Novel Variable Selection Method Based on Uninformative Variable Elimination and Ridge Extreme Learning Machine: CO Gas Concentration Retrieval Trial.

Variable selection is an essential part in spectroscopy analysis area. To overcome the problems of traditional interval selection methods, this paper proposed a novel variable selection and assessment method based on uninformative variable elimination (UVE) and ridge extreme learning machine (RELM) algorithms. Firstly, the UVE method was adopted to eliminate the uninformative wavelengths. Secondly, to solve the collinearity problem, RELM algorithm was adopted to replace the traditional modeling methods (PLS, BP neural network, etc.). Finally, the optimal combination of wavelength regions was selected by using feature selection path (FSP) plot and sparsity-error trade-off (SET) curve. The experiment results of CO gas concentration retrieval showed that (1) the UVE algorithm can select the most informative variables, which were the feature wavelengths of the CO gas transmittance spectrum; (2) the RELM algorithm has the advantage of rapid modeling, solving the collinearity problem, and high accuracy (the determined coefficient r of CO gas concentration retrieval can reach 0.995); (3) the FSP plot and SET curve were easy understanding, also intuitive to experts to find the best combination of wavelengths and extract useful domain knowledge.

[Passive ranging of infrared target using oxygen A-band and Elsasser model].

Passive ranging method of short range and single band was developed based on target radiation and attenuation characteristic of oxygen spectrum absorption. The relation between transmittance of oxygen A band and range of measured target was analyzed. Radiation strength distribution of measured target can be obtained according to the distribution law of absorption coefficient with environmental parameters. Passive ranging mathematical model of short ranges was established using Elsasser model with Lorentz line shape based on the computational methods of band average transmittance and high-temperature gas radiation narrowband model. The range of measured object was obtained using transmittance fitting with test data calculation and theoretical model. Besides, ranging precision was corrected considering the influence of oxygen absorption with enviromental parameter. The ranging experiment platform was established. The source was a 10 watt black body, and a grating spectrometer with 17 cm(-1) resolution was used. In order to improve the light receiving efficiency, light input was collected with 23 mm calibre telescope. The test data was processed for different range in 200 m. The results show that the transmittance accuracy was better than 2.18% in short range compared to the test data with predicted value in the same conditions.

[The study of photo-elastic modulator-based Fourier transform spectroscopy].

In order to enhance the spectrum resolution of current photo-elastic modulator-based Fourier transform spectrometer, a multi-reflected photo-elastic modulator-based interferometer structure was proposed in the present paper. Through coating reflecting film alternatingly on the photo-elastic crystal and light oblique incidence, and allowing the incident ray to have the multi-reflection in the crystal and exit from the other side of the crystal, the authors increased the light propagation distance in the crystal and enhanced the optical path difference at last. Based on this, the function of interference-spectrum retrieval was established, the optical system matched to the multi-reflected PEM-based interferometer was designed, and finally, the experimental system of multi-reflected PEM-based Fourier transform spectroscopy for telemetry was established. The principle of verification tests by using 671 nm laser and xenon lamp shows that the interferogram was clear and stable, and the feasibility of the principle of the system was verified. The expected result shows that the spectrum resolution of the designed PEM-FTs with multi-reflection achieved 13 cm(-1), and its luminous flux just didn't reduce too much, which ensured the SNR. Through spectral inversion of the interference fringes, the technical feasibility of the spectrum system developed was verified. This work established the basic condition of prototype fabrication, radiation precise calibration, spectral calibration and instrument signal-to-noise ratio test and so on.

[The technology of fast spectral reconstruction in the longer optical path difference PEM-FTS].

The optical path difference of the photoelastic modulator Fourier transform spectrometers (PEM-FTS) changes rapidly and nonlinearly, while the instrument preserves the speed as high as about 10(5) interferograms per second, so that the interferograms of PEM-FTS are sampled by equal interval. In order to fleetly and accurately reconstruct these spectrums, the principle of PEM-FTS and accelerated NUFFT algorithm were studied in the present article. The accelerating NUFFT algorithm integrates interpolation based on convolution kernel and fast Fourier transform (FFT). And the velocity and precision of the algorithm are affected by the type and parameter tau of kernel function, the single-side spreading distance q and the oversampling ratio micro, and so on. In the paper these parameters were analysed, under the condition N = 1 024, q = 10, micro = 2 and tau = 1 x 10(-6) in the Gaussian scaling factor, and the accelerated NUFFT algorithm was applied to the longer optical path difference PEM-FTS to rebuild the spectrums of 632. 8 nm laser and Xenon lamp, The frequency error of the rebuilt spectrums of 632.8 nm laser is less than 0.013 52, the spent time of interpolation is less than 0.267 s. the velocity is fast and the error is less. The accelerated nonuniform fast Fourier transform is fit for the longer optical path difference PEM-FTS.

[Research on concentration retrieval of gas FTIR spectra by interval extreme learning machine and genetic algorithm].

This paper proposed a novel effective quantitative analysis method for FTIR spectroscopy of polluted gases, which select the best wavenumbers based on the idea of interval dividing. Meanwhile, genetic algorithm was adopted to optimize the connect weights and thresholds of the input layer and the hidden layer of extreme learning machine (ELM) because of its global search ability. Firstly, the whole spectrum region was divided into several subintervals; Secondly, the quantitative analysis model was established in each subinterval by using optimized GA-ELM; Thirdly, the best combination of subintervals was selected according to the generalized performance of each subinterval model by computing the parameters root mean square error (RMSE) and determined coefficients r. In this paper, the mixture of CO, CO2 and N2 O gases were selected as the research object and the whole spectrum range was from 2 140 to 2 220 cm-1. The experiment results showed that the RMSE of model established with the selected wavenumbers was 154. 996 3, the corresponding r can reach 0. 987 4, and the running time was just 0. 8 seconds, which indicated that the concentration retrieval model established with the proposed Interval-GA-ELM (iGELM) method can not only reduce the modeling time, but also can improve the stability and predict accuracy, especially under the condition of the exist of interferents, which providing an effective approach to the remote analysis of polluted gases.

[Automatic classification method of star spectra data based on manifold-based discriminant anaysis and Support Vector Machine].

Although Support Vector Machine (SVM) is widely used in astronomy, it only takes the margin between classes into consideration while neglects the data distribution in each class, which seriously limits the classification efficiency. In view of this, a novel automatic classification method of star spectra data based on manifold-based discriminant analysis (MDA) and SVM is proposed in this paper. Two important concepts in MDA, manifold-based within-class scatter (MWCS) and manifold-based between-class scatter (MBCS), are introduced in the proposed method, the separating hyperplane found by which ensures MWCS is minimized and MBCS is maximized. Based on the above analysis, the corresponding optimal problem can be established, and then MDA transforms the original optimization problem to the QP dual form and we can obtain the support vectors and decision function. The classes of test samples are decided by the decision function. The advantage of the proposed method is that it not only focuses on the information between classes and distribution characteristics, but also preserves the manifold structure of each class. Experiments on SDSS star spectra datasets verify the effectiveness of the proposed method.