基于元胞自动机的遗传神经网络在土地利用变化模拟分析中的应用

元胞自动机模型在土地扩展的转换规则设计上具有随机性,受周围环境影响较大。文中建立基于BP神经网络和遗传神经网络算法优化的元胞自动机土地扩张模型,对广州市2009—2011年进行城市扩张模拟分析。实验结果显示:BP神经网络能够较好地模拟分布较集中的耕地和林地等区域,精度可达到70%以上,而对于面积较零碎的建筑用地区域,模拟效果较差;而遗传神经网络优化算法能够总体提高模拟精度约5%,部分精度能提高至20%。同时,该算法还能充分考虑影响土地变化的各种扰动因素,优化选择驱动因子和缩短迭代次数,对于城市土地扩张研究具有可行性。

Genetic neural network based on cellular automata applied to the simulation analysis of land use change

Cellular automata model comes up with the random in the design of transit mechanism for land expasion,which is affected by the surrounding environment.The BP neural network （BP） and the genetic algorithm （GA） combined with the cellular automata （CA） are used as a model into the land expansion simulation in order to analyze the case in Guangzhou from 2009 to 2011.Experimental results show that the BP neural network is suitable for the simulation of cultivated land and forest land,the precision of which can reach up to 70％,but can not be suitable for the simulation of the urban building areas of fragmentary.The BP genetic neural network can be improved with simulation precision in general abby 5％,and part of the precisions can go up to 20％.In addition,the GA-BP model can not only better choose the factors that influence the urban expansion,but also shorten the number of iterations to improve the processing speed.To sum up,it is feasible and effective to apply the genetic neural network to the predicting of land use change.