Deriving Hidden Junction in Solid Model Reconstruction Using Neural Network

Research on solid model reconstruction has been started since 1970s and it is still investigated since then. In other fields of research, it also takes important role such as in fields as diverse as product design, engineering and rapid prototyping, medical imaging and artistic applications. Furthermore, it has turned out to be essential needs. This paper presents a new algorithm in reconstructing solid model using neural network with back propagation that has been successfully used by few researchers in deriving depth values of visible junction. The algorithm presented in this paper extends the previous works by Matondang et al. in 2007, which is reconstruct solid model from twodimensional (2D) line drawing by deriving depth values. And now in this paper the reconstruction of solid model is continued by deriving the hidden junction and produces complete solid model. Affine transformation in form of rotation is employed to generate the coordinate values that are required in the development of the proposed algorithm. Besides the algorithm, this paper also presents a new framework in solid model reconstruction using neural network. Comparison among three algorithms of previous works in terms of the accuracy and the ease of the algorithm are presented. The experimental results of three 2D line drawings, cube, Lblock and stair, show that the proposed algorithm is capable in deriving hidden junction for solid model reconstruction and can be used as a new alternative in reconstructing solid model.