修正的归一化差分建成区指数及建成区提取后处理方法研究

Abstract

及时、有效地获取城市中的建成区信息对进行合理的城市规划和管理有十分重要的意义。遥感数据具有来源广、覆盖范围大、实时采集的特点，包含丰富的地面信息，是进行城市信息采集和监测的有效手段。针对现有的利用遥感影像进行建成区提取中存在的裸土和建成区混分问题，根据建成区的光谱特征，本文提出了一种基于Landsat影像的修正的归一化差分建成区指数(Modified Normalized Difference Built-up Index，MNDBI)，并针对建成区的指数提取结果提出了相应的后处理框架，主要包括聚类去裸土和形态学偏重建两步。后处理中，第一步可以去除建成区提取结果中混杂的裸土，提高建成区提取的正确率，第二部可以减少提取结果中建成区的漏分，降低漏检率。实验结果显示，MNDBI相对于现有的建成区指数，可以更好的区分裸土和建成区，提高建成区的指数提取精度。指数提取后处理步骤可以有效的降低建成区指数提取的漏检率，减少虚警，进一步提高提取精度。 As remotely sensed data can provide consistent and substantially information of the earth, it has become a common tool to explore the land surface timely and effectively. Specifically, it is widely applied in monitoring the built-up surfaces in urban area, which is significant for urban planning and management. To alleviate the confusion between built-up area and bare soil which is one of the major difficulties in the extraction of built-up area, this article proposes a novel Modified Normalized Difference Built-up Index (MNDBI). Moreover, a two-step post processing scheme, including clustering and morphological partial reconstruction, is also presented follow the built-up extraction. During the two-step post processing, missed built-up pixels can be filled up and the rest bare soil pixels can also be excluded. The proposed MNDBI index demonstrates better results and accuracies comparing to the existing built-up indices in our experiments. In addition, the two-step post processing can further alleviate false alarm and missed alarm and improve the extraction accuracy of built-up area.