基于WWSVDD多分类的遥感图像病害松树识别  

Infected Pine Recognition in Remote Sensing Images Using WWSVDD Multi-Classification

作　　者：胡根生[1] 张学敏[2] 梁栋[1] 

机构地区：[1]安徽大学计算智能与信号处理教育部重点实验室,合肥230601 [2]安徽大学电子信息工程学院,合肥230601

出　　处：《北京邮电大学学报》2014年第2期23-27,共5页Journal of Beijing University of Posts and Telecommunications

基　　金：国家自然科学基金项目(61172127);安徽省自然科学基金项目(1408085MF121)

摘　　要：针对支持向量数据描述多分类中模糊数据域的误判问题,提出了一种改进的加权小波支持向量数据描述(WWSVDD)多分类方法,并应用于遥感图像病害松树识别.利用无人机搭载双光谱相机获取高分辨率遥感图像,提取地物特征,构建特征向量.用WWSVDD模型描述每类样本,根据待测样本在特征空间中的不同分布,分别采用最小相对距离法和隶属度函数法进行决策分类,从而实现病害松树的识别.实验结果表明,与传统的K近邻和支持向量数据描述多分类方法相比,所提方法在识别病害松树方面准确性更高.For the problem of misjudgment in the fuzzy data domain,a modified multi-classification algorithm of weighted wavelet support vector data description（ WWSVDD） is proposed and applied to the infected pine recognition in remote sensing images. Firstly,utilizing the high resolution images acquired by the double spectrum camera fixed on the unmanned aerial vehicle,the features of ground objects are extracted to construct the corresponding feature vectors. Secondly,each kind of ground objects samples is described by WWSVDD model. Finally,according to the distribution of test samples in the feature space, the methods of minimum relative distance and membership function are respectively used to decide the labels of the samples,so that the infected pines are recognized ultimately. The experiment results show that the proposed method can recognize the infected pines more effectively than the traditional multi-classification methods of K-nearest neighbor（ KNN） and support vector data description（ SVDD）.

关 键 词：无人机 遥感图像 病害松树识别 加权小波支持向量数据描述 多分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]