一种基于改进双树复小波变换的地形纹理特征提取方法  

A New Method on Terrain Texture Characteristics Extraction Based on Improved Dual-tree Complex Wavelet Transform

作　　者：陶旸[1] 王春[2] 蒋圣 TAO Yang;WANG Chun;JIANG Sheng(Geomatias Center of Jiangsu, Nanjing 210013;Chuzhou University ,Chuzhou239000;Eazytec of Jiangsu ,Yixing 214200 ,China)

机构地区：[1]江苏省基础地理信息中心,江苏南京210013 [2]滁州学院地理信息与旅游学院,安徽滁州239000 [3]江苏卓易信息科技股份有限公司,江苏宜兴214200

出　　处：《地理与地理信息科学》2017年第4期47-50,共4页Geography and Geo-Information Science

基　　金：国家自然科学基金项目(41201415;41571398)

摘　　要：地形纹理特征提取和识别是地形自动分类的关键。该文在分析已有地形纹理特征提取方法的基础上,提出一种基于双树复小波变换的地形纹理特征提取和识别方法。该方法将DEM地形垂直特征信息引入双树复小波变换模型,设计并实现了面向DEM地形数据特点的地形纹理特征提取方法。以陕北黄土丘陵沟壑区为实验样区,选取6个不同地貌特征共109个典型地形纹理样本进行特征识别实验,并将实验结果与常用的灰度共生矩阵(GLCM)模型及Gabor滤波结合PCA方法进行对比。研究显示,基于地形垂直特征改进的双树复小波变换模型对6类样本有较高的识别精度,同时消耗最短的纹理特征构建时间。The extraction of terrain texture characteristics is the key step of DEM based terrain classification.A new terrain tex-ture characteristics ext raction method is presented based on improved dua-tree complex w^avelet transform.This new method tries to introduce traditional image texture algorithms into terrain texture analysis.At first,set1/4as the sample period of Q-shift DT-CWT,then the texture characteristics and the distance between different texture characteristics with mean and variance value of complex wavelet^efficients were calculated.In order to improve the accuracy of terrain classification,vertical charac-teristic information of terrain surfacewas introduced in dua-tree complex wavelet transform model.Meanw^hile,109test samples in hilly and gully region of Loess Plateau were selected for terrain texture recognition.Experiments show that vertical character-istic information of terrain surface could improve the rectignition accuracy of DEM based terrain test areas.Also,compared with Gray Level Co-occurrence Matrices(GLCM)model and Gabor filtering with PCA,the improved dual-tree complex wavelet transform method presented in this paper could effectively identify six typical test areas with different texture patterns.The a\^erage recognition accuracy of improved dua-tree complex wavelet transform method can reach up to6082%with the least time consumption of texture characteristics database construction.

关 键 词：双树复小波变换 DEM 纹理分析 纹理识别 

分 类 号：P208[天文地球—地图制图学与地理信息工程] TP391.41[天文地球—测绘科学与技术]