Hyperspectral Target Reconnaissance

Since the advent of hyperspectral remote sensing technology,it has been widely focused because of its rich information.Hyperspectral images play an important role in target detection and classification.So far,hyperspectral target detection technology has been greatly developed.However,the fast realization of hyperspectral target is still a problem to be solved.Hyperspectral target detection is based on the difference between target spectrum and background spectrum.When the detection target spectrum is very similar to the background spectrum,effective target recognition becomes a challenge.Based on the study of common data dimension reduction methods and target detection algorithms,two new data dimension reduction method are proposed,and a new target detection algorithm is proposed.The main work of this paper is as follows:Firstly,Airborne Hyperspectral Remote Sensing Technology is summarized.Airborne remote sensing characteristics,spatial sca nning methods and spectral reconstruction of airborne spectrometer are introduced.Also the basic characteristics of hyperspectral data,data reduction and target detection are studied.Secondly,there are many bands in hyperspectral remote sensing image,which affect the efficiency of detection algorithm.In order to achieve fast detection of hyperspectral targets,BSEF algorithm is prosed based on the OIF algorithm with information entropy introduced to calculate band information and Barkhausen coefficie nt introduced to measure the correlation between bands.The data dimension reduction method cost less time,and the selected band combination is convenient for unsupervised target detection.Compared with direct target detection,consumed time can be shortened to a quarter after dimensionality reduction of BSEF algorithm in unsupervised target detection with higher target detection rate and lower false alarm rate.On the basis of BSEF,BERF algorithm is proposed to improve the efficiency of supervised target detection which can shorten the running time to one third of the original band.Thirdly,after analyzing the factors affecting the results of C EM detection,it is found that the selection of image pixels can improve the autocorrelation coefficient.Therefore an improved C EM target detection algorithm is proposed.In this method,spectral reordering and first order derivation of hyperspectral data are firstly used to increase the difference between target and background.The similarity between target spec tral and spectral points of data set is calculated,and the pixels with high similarity with target are removed when the autocorrelation matrix of CEM algorithm is obtained.To further suppress the background,a logarithmic operator is added.Finally,experiments on synthetic hyperspectral data and real hyperspectral data show that the proposed algorithm can recognize camouflaged targets effectively,and is applicable to small targets and large area targets detection.At the same time,the algorithm has strong robustness and can run normally when the SNR is as low as 23 and the SAM value of the camouflage and background is as high as 1.3.