TOWARDS OPERATIVE FOREST INVENTORY BY EXTRACTION OF TREE LEVEL INFORMATION FROM VHR SATELLITE IMAGES Heikki Astola 1 , Heikki Ahola 1 , Kaj Andersson 1 , Tuomas Häme 1 , Jorma Kilpi 1 , Matthieu Molinier 1 , Yrjö Rauste 1 , Jussi Rasinmäki 2 1 VTT Technical Research Centre of Finland, Vuorimiehentie 3, Espoo, P.O. Box 1000, FI-02044 VTT, Finland, Email:firstname.lastname@vtt.fi 2 Simosol Oy, Asema-aukio 2, FI-11130 Riihimäki, Finland, Email:jussi.rasinmaki@simosol.fi ABSTRACT The work related to this paper is part of an on-going study called NewForest - Renewal of Forest Resource Mapping. In this study the methodologies developed for individual tree crown (ITC) recognition and crown width estimation will be combined with forest variable estimates that are produced using features calculated from segmented VHR satellite image. A field visit to Karttula, Eastern Finland, was conducted to collect the class and geo-location information for 1164 ground objects (900 trees and 264 non-tree objects). These data were used for the classifier model and feature selection and for species classification accuracy assessment. For testing the classifier ability to predict tree species proportions, an independent set of 178 forest field inventory plots was used. Seven classes were defined: pine, spruce, deciduous, shadow, open area, bare ground, green vegetation. A modified Local maximum (LM) filtering technique was used for individual tree crown (ITC) detection. The spectral signatures of an ITC were sampled with a radius of r=1.5 m around the ITC brightest pixel (feature set A). Also a set of 9 contextual features were extracted from circular neighbourhood (r=7.25 m) around the ITC (feature set B). A classifier model and feature selection was performed. A 5NN classifier provided the best overall performance in tree species classification in terms of classification accuracy and generalization. The overall classification accuracy for the seven classes was 73.8% with feature set A using 5NN classifier. With feature sets A and B combined the accuracy was 74.1%. The average RMS errors in species proportion prediction were 2.6% with feature set A and 2.5% with feature sets A and B combined. Key words: forestry; tree species classification; individual tree crown, ITC; GeoEye; VHR 1 INTRODUCTION The availability of VHR optical satellite data is constantly increasing with the emergence of new imagers. This facilitates the development of remotely sensed forest inventory methods with improved accuracy. The new methods have also potential to compete with the price paid per hectare when compared with the methods that are presently used, usually requiring costly field work. One of the biggest challenges for remote sensing is to develop data analysis methods that increase the species-wise forest variable estimation accuracy to meet the requirements of operative forest management planning and wood procurement industry. Today, space-borne imagery is available with a resolution comparable to the aerial photographs that are used in forest management planning. The latest advances in the interpretation of the very high resolution (VHR) satellite data and airborne laser scanner (ALS) data can give information at the tree level, which can be utilized in improving the estimation of forest variables. The algorithms and results of research have to be further transformed into a complete chain for the provision of all the key forest variables. The authors studied the separation of Scots pine (Pinus sylvestris L.) and Norway spruce (Picea abies Karst.) in mature single-species stands from Ikonos-2 images in boreal forest conditions (Astola et al. 2006), and the study was further extended to include also broadleaved class and mixed species stands. The obtained results indicated that the developed method could detect trees, identify their species and determine the species proportions in mixed forest. In the project NewForest – Renewal of Forest Resource Mapping the authors will integrate the developed single tree detection and species classification method to existing forest variable estimation method (Häme et al. 2001) expecting an improvement in the present species- wise forest variable estimates.