基于径向基函数神经网络的植烟土地适宜性评价

Suitability Assessment of Tobacco-Planted Land Based on RBF (Radial Basis Function) Neural Network

在Matlab7.0环境下,根据人工神经网络的理论和方法,以重庆市彭水县植烟土地的实测数据及评价标准构建径向基函数神经网络模型,并进行模型训练及样本评价;在ArcGIS技术支持下,进行不同尺度土地适宜性评价及精度检验.结果表明:采用最近邻聚类学习算法选取聚类中心,模型具有较强非线性处理能力和逼近能力,并具有学习时间短,网络运算速度快,性能稳定等优点;通过模型评价结果和检验值的验证,发现用径向基函数神经网络模型评价土地适宜性,具有评价精度高,使用方便,适应性强等优点,因此可望将其用于区域土地资源生态环境分类评价研究.

Taking into account the non-linear relationship between assessment indicators and assessment grades, the experimental results showed a high accuracy, thus demonstrating that this method could be widely extended. In this paper, a Radial Basis Function neural network model （RBF） was established for evaluating land suitability in Pengshui county of Chongqing based on the theory and methodology of neural network with Matlab 7.0 and the data obtained from the tobacco-planted land and criteria of suitability e- valuation. Land suitability evaluation and its accuracy tests were performed at different scales based on ArcGIS. The results were as follows. The model proved its capability to approach function and treat nonlinear problem by using nearest neighbor cluster algorithm to select the clustering center. The main advan- tage of the RBF-based model was its accuracy, time saving, fast running and stability behavior. The good agreement between the assessed and the tested data was observed. Because of these benefits, it is believed that the model could find application in classifying and assessing regional land resources.