Object-Oriented Automatic Identification of Forest Gaps Using Digital Orthophoto Maps and LiDAR Data

Abstract

Identification of forest gaps is a prerequisite for quantification of their size, shape, and dynamics, and for clarification of both complex structural forest species regeneration and understory species diversity. Although airborne LiDAR and digital orthophoto maps (DOM) have been used separately to identify forest gaps, few studies have considered integration of the two data sources for forest gap segmentation and classification. True color DOM (20 cm) and airborne LiDAR (3.7 points/m2) data were used to study object-oriented gap identification in the typical natural secondary forest of the Maoershan Experimental Forest Farm (China). Three segmentation schemes based on DOM only data, LiDAR data, and integrated DOM & LiDAR were adopted when processing the object-oriented classification. Based on the segmentation results, the support vector machine classifier was used with DOM spectral features, LiDAR height features, and integrated features from both data sources to identify forest gaps. The Modified Euclidean Distance 3 (ED3Modified) index was selected to assess segmentation quality. Comparison of the three segmentation schemes revealed that segmentation based on LiDAR was the best and the classification accuracy using integrated spectral and height features was the highest (OA = 87%, Kappa = 0.81). Those results could provide technical support for the quantitative analysis of forest gap features.