Bank Line Extraction by Integration of Orthoimages and Lidar Digital Elevation Model Using Principal Component Analysis and Alpha Matting

Riverbank lines change over time, causing loss of land and property. Accurate mapping of riverbank lines is essential for restoration and preservation. This paper presents a method to map riverbank lines by combining georeferenced orthoimages and lidar digital elevation model (DEM). This method uses the properties that lidar can provide elevations under trees and open water edges are visible in orthoimages to extract the planimetric locations of bank lines. The orthoimage pixels with less than 0.15% slope on the DEM were replaced by water pixels. Principal component analysis (PCA) was conducted using DEM, slope, and orthoimage bands. Training data of river body and the background were identified manually on the first three component images. An alpha matting–based method was implemented using the training data to extract the bank lines. Bankline using α value of 50% probability were statistically and visually better when compared to the manual bank lines.