Canadian Forest Service Publications

A crown-following approach to the automatic delineation of individual tree crowns in high spatial resolution aerial images. 1995. Gougeon, F.A. Canadian Journal of Remote Sensing 21(3): 274-284.

Year: 1995

Issued by: Pacific Forestry Centre

Catalog ID: 4583

Language: English

Availability: PDF (download)

Mark record


The availability of quality georeferenced high spatial resolution aerial multispectral digital images, such as those obtained with the MEIS-II, could lead to new ways of obtaining forest management inventories. The potential for computer-assisted on-screen human interpretation of colour-enhanced images is a good example of the possible benefits of such technology. A fuller use of the digital nature of this type of data is also possible by getting precise forest inventories in an almost completely automatic way. To reach this goal, while still quantifying adequately the species composition of forest stands, it is believed that individual tree crown delineation followed by crown species identification and regrouping is required. This paper describes the first step in this endeavour, an approach to the automatic delineation of tree crowns, which can lead to detailed crown outlines and accurate crown counts. The approach consists of first isolating crowns from the background vegetation and from each other using an image processing program that follows valleys of shaded material between crowns. These crowns are subsequently delineated more precisely and further separated using a rule-based program. Tested on a MEIS-II image of coniferous plantations at 31 cm/pixel, this method led to crown counts that are within 7.7% of ground counts, compared to 18.1% using photointerpretation. Examination of the errors of omission and commission shows that, in general, 81% of the crowns are the same as those obtained by the visual interpretation of the imagery.