Research Of Similarity Measure On Slope Sedimentation Based On Decision Fusion

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.