Individual tree recognition based on a synthetic tree crown image model

Individual tree crowns have been automatically recognized in monocular high -spatial resolution optical images of scenes containing boreal or cool temp erate forests in a leaved state. The procedure is based on a model of the i mage formation process at the scale of an individual tree, user-generated t raining data, and the exploitation of a basic constraint on the spatial rel ationship of neighbouring trees. The procedure does not rely on the existence of explicit tree crown image r egion boundaries or on a large amount of contrast between the tree crown im age region and the surrounding image region. The procedure has been tested on image data acquired with the MEIS II and CASI sensors for scenes in Onta rio and Alberta. The ground-projected IFOV that the data were acquired with is 36cm (MEIS II) and 60cm (CASI). The tests results show that the procedure is largely able to accommodate th e variation represented by the scenes, but that significant errors are rela ted to irregularities in tree crown form and irradiance that are caused by tight vertical and horizontal spacing of the crowns (relative to the spatia l resolution of the image data), and to the inability to resolve trees with crown diameters of approximately Im or less.