Evaluation of Matched Filter method for wind erosion mapping Landsat 8 OLI Imagery, (Central and North West province of Khuzestan)

Successful target detection, especially if there is a similarity between the target and the background area, always is a noticeable issue in remote sensing studies. Because of the spectral behavior similarity to the other phenomena and spatial distribution of these units, the mapping of wind erosion units is difficult. Thus, this study attempts to detect the favorable areas by using matched filter (MF) methods and satellite imagery. In this regard, the latest generation of Landsat satellite imagery (Landsat 8 OLI) in February 2015 was used to study the North West of Khuzestan province. To evaluate this method, the other techniques in this field, such as Support Vector Machine (SVM), Adaptive Coherence Estimator (ACE) and Spectral Angle Mapper (SAM) are adapted to the methods. In order to assess the accuracy of the results, the ground samples were used. The study indicated that this method (MF) with 81 percent was more accurate than other methods in identifying of the target areas. Also ACE showed a good performance after the MF method. SVM and SAM showed lower performance and ROC curve also confirms these results. It seems that MF method has a higher performance than other methods such as SVM to distinguish phenomena with similar spectral behavior and less affected by interference from the data cloud and can be used successfully to provide separation maps of phenomena such as wind erosion units