【摘要】 面向对象分类方法能够解决基于像素分类方法带来的"椒盐噪声"缺点,并能利用分割单元构建对象特征空间,从而提高分类精度。常规的面向对象分类通常会设定经验或最优分割参数,在此基础上进行面向对象遥感影像分类。然而,不同地物具有不同最优分割参数,这样会导致地物分类精度不佳。因此,采用分割参数分级化思想,使用大、中、小尺度进行分阶段分割,首先对水域和浓密植被进行分类,然后再对其他地物进行细分,能够有效提高分类精度。通过对南京市Landsat-8卫星的OLI影像进行实验,试验证明,本方法在精度和分类效率上具有一定优势,在实际工作中可以提供借鉴。更多还原

【Abstract】 Object-oriented classification method can solve the problem brought by "salt and pepper noise" shortcomings based on pixel classification method,and can use the segmentation unit to construct the object feature space,so as to improve the classification accuracy.Conventional object-oriented classification usually sets the experience or the optimal segmentation parameter,and then carries on the object-oriented remote sensing image classification.However,different objects have different optimal segmentation parameters,which can lead to poor classification accuracy.So use the segmentation stage,medium and small scale,the waters and dense vegetation classification,and then starting the other object classification,that can effectively improve the classification accuracy.Based on the experiments of Landsat-8satellite images of Nanjing City,the method has advantages in accuracy and classification efficiency.更多还原