Taking into account the theoretical resolution of the concave gratings, the aberration, the width and height of the slit and the size of the grating, mathematical expression for the resolution of the system was deduced and a new method for simulating and testing the system was proposed. We drew some following conclusions. The system can not improve the resolution by increasing indefinitely grating lines. The energy and the resolution are considered to select an appropriate width of the slit. The impact of the width of the slit on the system resolution is far greater than that of the height of the slit and the theoretical resolution of the concave grating. And then, a formula of the size of the gratings and a method for optimizing were proposed. The theoretical calculation values coincide with the experimental results.

In order to measure the absorption spectrum, whose radiation wavelength range is between 1 591 nm to 1 621 nm (near infrared band)，a kind of high spectral resolution spectral detector is designed for measuring the near infrared CO2 absorption bands, and the instrument underlying is in progress. Compared with existing optical systems applied to spaceborne hyperspectral imagers abroad, a simple planar reflection grating structure is selected for beam splitting. The whole detector is consisted of a fore optics system and a spectral imaging system. To fit the big opposite aperture (F1.8) and small IFOV, Petzval construction is applied to the fore optics system. The system’s simulation and optimization is completed by ZEMAX optical design software, which shows that the optical design completely and feasibly meet the application requirements. System realize the high spectral resolution of 0.1 nm, and the Modulation Transfer Function (MTF) of both the center and the edge wavelength are higher than 0.65 at the characteristic frequency of 20 lp/mm.

High spectral resolution is the inevitable trend in the future development of spectral imager. However, the traditional spectral calibration based on the monochromator in order to ensure enough output power made bandwidth of the calibration light source and spectral resolution of the spectral imager become very close, which would significantly affect the precision of the calibration. This paper quantitatively analyzed that the bandwidth of calibration light source bring the influence on the spectral calibration, and the methods of deconvolution or pixel merging was put forward to correct the calibration error. Finally, calibrated the spectral imager with different bandwidth light sources, and corrected by the before-mentioned methods. The results show that the accuracy of the spectral resolution is better than 0.2 nm, achieves the accuracy of the spectral calibration system and verifies the feasibility of the method.

To address problems of cloud phase retrieval using the threshold method, a cloud phase retrieval approach based on Self-Organizing Feature Map (SOFM) neural network was proposed. Cloud phase retrieval experiments were conducted using FengYun-3A/Visible and InfRared Radiometer (FY-3A/VIRR) multi-channel images, which cover the China’s territory. Experiment results indicated that the results from the SOFM neural network approach and the K-means method have good consistency, and the retrieval accuracy of the SOFM neural network exceeds that of the FY-3A operational product. Additionally, retrieval time consumed by the SOFM neural network approach is only about one third of that of the FY-3A operational product.

Aiming at SHS blind-data error and analyzing the formation mechanism and distribution characteristics, a global threshold judge method and windowing technique are proposed, which combines the detection and repair of interferogram blind-data. Through blind-data correction experiment on continuous light sources and atmospheric carbon dioxide absorption interferogram, the result shows that the algorithm does not destroy the interferogram quantity of information while the compensation of blind-data error, which improves the accuracy of spectral reconstruction.

Using the cascade structure of an Electro-absorption Modulator (EAM) and a frequency modulator (FM), we proposed a new Optical Frequency Comb (OFC) generation scheme. The program structure is simple, easy to be controlled and the output noise is low. We conducted a theoretical analysis on the principle scheme and did research on the simulation with software of Optisystem7.0. The simulation results show that: we generated a flat optical frequency comb, whose center wavelength and line spacing can be tuned independently. Also，the effective bandwidth of the OFC is 3.8 nm, its line spacing is 5 GHz and number of lines is 105.

Tunable Diode Laser Absorption Spectroscopy (TDLAS) technology, with the advantages of high selectivity and high accuracy, provides a reliable technology means for ammonia (NH3) on-line detection of the environment. There are many factors affecting the measurement accuracy of TDLAS system. Among them, temperature and pressure are the two basic influence conditions. First, the present paper introduces the effects of temperature change on NH3 concentration detection. Between -10 ℃ ~ 50 ℃,using Hollow Waveguide (HWG) cell to detect the NH3 of 50 ppm, the second harmonic spectra were obtained, which show that the amplitude of the second harmonic signal will decrease with increasing temperature. When the pressure changes from 0 kPa to 100 kPa, the amplitude of the second harmonic signal will decrease with the increasing of the pressure. According to the above results, the experience formula of temperature and pressure correction was raised. The maximum relative error of the 50 ppm concentration NH3 after temperature correction is -5.5% at different temperature condition. The long time monitoring results show that the system could adapt to on-line monitoring after correction.

Sparse representation had achieved very good results in hyperspectral imaging anomaly detections. A local joint sparse index method was employed, which combined local spectral sparse index and local spatial sparse index. The influence of the window design on the detection results was discussed. The algorithm combining the adaptive subspace decomposition and the detection method based on local joint sparse index was proposed to improve the detection effect. With synthetic and real hyperspectral imaging datasets in the simulation experiment, the results show that the algorithms utilizing the new models could improve the effectiveness of the detection results to a certain degree, and different window designs have an impact on the results.

