Methods And System Of Marine Target Detection In Hyperspectral Remote Sensing Based On Space Spectrum Combination

Hyperspectral remote sensing image has the characteristic of high resolution,and the data contains rich and detailed spectral information of ground objects.This feature can greatly improve the ability to identify ground targets,and it has a great advantage in small pixel and sub-pixel targets on the hyperspectral image.Therefore,hyperspectral remote sensing technology is widely used in gas detection,geological identification,military reconnaissance and marine monitoring and many other fields.Target detection in hyperspectral remote sensing image is an important technique in hyperspectral remote sensing,which is essentially a binary hypothesis testing problem.Each pixel is judged whether belong to a background or a target,in order to obtain the detection effect map of ground objects.However,the existence of mixed pixels in hyperspectral images and its huge amount of information also bring severe challenges to the image target detection.In this paper,we will give corresponding detection methods for maritime ships and red tide targets respectively by fully exploring the low rank and sparse characteristics of hyperspectral ocean remote sensing data on the basis of traditional marine target detection methods.The main work and research results of this paper are as follows:(1)We propose a kernel sparse representation for hyperspectral marine red tide target detection based on neighbor information.And first,we introduce sparse representation detailedly,by the way of the singular value decomposition algorithm to solve the problem of dictionary,and then use the orthogonal matching pursuit algorithm to solve the sparse coefficient of background and target respectively,finally by calculating the target detection operator to get the category of the pixel.Considering the nonlinear characteristics of hyperspectral data,the sparse representation is optimized by kernel method,and the dictionary learning KSVD and coefficient solving OMP algorithm are nucleated.Then we make full use of the spatial information in the hyperspectral image data and introduce the spectral information of adjacent pixels to be measured to conduct an average filtering smoothing for the algorithm.Finally,the superiority of the target detection algorithm based on kernel sparse representation is verified in the real Marine hyperspectral data.(2)We propose a low rank and sparse representation for hyperspectral ship target detection based on adaptive weighted kernel norm approximation.Aiming at the disadvantage that the traditional kernel norm minimization is not accurate in describing the background edge and texture,this paper introduces an adaptive weighted kernel norm approximation method which can set the weight adaptively to improve the accuracy of kernel norm approximation based on the traditional kernel norm approximation.Experiments on real ocean specular data show that this method has good detection ability.(3)Based on above model and algorithms,we design and implement a software interface system for Marine target detection in hyperspectral image under the framework of MATLAB GUI.The system adopts layered and modular design methods,and mainly includes a main system interface and several sub-interfaces with multiple functions,including image display,detection type and detection analysis.Besides,I deeply analyzed the design idea of the whole system and showed the specific function of the realization of the effect diagram.