KLT-CRKCN: Hyperspectral Image Classification via Karhunen Loeve Transformation and Collaborative Representation-Based K Closest Neighbor

Collaborative representation-based classification is extensively used in numerous application fields for e.g. hyperspectral image and face recognition. This paper presents a simple, stable as well as efficient classification algorithm Collaborative Representation based K Closest Neighbor classes (CRKCN) that depends on nearest neighbor method. In the CRKCN, primarily some testing samples are characterized in the way of linear arrangement of each accessible training samples, and later the representation weights are assessed through l(2)-norm minimization. Specifically, the leading k closest neighbor pattern classes which are situated in the neighborhood of the sampling data require to be tested and further utilized. The labels for testing samples are calculated through majority votes related to weights of k major representations. In addition, a Karhunen Loeve Transformation based dimensionality reduction is executed for the selection of related bands. Finally, the experimental evaluation of proposed structure on different hyperspectral image datasets validates that it can lead a promising improvement in classification performance in contrast to related competitive traditional classification methods.

Automated summary

The paper introduces a Collaborative Representation based K Closest Neighbor classes (CRKCN) classification algorithm that shows promising improvements in classification performance for hyperspectral images compared to traditional methods.