A Directional Message Propagation Convolutional Neural Network for Hyperspectral Images Classification

Convolutional neural networks (CNNs) have emerged as a powerful tool in remote sensing image analysis. However, the layer-by-layer convolutions (L
²
Convolutions) in CNNs cannot fully exploit the relativities of pixels in the 3D cube data, especially for hyperspectral images (HSIs). In this paper, a directional message propagation convolutional neural network framework (MPCNN), is proposed for the supervised classification of HSIs. In the proposed framework, we integrate a novel multi-directional message propagation mechanism, namely slice-by-slice convolutions (S
²
Convolutions), into the hidden feature layers which are generated by L
²
Convolutions to propagate feature information between feature maps in the same layer. Owing to the S
²
Convolutions, abundant and discriminative spectral-spatial feature learning can be enhanced compared with traditional CNNs without S
²
Convolutions. The performance of the proposed MPCNN is evaluated on benchmark dataset of HSIs, and quantitative and qualitative experiments show that the performance of the proposed method outperforms several state-of-the-art methods.