Linear versus single photon lidar height loss-specific change detection with subsequent object-based supervised classification of timber clear cuts and other disturbance based on post-change imagery
Graduate Student University of Minnesota - Natural Resources Science and Management Minneapolis, Minnesota, United States
Examined bi-temporal lidar datasets from different sensors to detect loss-changes in canopy height models in the Arrowhead region of Minnesota. The different types of lidar provided similar results, but are unreliable in short-term change detection. The subsequent loss-object-based, machine learning classifier performed well in classifying clear cuts versus other disturbances.
Learning Objectives:
Connecting the impacts of anthropogenic disturbance on forest ecology and the effectiveness of machine learning algorithms for detecting and classifying different types of disturbance in actively managed forest landscapes.
Understand the limitations and reliability of using different types of lidar data for detecting short-term changes in canopy height models and the importance of robust training data when using a support vector machine for forest disturbance detection and classification.
