LAYER PEELING THROUGH VOLUME-BASED CLUSTERING IN 3D POINT CLOUD MODELS
Kok-Why Ng, Abdullah Junaidi · 2013
Three-dimensional (3D) point cloud model presents limited geometric detail. Many existing research works have to apply Nearest-Neighbor (NN) search technique to implicitly find the closest appropriate point to form regular triangle mesh, in order to acquire more geometric detail (e.g. vectors, area and etc.) to define the model. This paper introduces a volume-based clustering technique to group the associated points in different regions. The region can adaptively be partitioned based on the density of the points. Each region will be consigned a key point to represent all its members. This can easily control the complexity of the model and alleviate the computational process. In our implementation, we further extend this to layer peeling which can reveal more of the inner structure of a model for other research application. We have tested on three models and produce different layers of peeling. Each gives unique and interesting region representation. This piece of work is significance for beginners who would deal with high density of 3D point cloud models.