Today with the advancement of technology and the use of data obtained from multispectral and hyperspectral sensors, new and advanced methods of remote sensing image analysis are considered in order to extract more information from multispectral and hyperspectral data. One of the analysis that is performed on the hyperspectral images is target detection. Target detection methods at hyperspectral images is usually done based on characteristics and spectral date. In this papar, we use LISSIII multispectral image in order to spectral detection of target using minimum Noise Fraction Transform (MNF) and Mixture Tuned Matched Filtering (MTMF).  MNF algorithm eliminate correlation and redundancy between bands and another one eliminate the noise of the image, simultaneously. Then on the next step MTMF algorithm were implemented on the image with reduced noise and spectral dependence. The results showed that the number of false positives have a key role in improving or not improving of classification accuracy in the detection of sand dunes. With eliminating the number of false positives in this method, we were gained acceptable results compared to image classification method and it was determined that MTMF method  compared with image classification methods is more accurate and efficient method to identify the sand dunes. Mixing sand's effect with each other and with other phenomena such as vegetation also reduces the accuracy of the algorithm. So understanding the effects of geomorphology, vegetation, and geological studies lead to reduce the error rate of algorithm ad will increase accuracy. Implementing MTMF and MNF algorithm on the hyperspectral image will have the better and more acceptable result compare to multispectral images. Since the LISSIII image has a few bands and MNF method is designed for hyperspectral images, in this study, this method played a key and great role in detecting the sand dunes.