Blind Source Separation Technique Used in Forest Images Analysis

With the purpose of forest remote sensing images analysis, this method utilizes a Blind Source Separation technique, named Robust Second Order Blind Identification. In orthogonalization stage, a Robust orthogonalization is performed which ensures that positive definite covariance matrix is not sensitive to the additive noise.The separation process considers the periodical spatial distribution of vegetation as source, namely the coverage along the forest plantation position. Consequently, the spectrum of vegetation was estimated as the mixture ratio of the source.This technique allows recognizing the qualitative change of vegetation from the estimated spectra and quantitative one from the coverage