Research On The Method And Application Of Multi-types Stochastic Graphics' Generation

Based on the summings-up and analyses of existing generation methods of stochastic graphics, several generation methods of stochastic graphics are investigated under a uniform framework aimed at improving its predictability and controllability. Both modeling techniques of free shapes and adaptive learning mechanism of neural network are introduced into our new generation methods of stochastic graphics, which provide new approaches to enhance the modeling function and simulating ability of stochastic graphics.Chapter 1 analyzes in the round both advantages and disadvantages of existing modeling methods of stochastic graphics and divers existent problems are pointed out as well. The main contents and thoughtway of the dissertation are proposed with a view to obtain the better predictability and controllability of stochastic graphics.Chapter 2 endows the control points with new meaning and puts forward a basic model of generating stochastic graphics through the operation of control points based on the modeling techniques of free shape. The method to obtain control points with various properties, the rules to adjust linear combination coefficients and the controlling method of adaptive neural network are presented, which greatly enhance the modeling ability of stochastic graphics and provide a new approach to generate stochastic graphics with different kinds of shape.Chapter 3 puts forward a method to generate stochastic graphics from dispersion to dispersion based on the random combination of control points, and discusses in detail how to control the feature, tendency, fuzziness as well as random gradual change of stochastic graphics by taking advantage of the randomness of linear combination coefficients and control points. The new method can effectively control the fuzzy transient and discontinuous shape feature between the discrete nubby structures of stochastic graphics.Chapter 4 presents two methods of genrerating stochastic graphics based on adaptive neural network control and on interpolation displacement respectively. The method based on adaptive neural network control utilizes free curved face as learning object and generates stochastic graphics, which possesses irregular structure and good controllability, without the introduction of random displacement disturbance. On the basis of de Casteljau algorithm for the discrete generation of free curved face, two interpolation displacement methods, i.e. uniform interpolation displacement method and nonuniform interpolation method are putforward. In our new methods, displacement disturbances with different properties are used to generate various complicated stochastic graphics. The creation of irregular elaborate structure is integrated into the process of interpolation by making use of nonuniform recursive interpolation. As a result, stochastic graphics with irregular structure can be generated without the introduction of randomness.Chapter 5 proposes a method to generate stochastic graphics based on multivariable random maps and hybrid control over diversified mappings. The randomness and irregularity of variate-value are employed to control the randomness and irregularity of stochastic graphics. Multi-point mapping, multi-variable mapping and various shapes are all used to control the randomness of stochastic graphics. The forming procedure of geometrical texture and synthesizing procedure of local color texture are manipulated by the Manhattan distance between two points, which is fairly good at achieving coherent change.In view of the fact that some important natural phenomena have the characteristic of random shape, chapter 6 applies the methods of generating stochastic graphics presented in this dissertation to the generation of geometrical shapes of cloud cluster, transformation of cloud layer, terrain and land form, tornado etc. and their result simulating figures are given in correspondence.Chapter 7 summarizes the research work conducted in the dissertation and points out the originality innovations. Recommendations for future research are also included.