Caracterisation de la densite (tiges/ha.) et de la defoliation d'un territoire forestier boreal a l'aide de la videographie aerienne multispectrale (VAM)

摘要

We use the multipsectral potential of the UQAC videographic sensor to acquire high resolution imagery (30cm/pixel) of a forest territory (86 km~2) exploited by Donohue Saint-Felicien(Lac-Saint-Jean).The systematic sampling o fthis coniferous sector led to the digitizing of 1 412 false color infrared images.From the classification of the Landsat TM imagery and the forestry maps of the area,this image batch was divided into groups caracterizing forest densities from A to D.30 images/group were randomly selected to analyze the performance of a stem counting algorithm basded on a local maximum extractin from the preprocessed luminance spectral band and to classify their defoliation stage by evaluating the band and to classify their defoliation stage by evaluating the hue of these maxima.While images of D density were clearly overcounted,the comparision algorithm vs interpret,including omission and comission errors,showed respectively 5