自然边坡的形成因素较多，属性也较复杂。由于边坡沉降产生的滑坡、坍塌、泥石流等灾害对人类造成的损失严重，因此对于边坡监测的研究显得非常必要。本文在探讨当前边坡监测技术的基础上，着重研究一种基于判决决策的边坡沉降相似度度量方法。该方法首先使用三频彩色条纹投影轮廓术来快速精确地恢复被测边坡的三维高程信息，在此基础上采用基于相似度度量的结果进行判决决策，进而判定被测边坡是否发生沉降，是否产生虚警。本文的主要工作内容包括彩色条纹投影轮廓术、相似度度量以及基于决策级的融合判决，具体如下： 首先，在边坡监测数据获取阶段，本文先对常用的光学三维测量方法进行研究，分析比较它们的特点和优势，确定选择三频彩色条纹投影轮廓术作为边坡监测数据获取的方法。重点研究二维经验模式分解、背景消减和颜色解耦、三频变精度相位展开等方法，并开展了边坡模拟实验，通过仿真实验，证明三频彩色条纹投影轮廓术用于边坡沉降监测的可行性。 然后，在边坡监测数据处理阶段，主要研究如何利用三频彩色条纹投影轮廓术恢复的高程信息数据来对边坡的沉降情况做出判决。重点研究了基于Hausdorff距离和欧氏距离两种相似度度量判决方法。使用Hausdorff距离做判决时，因为原始的Hausdorff距离对虚警现象敏感，本文采用改进的Hausdorff距离解决了此问题。使用欧氏距离做判决时，由于获取到的边坡高程信息矩阵较大，不利于分析和计算，本文对边坡高程信息进行二维主成分分析，实现数据降维，再通过不同时刻的欧氏距离对边坡的沉降情况做出判决。 最后，针对单一的相似度度量方法判决准确度不高的问题，本文在决策级将基于改进的Hasdorff距离和2DPCA的欧氏距离这两种相似度判决结果进行有效融合。边坡相似模拟实验表明，本文的融合算法比单一的判决方法判决率更高。

Natural slope is more forming factors and complex attributes. It is necessary to research the slope monitoring, because the landslide, collapse, debris flow and other disasters produced by slope sedimentation have caused serious losses for the human being. The paper focuses on a method of similarity measure on slope sedimentation based on decision fusion, at the basis of discussion about the current technology. Tt utilizes triple-frequency color fringe projection profilometry to fast and precisely reestablish the height information of the measured slope, then decision is be made based on the similarity measure about the acquired height information to judge whether there is an deformation and false alarm or not. The paper involves triple-frequency color fringe projection profilometry, similarity measure and judgment based on the decision fusion. The works in concrete are mainly as follows： Firstly, in the data acquiring of measured slope, the common optical 3D measurement is researched, their characteristics and advantages are analyzed and compared, and the approach of this paper about the data obtained is confirmed—color fringe projection profilometry. The paper is emphasis on the bidimensional empirical mode decomposition, background subtraction and color decoupling, variable precision phase unwrapping and so on. The slope simulation tests conducted, the results demonstrate that the feasibility of color fringe projection profilometry used to monitor the slope sedimentation. Secondly, in the data processing of the acquired slope, this paper mainly studies how to use height information regained from the color fringe projection profilometry to make decision on the slope sedimentation. Two similarity measure based on the Euclidean distance and Hausdorff distance are focused studied. When Hausdorff distance as a decision, the original algorithm is sensitive to false alarm, the problem solved by the improved Hausdorff distance. When Euclidean distance as a decision, the original data is so large that it must be realized dimensionality reduction by 2D principal component analysis, and then according to the Euclidean distance of slope corresponding to different time to make decision. Lastly, for the single similarity measurement accuracy is not high, this paper adopts the decision-making level fusion between modified Hausdorff distance and Euclidean distance. The experimental results indicate that the fusion algorithm has a higher judgment rate than the single decision.