论文部分内容阅读
针对单源遥感数据分类精度不高的问题,提出一种基于多特征融合的面向对象分类方法。该方法利用LiDAR点云数据的高程信息,并融合地物粗糙度特征,以及航空影像的地物光谱、形状和上下文信息等多种特征,再基于SVM分类器构建面向对象的分类方法,以提高城区环境下遥感数据分类的可靠性。试验表明,该方法可有效地提高城区地物的分类精度,且分类结果更符合人的视觉认知规律。
Aiming at the problem of single-source remote sensing data classification accuracy is not high, an object-oriented classification method based on multi-feature fusion is proposed. The method utilizes the elevation information of LiDAR point cloud data and combines the features of roughness of features with the features of spectrum, shape and context information of aerial images, and then constructs object-oriented classification method based on SVM classifier so as to improve Reliability of Remote Sensing Data Classification in Urban Environment. Experiments show that this method can effectively improve the classification accuracy of urban landscape objects, and the classification results more in line with human visual cognitive rules.