<正>介绍了图像融合的框架层次结构，以及像素层、特征层和决策层3层图像融合的方法及其相互关系。分析了图像融合平台的设计与实现方法，选择基于DS证据理论和模糊Kohonen神经网络聚类算法，进行了适当改进，并加以验证。结果表明，模糊Kohonen神经网络聚类算法的聚类精度和聚类速度都要优干传统算法。

According to the different plant communities, the arbor, shrub, herb and liana plant layers were distinguished. Main plant community species diversity was studied along Taihu roadside in Wuxi Mashan National Tourist Holiday Resort, by using species richness index, Shannon-Wiener index, Simpson index and Pielou evenness index. The results showed as follows: (1) The four species diversity indices of the arbor layer and the shrub layer are interrelated. However the herb layer is different. (2)By artificial vegetation renovating, plant species diversity along the road has been more remarkably increased than non-restored spot, so that the roadsides have been protected and landscape improved. (3)The plant community in restored lake shoal has high community rate of coverage, but designed plant variety is single, so that it’s species diversity is less than the non-restored. (4)Original plant community species diversity was most above average. Reasonable artificial restoration can increase biodiversity for original destroyed vegetation. (5)By several calculating methods of species diversity, the species diversity in an area can be objectively reflected.