Research And Application Of Partial Fitting Hybrid Technology For Discrete Data

Modeling curves and surfaces is the basis of Computer-Aided Design and Computer-Aided Manufacturing, which is one of the most important problems in reverse engineering. Both of curve fitting and surface fitting have been hot areas of approximation theory in these years. They have a wide range of applications in many fields, such as shape design and manufacturing in aviation, aerospace, automobile, shipbuilding and other engineering fields, data processing and analyzing in geology, ocean, meteorology and other fields.Surface reconstruction is also of great significance in three-dimensional terrain modeling. Synthesizing large-scale terrain is one of the key technologies in the process of modeling virtual scenes. It has practical significance in enhancing the sense of reality and visual impact of virtual scene. It has been a common hot academic theme among the fields of Virtual Reality, Computer Simulation, Geographical Information System and others in recent years. Relative achievements on research of it play an important role in battlefield simulation, disasters simulation, outdoor games, video games, video production, road design, land planning, virtual reality and other application fields of geographic information system.Based on existing methods of curve fitting and terrain synthesizing, a novel fitting method by blending local least squares is proposed in this paper. Also the use of this method in curve fitting and terrain synthesizing is illustrated in this paper. For curve fitting, at first, the discrete data points are divided into several pieces. Then the algorithm generates piecewise approximation to the points of each piece. Finally, the approximation curve is blended together according to their weights, which is determined by the influence degree they have on each other. A smooth curve approximating all the discrete data points in a high precision is obtained. There are several controllable parameters such as radius of piece, weight function and so on, which can be used to control the approximating process in need of different sampling density and application. Experimental results show that this method has a stable performance and can be used to process all kinds of data in different resolutions.For terrain synthesizing, a novel method to synthesize large-scale digital virtual terrain from a set of halftone images is presented in this paper. At first, the algorithm obtain planar point set by a series of processing to the images, such as standardizing, soften, halftone. That is to say, algorithm converts continuous-tone image into halftone image composed of black points which are spatially non-uniform. Then the planar points are simplified and each point in it is given a value at random in certain ranges. After that algorithm creates terrain grids on the image area and evaluates the height value of each grid point by weighted blending. Finally, in order to eliminate the crack between two terrain grids, algorithm patches the terrain grids together by weighted blending. Experiments show that this algorithm realizes smooth transition among digital elevation models. Also, it proposes user several controllable parameters for varies terrains.