Classification of single trees

Classification of single trees using VHR satellite imagery and lidar

Background and Objectives

Information on tree-species is fundamental for many silvicultural and forest economical calculations. This project will investigate the potential synergistic use of VHR satellite imagery and airborne lidar data for tree-species classification of single trees. This represents an alternative approach to using airborne imagery with lidar since the aerial imagery has been shown to constrain the process in several important ways. For example, aerial images typically show significant radiometric and geometric variations from nadir to scene edge. Moreover, image capture usually occurs simultaneously with lidar surveys. Since these often take many days or weeks to complete, the aerial imagery show considerable variations in lighting, which significantly complicates the analysis.

The project aims to address the shortcomings of aerial imagery by investigating very high-resolution satellite imagery as an alternative image source. Such imagery is both more homogeneous and operationally decoupled from the lidar survey. As a consequence it is hoped to develop a methodology for large-scale provision of single-tree classification on an operational basis
that will lead to significantly cheaper single-tree survey products and hence a commercially more attractive offering. 

Approach / Implementation

The overall approach is to make synergistic use of VHR imagery and lidar data for specie- classification of individual trees. The principal geometric challenge is to accurately source the radiometric values for each tree crown from within the satellite imagery. Established methods for tree-species classification in multispectral imagery will then be used to distinguish the various classes of tree, in this project the following classes will be used; pine, spruce and broadleaf trees.

The research and development work is divided in three main phases: i) Geometrical co-registration of satellite imagery and lidar data and assessment of geometrical accuracy using lidar and field data, ii) association of radiometric information from one or more satellite images to tree individuals and iii) species classification of individual trees identified in lidar data with the help of VHR satellite imagery and assessment of classification result using field data. The work in theses phases will leverage on existing tools at each of the partner organisations. The co-registration and orthorectification development work will be implemented using the Spacemetric Keystone platform. This ensures the highest accuracy in processing and makes available a framework using rigorous photogrammetric methods that is necessary for the image geometrical challenges in the project. Single Tree data will be obtained by the use of FORAN Remote Sensing in-house developed lidar data analysis platform. This system is designed to handle very large datsets and automatically identifies (dominant) individual trees and provides estimates on several important tree features, e.g. position, height and crown area. 

Expected results

The expected results of the project are to:
Knowledge and experience on with what geometric accuracy the satellite imagery and lidar data can be geometrically co-  registered
Knowledge and experience on with what accuracy a single tree can be classified using radiometric information from an co-registred VHR satellite imagery
Knowledge on to what degree the workflow can be automated. 


Dr. Ulf Söderman, FORAN Remote Sensing AB (, phone: +46 (0)73 552 1046

Ian Spence, Spacemetric AB (, phone: +46 8 594 770 83, Email:

Senast uppdaterad: 20 maj 2009