multiple characteristics of DOM features for Benggang, and then establish feature fusion policy of Benggang DSM and DOM based on Probabilistic Topic Model (PTM), in order to comprehensively and distinctively describe Benggang..Last but not the least, our research will design semi-supervised support vector machine for Benggang recognition, implement automatic Benggang recognition based on the latent sematic descriptions of multiple DSM and DOM on High Performance Computing platform..This research improves the accuracy of Benggang recognition in the feature description, feature fusion, and semi-supervised classification three procedures to the greatest extent, and therefore provides fundamental support to the research of Benggang erosion mechanism and cure measures.