Hopfield Neural Network for Change Detection in Multitemporal Images

This paper proposes a supervised change detection technique for multitemporal remote sensing images. The technique is presented after studying three different models for change detection using neural network and assimilating the unique feature of each of the model. The technique is based on Hopfield neural network modified to model spatial correlation between neighboring pixels of the difference image. Each pixel in the difference image is represented by a neuron in the Hopfield network that is connected to its neighbors. These connections to the neighboring units model the spatial correlation between pixels and are assigned weights according to their influence on each other with help of training sets. The information about the status of the network is rendered through an energy function allocated to the network. A threshold is defined for segmenting the pixels into two classes of pixel-changed and unchanged. Change detection map is obtained by iteratively updating the output status of the neurons until a minimum of the energy function is reached and the network assumes a stable state. Experimental results carried out on two multispectral multitemporal remote sensing images confirm the effectiveness of the proposed technique