Deep Convolutional Asymmetric Autoencoder-Based Spatial-Spectral Clustering Network for Hyperspectral Image

Algorithm 1

Flow chart of over-clustering -means algorithm.

Over-clustering-means

Input: HSI feature vectors, , .

1: Set according goal clusters, is 2~3 times of

2: repeat

3: Do -means cluster with clusters

4: Keep the top clusters with the most elements, the other clusters are set to the background cluster.

5: Observe the clustering result and adjust

6: until Remove noisy class effectively

7: output: HSI clustering result