Research On Model And Algorithm Of Hyperspectral Image Unmixing

In the process of hyperspectral image imaging,due to the limitation of the resolution of imaging equipment,and the complexity of weather and ground features,a single pixel is usually mixed with multiple pure materials in a certain proportion,which is also called a mixed pixel.The existence of mixed pixels makes the hyperspectral image lose some information.Hyperspectral image unmixing is a process of decomposing the measured spectrum of mixed pixels in the hyperspectral image into the endmembers spectrum and the corresponding abundances.This technology plays a great role in the military and civil fields.Sparse hyperspectral unmixing has been a hot topic in recent years.There are abundant spatial information and spectral information in the hyperspectral images.Making full use of the spatial information and spectral information of hyperspectral images will improve the accuracy of unmixing results.In view of the current research work can be improved,we put forward new methods.The main work and innovation of this paper are as follows,First,recently,the low-rank representation provides a new perspective for spatial correlation and the weighted nuclear norm regularization has been well studied to enhance the low-rankness of the abundance matrix.However,the commonly used weights only depend on respective singular values,ignoring other singular values’ information.In this paper,we propose a new weighting scheme for the weighted nuclear norm to further enhance the sparsity of the singular values of the abundance matrix.The proposed weight for each singular value considers information of all singular values,instead of particular singular value only.In this way,we further increase the punishment on smaller singular values and decrease the punishment on large singular values simultaneously.Then we refine two sparsity and low-rankness based unmixing algorithms.Simulated and real-data experiments demonstrate the effectiveness of the resulting unmixing algorithms.Second,j oint sparsity is a hypothesis based on the spatial information of hyperspectral images.Joint sparsity assumes that each pixel in a small neighborhood of hyper-spectral images is composed of the same endmembers.Hyperspectral images are three-dimensional tensors.Recall that a plethora of unmixing algorithms reshapes a 3-D HSI into a 2-D matrix with vertical priority.To make further use of the spatial information of HSIs,in this article,we propose a bilateral joint-sparse structure for hyperspectral unmixing in an attempt to exploit the local joint sparsity of the abundance matrix in both the vertical and horizontal directions.On this basis,a bilateral joint sparse regularization term is proposed.Then,we propose a new unmixing model considering both low rank and bilateral joint sparsity,and use the alternating direction multiplier method（ADMM）to solve the model to obtain the bilateral joint-sparse and low-rank unmixing（BiJSpLRU）.Finally,we test the algorithm on the real and simulated data set,and compare the performance of the algorithm with other classical hyperspectral unmixing algorithms.Experimental results show the effectiveness of the proposed algorithm.