Multi-resolution vegetation classification along lower Rio Grande

We present a quasi-wavelet based approach for multiresolution unsupervised and supervised vegetation classification of videography imagery. Due to certain limitations of traditional wavelet decomposition, we investigate a new quasi-wavelet decomposition to generate an alternative feature space for vegetation classification. Most existing wavelet-based classification approaches only consider the information on low-low frequency subspaces at each decomposed stage. However, both low and high frequency features are applied in our algorithms and are controlled by weight setting on the vector feature space. The thresholding method is used for generating the original class center to achieve a convergent result for unsupervised classification. The results on lower Rio Grande Valley videography imagery are presented.