This paper presents the application of iterated function system (IFS) based three-dimensional (3D) fractal interpolation to elevation data compression. The parameters of contractive transformations are simplified by a...This paper presents the application of iterated function system (IFS) based three-dimensional (3D) fractal interpolation to elevation data compression. The parameters of contractive transformations are simplified by a concise fractal iteration form with geometric meaning. A local iteration algorithm is proposed, which can solve the non-separation problem when Collage theorem is applied to find the appropriate fractal parameters. The elevation data compression is proved experimentally to be effective in. reconstruction quality and time-saving.展开更多
基金国家自然科学基金(the National Natural Science Foundation of China under Grant No.60573096)甘肃省教育厅基金(the Foundationof Gansu Province Educational Department under Grant No.0604-09)
文摘属性规约是应对"维数灾难"的有效技术,分形属性规约FDR(Fractal Dimensionality Reduction)是近年来出现的一种无监督属性选择技术,令人遗憾的是其需要多遍扫描数据集,因而难于应对高维数据集情况;基于遗传算法的属性规约技术对于高维数据而言优越于传统属性选择技术,但其无法应用于无监督学习领域。为此,结合遗传算法内在随机并行寻优机制及分形属性选择的无监督特点,设计并实现了基于遗传算法的无监督分形属性子集选择算法GABUFSS(Genetic Algorithm Based Unsupervised Feature Subset Selection)。基于合成与实际数据集的实验对比分析了GABUFSS算法与FDR算法的性能,结果表明GABUFSS相对优于FDR算法,并具有发现等价结果属性子集的特点。
文摘This paper presents the application of iterated function system (IFS) based three-dimensional (3D) fractal interpolation to elevation data compression. The parameters of contractive transformations are simplified by a concise fractal iteration form with geometric meaning. A local iteration algorithm is proposed, which can solve the non-separation problem when Collage theorem is applied to find the appropriate fractal parameters. The elevation data compression is proved experimentally to be effective in. reconstruction quality and time-saving.