基于支持向量机的遥感影像分类比较研究

支持向量机是建立在统计学习理论基础上的一种新的人工智能算法,较好地克服了传统分类方法中存在的小样本、非线性、过学习、高维数、局部极小点等问题,是一种极具潜力的遥感影像分类算法。本研究采用Landsat-5的TM影像,用支持向量分类法对影像进行分类,分析了支持向量机不同参数组合情况下的分类精度,并对支持向量分类法与传统分类方法进行了比较,发现支持向量分类算法具有参数选择范围宽,不要求对待分类区域地物光谱特征和影像分布特征具有先验知识,分类精度高等特点,对于在没有现场同步实测数据的区域进行精确的分类具有特别重要的价值。

Comparative Study on Classification of Remote Sensing Image by Support Vector Machine

Support vector machine(SVM) is an artificial intelligent algorithm based on theory of statistics learning.It is a promising classification algorithm and can overcome the limitation of traditional classification algorithm such as small data set,nonlinear,overfitting,high dimension and local minimum.In this paper,the TM image of Landsat-5 is used for classification by the method of support vector machine.The results and precisions of classification are compared between the different parameter combinations.Further more,precisions are compared between the SVM and traditional algorithm.The results indicate that SVM classification algorithm has the advantage of broad parameters range without prior knowledge of image and samples.The precision of SVM algorithm is much higher than traditional algorithm adapting to the area without in situ measurement.