Automatic Radiometric Normalization of Multi-Temporal Satellite Image based on IR-MAD Transformation and Artificial Neural

Relative radiometric normalization often used in multi-temporal satellite image analysis، especially land use change detection. In this paper، IR-MAD transformation has been reviewed and a new method has been developed based on this transformation and artificial neural networks، also. The proposed method is implemented on multi-temporal Landsat TM satellite images captured in 1989 and 2010. Study area is located in Tabriz. According to Linear combination of multi-temporal satellite images، transfer the images to another space and iterative process of IR-MAD transformation، the transformation has led to independent method of statistical noise and atmospheric conditions and Used in this study for change detection and selection radiometric controls point. The capability and flexibility of ANN in approximation of nonlinear and linear continuous functions in the hybrid space has led to the networks used for modeling of relationship between radiometric controls point in multi-temporal satellite images. Evaluation metrics in this paper، include root mean square error، T-test and F-test. The results show that the proposed method increases the accuracy and performance relative radiometric normalization. The proposed method has increased. Root mean square error in all spectral bands than IR-MAD and raw data respectively 0. 11 and 8. 13%.