Level-Wise Band-Partition-Based Hierarchical Representation Residual Feature Learning for Hyperspectral Target Detection

This paper proposes a Level-wise Band-partition-based Hierarchical Representation residual Feature (LBHRF) learning method for hyperspectral target detection (HTD). Specifically, the LBHRR method proposes to partition and fuse different levels of sub-band spectra combinations, and take full advantages of the discriminate information in representation residuals from different levels of band-partition. First, the original full spectral bands are partitioned in a parallel level-wise manner to obtain the augmented representation residual feature through level-wise band-partition-based representation residual learning, such that the global spectral integrity as well as the contextual information of local adjacent sub-bands can are flexibly fused. The SoftMax transformation, pooling operation, and augmented representation residual feature reuse among different layers are introduced in cascade to enhance the discriminative ability of the nonlinear and discriminant hierarchical representation residual features of the method. Finally, a hierarchical representation residual feature-based HTD method is developed in an efficient stepwise learning manner, instead of back-propagation optimization. Experimental results on several HSI datasets demonstrate that the proposed model can yield promising detection performance in comparison to some state-of-the-art counterparts.