高分辨率遥感影像5种面向对象分类方法对比研究

发布时间：2018-05-23 07:35

本文选题：面向对象分类 + SVM分类　；参考：《测绘通报》2017年11期

【摘要】：针对主流的面向对象分类方法在遥感影像处理中的使用范围不明确的问题,以e-Cognition软件平台为基础,处理标准数据集,综合考虑视觉效果、总体精度和用户精度3方面,系统地比较分析了主流的面向对象分类方法在高分辨率影像中的分类效果和精度分析。试验结果表明:使用不同的分类方法均存在混分现象且混分对象不完全一样。在处理同一标准数据集时,隶属度函数分类方法的精度最高但分类速度最慢,Bayes的分类效果最差但操作简单,支持向量机(SVM)、决策树(DT)、随机森林(RF)的分类速度均较快且都有较高的精度,其中SVM分类方法在区分相似性高的对象方面具有明显优势。在选择高分影像分类方法时,要充分考虑分类影像的特征选择从而选择合适的分类方法。
[Abstract]:Aiming at the problem that the application scope of the mainstream object oriented classification method in remote sensing image processing is not clear, based on the e-Cognition software platform, the standard data set is processed, the visual effect, the overall accuracy and the user accuracy are considered, and the mainstream object oriented classification method in the high resolution image is systematically compared and analyzed. The result of classification results and accuracy analysis. The result of the experiment shows that there is a mixture phenomenon and the mixed object is not exactly the same. When dealing with the same standard data set, the membership degree function classification method has the highest precision but the slowest classification speed, the Bayes classification effect is worst but the operation is simple, the support vector machine (SVM), decision tree (DT) The classification speed of random forest (RF) is relatively fast and has high precision, and the SVM classification method has obvious advantages in distinguishing objects with high similarity. When selecting high score image classification methods, we should take full consideration of the feature selection of classified images so as to choose the appropriate classification method.
【作者单位】：江苏师范大学地理测绘与城乡规划学院;中国测绘科学研究院;
【基金】：国家自然科学青年基金(41401397) 江苏省自然基金青年项目(BK20140237)
【分类号】：P237

【相似文献】