针对山区样本数据有限的情况,提出一种支持小样本数据处理和高抗噪声能力的模糊最小二乘支持向量机(FLS-SVM)方法。采用基于混沌粒子群(CPSO)的向量机模型参数优化方法,避免了人为参数选择的不确定性,并考虑山区地形起伏的影响,引入地形改正量构建支持向量机训练模型。试验结果表明,考虑地形起伏在内的基于CPSO优化的FLS-SVM模型与传统拟合算法(如PSO-LSSVM和GA-LSSVM)相比,能够利用有限样本获得更高精度的拟合效果,适用于SRTM正高数据获得正常高的应用。 Aiming at the limited sample data in mountain area, a fuzzy least square support vector machine (FLS-SVM) method is proposed to support small sample data processing and high noise immunity. The chaos particle swarm optimization (CPSO) -based vector machine (SVM) model parameter optimization method is adopted to avoid the uncertainty of human parameter selection. Considering the impact of mountainous terrain fluctuation, the terrain correction is introduced to construct SVM training model. The experimental results show that the FLS-SVM model based on CPSO optimization, which considers the relief of topography, can obtain more accurate fitting results by using the finite samples than the traditional fitting algorithms (such as PSO-LSSVM and GA-LSSVM) Normal high data is obtained for SRTM positive applications.