Study Of Imaging Hyperspectral Technique Based On Linear Variable Filter

As one of the important payloads for remote sensing observations,hyperspectral imagers has the ability to analyze substances qualitatively and quantitatively,and is widely used in agriculture,food processing,mineralogy,environmental monitoring,art,deep space exploration and many other fields.The miniaturization and cost-efficiency has become an important development direction of the hyperspectral imagers.Spectral imagers based on linear variable filters have the advantages of compact design,simple structure and low cost compared with the traditional grating,prism and interference based hyperspectral imagers.In this thesis,a thorough study on applying linear varaible filters in hyperspectral imagers is carried out: a forward simulation model of linear varaible filter based hyperspectral imager is established;a prototype of visible near-infrared hyperspectral imager covering between 650 nm and 810 nm,with average spectral resolution about 14.4 nm,is built and tested;An outdoor pushing-broom experiment is carried out,and the 3D hyperspectral data cube is successfully obtained;the data processing flow of the hyperspectral imagers based on linear variable filters is studied.The main innovative work In this thesis is as follows:1.A novel forward system simulation model for hyperspectral imagers based on linear variable filters is proposed.The proposed model can be used to quantitatively analyze the influences of the focal ratio of optical system,the distance between the linear variable filter and the detector,and the detector pixel size on the spectral response function of the system.The proposed model is proved to be an indispensable tool in system designing phase.2.A novel method to automatically determine the double sampling clock phases based on pseudo-contrast is proposed.With the new evaluation function named pseudocontrast taking advantages of characteristics of CCD video signals,an optimization algorithm is used to automatically determine the phases of the correlated double sampling clock.Verification experiments show that the method can be accurately determine the phases of the correlated double sampling clock with an error as low as 3%,and is robust to both the brightness variation in the scene and system noise.Compared with traditional methods,the proposed method can effectively improve the signal to noise ratio of the hyperspectral imager's output.3.A novel fast lossless compression method for hyperspectral imagers based on linear variable filters is proposed.The proposed method redefines the local difference vector in CCSDS123,taking fully advantage of the spatial-spectral co-modulation characteristics of the linear variable filter based spectral imagers.To verify the compression performance of the proposed method,experiments are taken on 6 raw datasets containing different scenes.The results show that the proposed method surpasses the original CCSDS123 method by about 21.62% higher compression ratio on the test datasets.