Research On Hyperspectral Characteristic Of Camouflage Clothing

Due to the characteristics of hyperspectral, such as high spectral resolution, more band number and large amount of data, makes the hyperspectral disguise and recognition detection technology has gradually become an important means of battlefield reconnaissance. And furthermore, the recognition and detection technology of Camouflage clothing are an important method for military reconnaissance. Thanks to hyperspectral technology, it is possible to accurately Classify and detect the Camouflage clothing.In this paper, the hyperspectral image data of Camouflage clothing and background features were obtained by the experiment. The camouflage hyperspectral data were analyzing and processing.Respectively studying Spectral Encoding, Spectral Angle Mapping, Continuous Removal, Derivative Spectra methods for spectral transform and processing, to analyze the spectral curve features of camouflage and background feature. Using Principal Component Analysis and the Minimun Noise Fraction methods to effective reduce band dimensions of hyperspectral data. Based on the definition of training sample, respectively applied classification such as Minimum Distance, Markov Distance, Spectral Information Divergence and Spectral Angle Papping to hyperspectral image classification, and verified accuracy evaluation with the confusion matrix.Using Spectral Angle Mapping and Spectral Information Divergence methods to camouflage target detection in hyperspectral image.Results show that the Markov Distance classification method has a higher classification accuracy, the overall accuracy evaluation up to95.99%. This method can effectively detect the camouflage target samples, the detection results were quite acceptable. It is also showed that the target detecting result of Spectral Information Divergence method is better than Spectral Angle Mapping.