Research On Ship Targets Detection Based On Hyperspectral Data

Hyperspectral Data can provide spatial information and diagnostic spectral characteristics,in the range of visible and near-infrared wavelength,about the attributes of materials in the scene,developing a reliable new method for target detection.has been widely used In many areas.However,hyperspectral target detection has some shortcomings,such as the classical hyperspectral target detection algorithm from the spectral dimension detection target,there is the lack of image information estimation,hyperspectral data on the light conditions are more dependent on the shadow of the suppression of the spectrum The feature of the shadow area target detection effect is not ideal.In this paper,the hyperspectral target detection algorithm and its improved method are studied in order to apply the natural advantage of the hyperspectral target detection algorithm to the complex and complex ship target detection.In this paper,we first design a single spectral feature target detection experiment,low spatial resolution and high spatial resolution multi-spectral feature target detection experiment,from the perspective of detection accuracy and time efficiency of the classical single target and multi-target algorithm for ship target detection performance And then on this basis,the classical algorithm is based on histogram distribution for secondary detection and improved morphological improvement based on the study to improve the detection accuracy of ship targets and proposed based on 3D ROC area index and resolution index The detection criteria of the ship target in the shadow area are studied by using the improved moment matching method for the defects that are more dependent on the hyperspectral data.Based on the research work,the main conclusions are as follows:(i)The OSP algorithm with known background is not required to estimate the image information,and the execution time of the algorithm is small.The detection efficiency is very good in terms of accuracy and time efficiency.But the efficiency of the algorithm with the precision of prior knowledge acquisition is closely related,is not suitable for ship target detection scene detection algorithm RXD the myriads of changes;only the unknown targets in unknown background as a high dimensional hyperspectral data processing,only suitable for ship targets just for high dimensional point weak components;the known target of unknown background AMF algorithm and ACE algorithm in time efficiency is low,and the CEM algorithm in terms of detection accuracy and time efficiency considerations,the detection performance is very ideal.(ii)For a single target detector to improve the detection accuracy,efficiency and time,the Mor CEM detector performance is the best,followed by MaskCEM detector,the other detectors are behind a single grade;if from the detection accuracy,followed by Mor CEM,MaskCEM,MorACE,MorRXD,MaskACE,the precision reaches more than 99%.If the individual detector;considering the time efficiency,the CEM detector is the best choice,followed by MaskCEM and MorCEM detector,in fact the accuracy of 95.45% CEM detector at a high level.(iii)priority selection problem in multi target detectors,when the spatial resolution is not high enough,only a single target spectrum curve(such as the HYPERION data),from the perspective of the detection accuracy of the detection performance of the top of the order of SumCEM,MtCEM,SumAMF,SumACE,MtAMF;from the perspective of time efficiency of detection performance the top of the order of MtCEM,SumCEM,WtaCEM,SumACE,WtaACE;efficiency and detection precision of time,the detection performance of the best SumCEM,MtCEM,SumACE.The target detector to improve the detection accuracy,efficiency and time,the optimal time of MorMtCEM detector,MaskMtCEM;only considering the detection accuracy,MorMtCEM detector is the best,followed by MorSumCEM,MorSumAMF,MorSumACE,MaskSumCEM,MorMtAMF,the detection accuracy reached more than 96%;if the pursuit of efficiency,MtCEM the detector has obvious advantages,the detection accuracy is above 94%.When the spatial resolution is higher,the spectra of several kinds of materials can be extracted from a target,and the spectrum has some difference(such as flight data).From the perspective of detection accuracy,the first few cases are MtACE,SumACE,SumtaM,MtAMF,Sumtom,MtAMF,Sumtom,and MtAMF,respectively.The results show that the detection performance is MtAMF,SumCEM,and the detection performance is the best.The improvement of the multi-target detector,the integrated time efficiency and the detection accuracy are still the best MorMtCEM detector,followed by MaskMtCEM.From the point of view of detection accuracy,the degree of the detector is MorMtACE,MaskMtACE,MaskSumACE,MorSumCEM,MorMtCEM,MaskSumCEM,MorMtCEM,detection accuracy of more than 96%;from the point of view of time efficiency,MtCEM detector the best detection accuracy is more than 90%.(iv)from the perspective of the degree of recovery of spectral characteristics of target shadow area,three kinds of spectral information of ship target material recovery degree reached 97.95%,96.95%,96.94%;the improvement of detection precision,6 kinds of detector accuracy from the original image of the 0.7171,0.7172,0.7486,0.6825,0.7334,0.6551 to the shadow area the spectrum after a recovery of 0.9008 and 0.9007,0.9445,0.8953,0.9395,0.8785.The improved moment matching method based on correlation coefficient is proved to be feasible for shadow removal.