高分辨率遥感影像5种面向对象分类方法对比研究

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。

Five Object-oriented Classification Methods Analysis Based on High-resolution Remote Sensing Image

In view of the mainstream object-based classification method in remote sensing image processing using range is not clear, based on e-Cognition software, the standard data set is processed. Comprehensively considering the visual effects, overall accuracy and user accuracy, classification results and precision analysis of mainstream object-based classification are systematic analysis in the high resolution image. The experimental results show that there are mixed phenomena using different classification methods and the mixed objects are not exactly the same. In dealing with the same standard data set, the membership function has the highest accuracy but the slowest speed. The classification effect of Bayes is the worst, but the operation is simple. The classification speed of SVM, RF, DT are faster and have higher accuracy. Meanwhile SVM has obvious advantages in distinguishing objects with high similarity. In the selection of high resolution image classification method, we should fully consider the feature selection of classified images to select the appropriate classification method.