Matching of Remote Sensing Images based on Projective Transformation and Using Hopfield Neural Network

Because there are differences of scale، rotation and intensity and deformation caused by the high of the satellite images، matching is one of the challenging problems in remote sensing. This activity intensively is used in remote sensing for many purposes such as multi source classification، environment monitoring، change detection، image mosaicking، etc. Among the variety of matching methods، one of them is global matching by Hopfield neural network. This method is mainly applied in object recognition and close range dominant and up to now it is not evaluated on aerial and remote sensing images. Implementing the matching method using Hopfield neural network on remote sensing images is the main goal of this study. In research conducted، fifth-order Hopfield neural network is used to perform the projective invariant matching. This network with increasing number of points loses its speed and efficiency in matching. Therefore، in this paper by reducing the order of Hopfield neural network from fifth-order to second-order improves the efficiency and speed of matching method. Practical results on two pairs remote sensing images shows the efficiency of this method.