基于高分辨率遥感影像的输电线走廊场景分类

Scene classification of power transmission line corridor using high resolution remote sensing imagery

提出了一种将纹理特征和颜色特征相结合的输电线走廊遥感图像分类方法.该方法首先采用简单线性迭代聚类(SLIC)过分割技术将一幅大场景遥感图像分割为若干个尺寸大致规则的超像素块,然后对这些超像素块进行联合散射纹理特征和颜色词袋(BOC)特征提取,接着将这两种特征级联融合,最后将组合后的特征输入到直方图交叉核支持向量机(HIK-SVM)中训练分类器并进行场景分类.武汉地区输电线走廊场景的高分辨率遥感影像分类实验结果表明,与仅利用单个特征相比,两种互补特征的组合具有更高的分类准确率,可获得更为满意的场景分类结果.

Classification of power transmission line corridor using high resolution remote sensing images by combining texture and color features is investigated.A simple linear iterative clustering algorithm is adopted to over-segment the large-scale remote scene into nearly uniform superpixels.Then texture（combined scattering）and color（bag of colors）features are extracted to characterize each superpixel.Next the derived two features are concatenated and fed into a support vector machine with histogram intersection kernel（HIK-SVM）to train classifier and conduct scene classification task.Experimental results of power transmission line corridor scene in Wuhan show that the combination of two features obtains more satisfactory classification map and higher accuracy than the single feature channels.