Abstract

The accurate extraction of Built-Up Areas (BUAs) is of great significance for analysing urban spatial evolution patterns. In this study, we proposed a new method for high-precision extraction of BUA based on multi-source remote sensing data. Firstly, Built-Up Area Extraction Index (BAEI) was used to preliminarily identify the BUA based on Landsat 8 imagery. Secondly, the Support Vector Machine (SVM) algorithm was used for improving the extraction precision of BUA, whose selected training sample was established on the Nighttime Light (NTL) data. Then, images fusion and continuity correction were carried out. Finally, the Neighbourhood Statistics Analysis (NSA) was used to adjust and remove the part of the non-urban centre which was misjudged as the BUA. Our results show that this method has better performance on both overall accuracy and Kappa coefficient compared with other classic methods, which provides empirical reference for understanding law of land expansion and rational land planning.

摘要

建成区（BUA）的准确提取对于分析城市空间演化格局具有重要意义。在本研究中，我们提出了一种基于多源遥感数据的高精度BUA提取方法。首先，利用建筑面积提取指数（BAEI）基于Landsat 8影像初步识别BUA。其次，采用支持向量机（SVM）算法提高BUA的提取精度，其选择的训练样本建立在夜间灯光（NTL）数据上。然后进行图像融合和连续性校正。最后，通过邻域统计分析（NSA）对误判为BUA的非城市中心部分进行调整和剔除。