Land use/land cover mapping using recurrent neural network with PSO

Fused LISS IV multi spectral and PAN image provide rich set of spectral and spatial information that used to increase accuracy rate of classification fused LISS IV+PAN dataset. LISS IV image data set has only high spectral information. PAN image data set has only high spatial information. To reduce computation complexity as well as increase performance of classification a new model proposed in this paper. Two step approach proposed. First step use QIM with DCT for fuse LISS IV and PAN image effectively that provide rich set of spectral and spatial information. The second step involves training fused image using RNN to extract rich set of features from companied LISS IV+PAN image and PSO optimization used to optimized feature set for reduce computation complexity of classification process. In this work, Madurai region LISS IV and PAN image datasets used for fusion and classification approach. Experimental results shows that proposed RNN+PSO based approach outperform compare to other approaches.