Comparison of Parametric and Non-Parametric Techniques to Accurate Classification of Forest Attributes on Satellite Image Data

نوع مقاله : مقاله پژوهشی


گروه جنگلداری، دانشکده کشاورزی و منابع طبیعی اهر، دانشگاه تبریز


Satellite images classification techniques obtain large information in forest areas. Among various techniques, k Nearest Neighbors (kNN) as a non-parametric and MLC (Maximum Likelihood Classification) as a parametric technique have popularity in predict forest attributes by integrating the inventory field data and satellite images. In this study, comparisons between k Nearest Neighbor (kNN) non-parametric method and Maximum Likelihood Classification (MLC) parametric method was performed in forest attributes consists of volume, basal area, density, and tree cover type estimation in the north forest of Iran. Results showed that kNN non-parametric method produces an accurate classification map in comparison to the MLC parametric method and the accuracy of kNN has the most amount in all attributes. Kappa coefficient estimation showed that the kNN method had the most amount of this coefficient in all attributes. Accordingly, the kNN non-parametric technique was identified as a feasible classification technique to produce forest attributes thematic maps.