Modeling of Curves and Surfaces using Soft Computing Techniques

Abstract

The main goal of the thesis is to design a robust, efficient and accurate method for the reconstruction of curves and surfaces. Modelling of shapes from a set of unorganized points with no information about the connectivity of the points is a problem with lot of practical significance and is encountered in many fields of computer graphics, image processing and multimedia applications. In this thesis, some new techniques for the reconstruction of curves and surfaces have been proposed. The performance of these techniques has also been compared with the existing techniques. The problem of curve and surface reconstruction has been tackled in two different ways in this thesis. First is the reconstruction of curves and surfaces from dense point cloud and the second is reconstruction of curves and surfaces using extracted features. newlineIn the first scheme various neural networks such as a multi-layer perceptron (MLP) and a radial basis function neural network (RBFNN) have been used for reconstructing curves and surfaces from unorganized point cloud data. A comparison of these neural networks for reconstruction problem is given with theoretical and experimental analysis. Some standard curves and surfaces are reconstructed using the neural network techniques. Also a new approach based on fuzzy logic and ant colony optimization (ACO) is presented which is able to reconstruct complicated curves including open, closed, self- intersecting and multiple components of a curve. This proposed scheme is robust to noise, non- uniform data and small gaps in the data as well. The advantages of the method have been established by comparing it with the existing methods. It has been shown that this hybrid approach reconstructs the curve with a good approximation and so is accurate, robust and efficient.