Hyperspectral image (HSI) can provide rich spectral information which can be helpful for accurate classification in many applications. Yet, incorporating spatial information in the classification process can improve the classification accuracy even further. Existing convolutional neural network (CNN) usually only focuses on local features in hyperspectral cubes, whereas the burgeoning vision transformer (ViT) is interested in global features in HSIs. In this letter, we propose a deep aggregated framework for HSI classification called convolution transformer mixer (CTMixer) to combine the advantages of the above two paradigms effectively. A group parallel residual block is first applied to capture local spectral-spatial features in the HSI patches. Second, a double-branch structure, consisting of the CNN and transformer branches, is developed to capture local–global hyperspectral features. Finally, to achieve an elegant combination of CNN and ViT, a novel local–global multihead self-attention mechanism (MHSA) is proposed by introducing convolution operations in the MHSA to further improve the classification accuracy. Extensive experiments demonstrate that the CTMixer achieves competitive classification results on several common HSI datasets compared with other state-of-the-art networks. The source code for this work will be available at https://github.com/ZJier/CTMixer .
Fuzzy set (FS) theory is a useful tool for dealing with the uncertainty in images. Nonetheless, images are always contaminated by noise which further increases the uncertainty. Comparing with FS, intuitionistic fuzzy set (IFS) can better describe the blurred characteristic in images due to taking into account the membership degree as well as the non-membership degree and the hesitation degree. However, when applied to image segmentation, IFS cannot completely overcome the influence of noise. With the aim of performing noisy image segmentation under several criteria, this paper defines a noise robust intuitionistic fuzzy set (NR-IFS) for an image and then presents a novel noise robust multi-objective evolutionary intuitionistic fuzzy clustering algorithm (NR-MOEIFC). A majority dominated suppressed similarity measure using the neighborhood statistics and the competitive learning is proposed to obtain the NR-IFS representation for the image corrupted by noise. Then NR-IFS is fully used to motivate the whole process of multi-objective evolutionary clustering: (1) computing a three-parameter intuitionistic fuzzy distance measure, (2) constructing intuitionistic fuzzy fitness functions, (3) designing a non-uniform intuitionistic fuzzy mutation operator, and (4) defining an intuitionistic fuzzy cluster validity index to select the optimal solution from the final non-dominated solution set. Especially, the histogram statistics of NR-IFS are adopted in NR-MOEIFC to greatly reduce the computational complexity. Experimental results on Berkeley and real magnetic resonance images reveal that NR-MOEIFC behaves well in noise robustness and segmentation performance, and meanwhile requires a low time cost.
6 April 2021 in IEEE Geoscience and Remote Sensing Letters
Convolutional neural networks (CNNs) showed impressive performance for hyperspectral image (HSI) classification. Nevertheless, convolutional layers contain massive parameters, which restrict the deployment of CNNs on satellite and airborne platforms with limited storage and computing resources. In this letter, we propose a lightweight spectral-spatial convolution module (LS 2 CM) as an alternative to the convolutional layer. The proposed LS 2 CM can greatly reduce network parameters and computational complexity in terms of multiply-accumulate operations (MACs) while maintaining or even improving the classification performance. Furthermore, it is a plug-and-play component and can be used to upgrade existing CNN-based models for HSI classification. Experimental results on two benchmark HSI data sets demonstrate that the proposed LS 2 CM achieves competitive results in comparison with other state-of-the-art methods.
Interval type-2 fuzzy c-means clustering (IT2FCM) algorithm describes more uncertainty than fuzzy c-means clustering (FCM) algorithm. How to obtain appropriate fuzzifiers and initialize cluster centers are essential tasks for IT2FCM. To effectively solve these problems, this paper proposes an alternate particle swarm optimization -based adaptive interval type-2 intuitionistic fuzzy c-means clustering (A-PSO-IT2IFCM) algorithm and applies it to color image segmentation. Firstly, in order to further deal with the uncertainty, a novel interval type-2 fuzzy clustering objective function is constructed by utilizing the intuitionistic fuzzy information extracted from images. Then an alternate particle swarm optimization (PSO) scheme is designed to optimize fuzzifiers and cluster centers alternatively. In addition, a multiscale update strategy for the position of particles is introduced into A-PSO-IT2IFCM to increase the diversity of swarm and boost the convergence of optimization. Color image segmentation experiments on Berkeley and UC Merced Land Use datasets show that the proposed algorithm can adaptively determine fuzzifiers and cluster centers and achieve good segmentation results.