Turbulence is an important motion of atmosphere, which affects the imaging system seriously, and causes image distortion and blur. The randomly changing turbulence makes it difficult to estimate the devolution kernel and the atmospheric turbulence image restoration is ill-conditioned. we propose an atmospheric turbulence image restoration method based on priors of distinct edges. Firstly, the images are rectified. Then, we fuse the registered images and deblur the image to obtain the final image using a new method based on the distinct edge prior. In the single image deblurring schedule, we detect the distinct edges and predict the “sharp vision” of them, and use them to estimate the value of blur kernel. Finally, in order to restore the sharp images, we employ winner filter to deconvolute the images. The algorithm archives recovering much image details and sharpening edges and avoids the artifacts, which is of great practice.

LIDAR which plays a very important role in the cloud observation is one of the main instruments of cloud remote sensing. At present, most LIDAR is pointed vertically or to a certain angle. In this mode, sampling of cloud distribution is limited. A scanning LIDAR for cloud observation is constructed, in which a small solid-state laser is bearded on the receiving telescope. The telescope is fixed to the 3D scanning platform, so the whole system can work in different scanning modes in order to realize the monitoring of the whole sky clouds. Then the scanning LIDAR is used to the observation of cloud，and the result is compared with the existing RRML. The cloud measurement result shows that the scanning LIDAR can describe the distribution of the height and the optical depth with high spatial resolution, which reflects the advantage of scanning LIDAR in cloud detection.

To improve the performance of polarization navigation, a new method to calculate the heading angle using all-sky polarization pattern is proposed. Through analyzing the distribution feature of atmospheric polarization pattern, we propose a method to identify the sample points of solar meridian area, from which we can decide the azimuth angle of solar meridian in carrier coordinate by least square algorithm. The DOP distribution feature along meridian is used to distinguish the solar meridian and anti-solar meridian, thus solving the problem of ambiguity in acquiring the heading angle. And then the heading angle is figured out since we can get the azimuth angle in geographic coordinate by astronomic theory. The simulated results show that under sampling mode that we set, the error of heading angle is below 0.05°; under different simulated conditions namely different time of a day, different input heading angle and different amount of sampling points. When there is an error within 8° with angle of polarization, the average error of heading angle stay within 0.2°, thus proving the high precision and well adaptation of the method.

In order to improve the accuracy in thyroid tumor detection, the selection, crossover and mutation operation of genetic idea was introduced into the bat algorithm and a novel registration method of SPECT image and B-type ultrasound image was proposed based on genetic idea bat algorithm. First, according to the common contour characteristic of thyroid and tumor, improved gradient normalized mutual information was used for similarity measurement. Second, the genetic idea bat algorithm was used to optimize the space registration transform parameters. The experimental results show that improved gradient normalized mutual information has high locating accuracy of registration and less local extremum, genetic idea bat algorithm improves the rate and accuracy of convergence of registration. The method achieves desired effect in registration of SPECT image and B-type ultrasound image.

Super-resolution reconstruction plays an important role in adding the image details and improving the visual perception. In order to effectively exploit the effective information hidden in the image itself, we proposed a single image super-resolution reconstruction method based on self-similarity and sparse representation. The method combines sparse K-SVD dictionary learning and nonlocal means，which are used to add the effective information hidden in the same scale and across different scales structural self-similarity into the maximum a posteriori probability estimation framework by two different regularization terms. Then, a local optimal solution is obtained by using the gradient descent algorithm. The experimental results show that our method has a better improvement both visually and quantitatively.

In order to get the performances of the high-speed video processer in TDICCD mosaic system with high resolution, analyzed the control methods of the video processor parameters and the noise types and their reduction methods in high resolution TDICCD system, and designed an intelligent method of controlling to LM98640 based on FPGA. The functions of time-stretching quantization and sampling adjustment with high resolution were implemented, the noise level of the TDICCD imaging electronics system was decreased, and the SNR of the system was improved. By experiments comparison and analyzing of the image indexes, it is proved that the TDICCD image processing system based on LM98640 could satisfy higher specifications besides of increasing the pixel processing velocity and the image resolutions, and the final SNR of the system could be achieved at 47.4 dB.

To quickly and accurately grasp the spatial and temporal distribution of precipitation, for regional climate, hydrology and ecological environment is essential. The Yangtze River Delta is taken as the study area, extracting spectral characteristics of FY satellite imagery and performing characteristic analysis, combined with the receiving precipitation data, the characteristic set to describe the precipitation parameters was obtained and using the support vector machine superior performance of nonlinear regression put forward a kind of self-adaptive and self-learning precipitation estimation method. The results show that: satellite imagery is applied to interpretation cloud precipitation mechanism, combined with SVM, which can well express the nonlinear relationship of the Yangtze River Delta region between precipitation and cloud features. The results also demonstrate that correlation coefficient between values of modeling estimation data and receiving precipitation data is 0.85. It is shown that this method can estimate the precipitation in the region to play a role.