Kazem Aliabadi; omid baghani
Abstract
This study aims to provide a computational-approximate algorithm based on Rationalized Haar (RH) to estimate the vegetation of the Landsat image using reflecting this phenomenon in the near-infrared band. This band is in the RGB color combination and located in the R section.This algorithm, using Digital ...
Read More
This study aims to provide a computational-approximate algorithm based on Rationalized Haar (RH) to estimate the vegetation of the Landsat image using reflecting this phenomenon in the near-infrared band. This band is in the RGB color combination and located in the R section.This algorithm, using Digital Number (DN) vegetation in 200 selected pixels of R band (infrared band) from the study area, tries to extract the features and vegetation of the whole study area. The number of selected pixels is distributed uniformly and only covers the vegetation.Due to using the matrix format in the input data, first vegetation reflection matrices for 4 and 8 wavelets are constructed using the assumed 200 pixels. Then, these matrices are extended to 16 and 64 parts respectively, through blocking the Landsat image of the region.Each matrix element represents the average vegetation of the area in its corresponding block. Then, by introducing an efficient mathematical equation, the vegetation of the entire study area is extracted. In addition, each pixel is reconstructed. Due to matrix calculations, speed and accuracy of calculations at the pixel scale will be listed as advantage of this approach.In this study, vegetation extraction with 4 and 8 RH Wavelets was performed with 75 and 87.5% accuracy, respectively. As the number of wavelets increases, the accuracy of the RH wavelet algorithm increases. However, rounding error and the increase in computational cost in high number of wavelet can be listed as disadvantage of this method. Such that, time and space memory will be increased exponentially. In remote sensing, extraction techniques such as classification have been proposed by remote sensing software. The accuracy of vegetation pixel extracted using this approach will be as advantage in comparison with those common methods. In processing and analytical techniques (for vegetation extraction and classification) in remote sensing, many pixels contain vegetation depicted as single or clustered (but in small numbers) while, in other classes such as barren or Urban land will be merged, which RH wavelet overcomes this shortcoming.
