论文部分内容阅读
通过元胞自动机(CA)模拟和重建城市演化的复杂非线性过程,对于城市土地利用规划和决策具有指导意义。利用传统线性方法获取的地理CA转换规则,较难刻画城市演化的时空动力学过程。基于核主成分分析方法(KPCA),通过核函数映射,在高维特征空间下不仅能够对多重共线的空间变量进行非线性降维,且由此建立的地理元胞模型KPCA-CA参数物理意义明确,能够较好地体现城市化过程的非线性本质。基于GIS环境下自主研发的地理模拟框架SimUrban,利用该KPCA-CA模型模拟和重建了快速城市化区域上海市嘉定区1989-2006年城市演化过程,并预测了研究区2010年的城市空间格局。模拟结果显示,嘉定区城市主要沿中心区域及主干道路而扩展,体现了KPCA方法提取的前两个主成分的作用,与城市实际发展情况相符。利用混淆矩阵和面积控制精度等指标,对模拟结果进行了评价,得到总体精度为80.67%、Kappa系数为61.02%,表明模拟结果与遥感分类结果及统计结果符合程度较好;与传统基于线性方法的地理CA模型比较,KPCA-CA模型模拟结果精度更高。
It is instructive for urban land use planning and decision-making to simulate and reconstruct the complex non-linear process of urban evolution through cellular automata (CA). It is difficult to characterize the spatio-temporal dynamic process of urban evolution by using the traditional linear method to obtain the geographic CA conversion rules. Based on kernel principal component analysis (KPCA), KPCA-CA parametric physics can not only reduce nonlinearly the multi-collinear spatial variables through kernel function mapping in high-dimensional feature space, The meaning is clear, which can better reflect the non-linear nature of the urbanization process. Based on SimUrban, a geographical simulation framework developed independently in GIS environment, the KPCA-CA model was used to simulate and reconstruct the urbanization process of Jiading District in Shanghai from 1989 to 2006 in the rapid urbanization area. The urban spatial pattern in 2010 was also predicted. The simulation results show that the city of Jiading mainly extends along the central area and trunk roads, which reflects the role of the first two principal components extracted by the KPCA method, which is consistent with the actual development of the city. Using the confusion matrix and area control accuracy, the simulation results were evaluated. The overall accuracy was 80.67% and the Kappa coefficient was 61.02%, indicating that the simulation results were in good agreement with the remote sensing classification results and statistical results. Compared with the traditional linear method Compared with the geographic CA model, the KPCA-CA model simulation result is more accurate.