Hyperspectral Image Classification by Using K-Nearest Neighbor Algorithm
1Perepi. Rajarajeswari, R. Hemashri, S. Jayapriya and M.M. Ravikumar
Recently, Deep learning has been acknowledged as one of the strong tool for feature-extraction to effectually address non linear problems and is employed in image processing tasks and has attained good performance. However, excessively increasing the depth of the network will lead to overfitting and gradient vanishing. To address these issues, a deep feature fusion network (DFFN) is introduced. With the application of knearest neighbor (KNN) algorithm in combination with residual learning, improves the classification accuracy by extracting much more discriminative features of HSI and also ease the training of deep network. It also extracts the hybrid features of the various classes in the image. The proposed model combines the results of various hierarchical layers that improve the classification accuracy.
K-Nearest Neighbour (KNN), Hybrid Features, Residual Learning, Feature Fusion.