【作者】 王长缨；

【导师】 周明全；

【作者基本信息】 西北大学 ， 计算机软件与理论， 2006， 博士

【摘要】 时态GIS是近年来GIS领域在认识和实践方面的一项重要发展。本文报告了源于国家自然科学基金项目“林火行为的动态模拟信息系统”的时态GIS领域若干关键技术的研究。这些关键技术领域涉及GIS时空数据模型、图像融合和GIS时空分析领域的元胞自动机模型等,是构建时态GIS应用的具有一般意义的基础问题。本文在分析传统的GIS数据模型的基础上,面向课题应用提出了一个时空数据模型。对GIS领域涉及的图像融合与拼接问题进行了研究,提出了利用图像特征信息的进行拼接配准和降低计算量的像素级融合与拼接算法。在时空分析的元胞自动机模型方面,对地理元胞自动机模拟实际地理环境所存在的本质问题进行了研究,提出了规则学习的思路和方法,并研究了元胞自动机在地理图像处理方面的应用。最后综合上述研究,提出了一个以元胞自动机作为推理机驱动的林火蔓延专家系统模型。本文的主要工作和创新点如下: 1.分析了林火蔓延空间对象及其属性的时态性质,在时空数据模型ESTDM的基础上,提出了一个基于地理对象时态性质分类的、描述元胞空间的时空数据模型。该模型提高了空间数据的时态表达能力,有着较少的数据冗余和更好的操作性能,较好地反映了林火蔓延的时空特征,便于元胞自动机操作。 2.提出了利用图像特征信息的像素级图像(栅格DEM数据)融合与拼接算法,改善了图像配准并降低了拼接计算量。该方法对实时CCD多区输出和多CCD遥感图像拼接的也有重要的价值。 3.标准元胞自动机是按照一组确定的规则演进的,因此元胞空间的状态传播与地理空间的地理流传播有本质的不同,这使元胞自动机模型的准确性受到损害。为解决该问题提出了元胞自动机规则的免疫学习机制。 4.研究了元胞自动机进行地理图像处理的原理及其在图像降噪、边缘提取中的应用,提出了一个基于多尺度元胞的水体中心线提取高效算法。 5.提出了一个用元胞自动机推理驱动的林火蔓延专家系统模型。更多还原

【Abstract】 In recent years temporal GIS appears to be a new development of GIS. This dissertation focuses on some key techniques and algorithms of temporal GIS. Aim to build a forestry fire spreading model, a new temporal data model, data-image fusion and fast image mosaicing approaches, cellular automata model and it’s rule evolution algorithms are explored. The research work can be summaried as following:1. Based on the current data models of GIS and regarding the forestry fire speading model, a new scalabile temporal data model is proposed, which gives a natural representation of the real forestry fire behavior with less data redundancy and better fits for the operation of the cellular automata.2. Data integration is one of most time costy work for building a GIS application. And data-image fusion is the important step of data integration. Aiming to promote efficiency of data integration of GIS, a new approach of multiresolution data-image fusion is introduced, which utilizes features directed auto-align. The calculation time of the same approach is further reduced by localized calculation of Laplacian pyramid to a statistically defined boundary neighbor. The time response of the image mosaicing of multiple partition images of CCD is also benefit from the approach.3. A typical cellular automata runs with a set of deterministic rules. However in case of the environment change there is a need of rule adjustment. So an immunity learning method for the rule adjustment of 2d cellular automata is proposed to improve the precision of geo-cellular automata model.4. Some principles of image process with cellular automata are also discussed with two examples. One of the examples is the edge extraction of image which shows better result than typical operators. And then an approach of image skelecton extraction is presented, which is useful for compare and aligning of RS image and DEM data when data fusion is concerned. The dissertation provides a new algorithm which employs cellular automata to search the contour of water. With multiscale iterativing the algorithm aperently improved the efficiency of image skelecton extraction.5. A forestry fire spreading expert system model with cellular automata as inference engine which is driven by the knowledge is finally presented.This research is supported by the National Natural Science Foundation of China.更多还原