分析了基本粒子群算法(PSO)、 混合粒子群优化算法(HPSO)和模糊C-均值算法(FCM)的特点, 将模糊C-均值算法引入到混合粒子群优化算法中, 发展和改进了HPSO-FCM算法, 并在Fortran语言和MATLAB环境下开发实现HPSO-FCM程序. 以2009年6月份的环境一号卫星多光谱可见光图像和ENVISAT的ASAR微波图像为基础数据, 通过波段叠加和主成分分析, 得到前3个主成分合成图像. 利用HPSO-FCM算法和非监督学习动态聚类算法(ISODATA)分别对湖南东洞庭湖3个主成分合成图像, 进行湿地分类实验. 结果表明：(1)将模糊C-均值算法引入到混合粒子群优化算法中, 具有较好的搜索速度和收敛精度, 能有效寻找和优化最佳聚类中心. (2)HPSO-FCM算法在多光谱遥感图像湿地分类精度比较高, 是一种有效的遥感图像分类方法.

The present paper analyzed the characteristics of particle swarm optimization(PSO), hybrid particle swarm optimization (HPSO) and fuzzy C-means (FCM), imported FCM into HPSO, and improved the HPSO-FCM arithmetic. An HPSO-FCM program was developed using Fortran language in MATLAB. Besides, a synthesis image combined with the former three principal components was obtained through band stacking and principal component analysis, taking the multispectral visible image of HJ-1 Satellite shot in June 2009 and the ASAR radar image of ENVISAT as basic data. And the paper has done a wetlands classification experiment in the synthesis image of the East Dongting Lake of Hunan province, using HPSO-FCM arithmetic and ISODATA separately. The results indicated: (1) The arithmetic which imported crossover operator of genetic algorithms and FCM into HPSO had better search speed and convergent precision, and it could search and optimize the best cluster center more efficiently. (2) The HPSO-FCM arithmetic has better precision in wetlands classification in multispectral remote sensing image, and it is an effective method in remote sensing image classification.