科研之友
微信
新浪微博
Facebook
分享链接
摘要
Tidal creeks (TCs) are transitional waterways between terrestrial and marine environments. Extracting geometric information for tidal creek networks (TCN) geometry from remote sensing is essential to understanding their characteristics, formation, and evolution. Currently, the major obstacles to automated recognition using digital elevation models (DEMs) derived from airborne light detection and ranging (LiDAR) data are low relief, varying widths, high density, strong anisotropy, and complicated patterns. Conventional methods, such as the optimal-elevation threshold method, the optimal-curvature threshold method, and the D8 method, cannot achieve satisfactory performance under these conditions. We propose an automated method for extracting tidal creeks (AMETC) using topographic features detected from LiDAR. Specifically, a multi-window median neighborhood analysis was designed to enhance depressions both in mudflat and marsh environments; a multi-scale and multi-directional Gaussian-matched filtering method was incorporated to enhance width-variant TCs; and a two-stage adaptive thresholding algorithm was implemented to segment low-contrast TCs. The AMETC was tested on two large LiDAR datasets of the Jiangsu coast with different resolutions. The quantitative assessments show that AMETC successfully extracted both small and large TCs from our study areas. The true positive extraction rate reached 95%, outperforming conventional methods. The AMETC is robust and weakly dependent on scale, and rarely requires manual intervention. Further applications suggests that the AMETC has potential for extraction of other types of channel features (e.g., badland networks and ravines).
- 出版日期2015-8
- 单位南京大学
