Research On 3D Point Cloud Simplification Algorithm Based On Two-stage Fuzzy Pattern Recognition Model

Reverse engineering is a key technology that gradually developed in the nineties of the last century, it can be regarded as a process that turning a physical model into a digital model. In the process of reverse engineering, data acquisition equipment is used to collect data from the original model. With the wide use of laser scanning equipment in data acquisition, the data collected by which becomes more and more huge. If these data are processed directly, it will not only hard to get an ideal result, but also will consume vast amount of time and memory space. Therefore, it is significantly necessary to simplify the point cloud model.The main research object of this paper is three dimensional cloud points model of human organs. The three dimensional cloud points model gained by laser scanning device do not have any spatial indexes structure and neighborhood relations. So, when establish spatial indexes and neighborhood relations for these points, we have a better management of them, and it's also very convenient for us to do some operations such as calculate curvature and neighborhood search in the process of point cloud simplification and surface reconstruction. After the establish of spatial indexes and neighborhood relations for these points, then begin to simplify points cloud. The simplification methods in the past always only focus on one particular feature of points cloud which cannot keep multiple features of original model very well. In this paper, fuzzy optimization pattern has been optimized, a new simplification algorithm of three dimensional cloud points based on two-stage fuzzy pattern recognition model has been proposed. This method can simplify points cloud base on multiple features at the same time, which can lead to a more comprehensive result of the features details of the original model.To be specific. In the first, we establish the octree of spatial point cloud data, and establish spatial data index structure of point cloud, then we build K- neighborhood relations for the point based on the index. In the second, in order to guarantee the simplification of point cloud can consider multiple characteristics index, the fuzzy optimization model is improved to the two level fuzzy pattern recognition model, which allows algorithm to evaluate object with multiple parameters at the same time, then get the value of relative optimal membership degree, and accordingly we can evaluate all objects. Finally, in order to take more valuable information into account in the process of simplification, some appropriate characteristic parameters need to selected as the indicators of simplification. There are a lot of surface characteristic parameters in the model, the point of curvature reflects the bending degree of local curved surface where the point located, but the curvature is too sensitive of the characteristics of flat area, many details will lose when the simplification happen in flat area, which produce hollow in the model; And the deviation of the normal vector represent the deviation of normal direction in local surface when remove one point, which reflects the point's contribution to the local characteristic information, and it is sensitive to the larger bending area, simplification of flat areas have very good effect. Therefore, we select the parameters of the curvature and normal vector deviation as indicators of simplification, and then using the proposed approach for simplification.