Xinhong Wang

A practical SNR estimation scheme for remotely sensed optical imagery

Signal-to-Noise Ratio (SNR) is one of the basic and commonly used statistic parameters to evaluate the imaging quality of optical sensors. A lot of SNR estimation algorithms have been developed in various research fields. However, one intrinsic fact is usually ignored that SNR is not a constant value, but a quantity changing with the incident radiance received by the sensor. So SNR values estimated on different images through commonly used method are not comparable due to their distinct intensity levels between the images. Here we proposed a normalized SNR estimation scheme which can be readily applied to remotely sensed optical images. With this scheme SNR values obtained from different images can be of comparability, thus we can easily evaluate the performance degeneration of the sensor with more sufficient reliability.

Earthquake prediction based on Levenberg-Marquardt algorithm constrained back-propagation neural network using DEMETER data

It is a popular problem that the mechanisms of earthquake are still not quite clear. The self-adaptive artificial neural network (ANN) method to combine contributions of various symptom factors of earthquake would be a feasible and useful tool. The back-propagation (BP) neural network can reflect the nonlinear relation between earthquake and various anomalies, therefore physical quantities measured by the DEMETER satellite including Electron density (Ne), Electron temperature (Te), ions temperature (Ti) and oxygen ion density (NO+), are collected to provide sample sets for a BP neural network. In order to improve the speed and the stability of BP neural network, the Levenberg-Marquardt algorithm is introduced to construct the model, and then model validation is performed on near 100 seismic events happened in 2008.

Progress of calibration and validation for quantitative remote sensing in China

Calibration and validation (Cal and Val) is one of the most important quality assurance means for satellite payload performance and data quality which has actually restricted RS applicable scope. It has aroused various attentions from academia and industries in recent few decades. The challenges include the lack of consistent RS assessment standard, the uncertainties introduced by atmospheric effect, as well as the gaps in non-synchronous measurements between satellite and field observation. As one of the countries which launched the largest number of earth observation satellites/payloads in last five years, China engaged to solve various challenges of Cal and Val for quantitative RS applications. Several reprehensive works were introduced, including the development of remote sensing technology standardization, the stepwise Cal and Val system, China’s Baotou comprehensive Cal/Val site, automatic in-situ calibration exploration, etc. All these works mitigated the uncertainties of RS measurement and enhanced the precision of quantitative remote sensing.

A stripe noise removal method of interference hyperspectral imagery based on interferogram correction

An Spatially Modulated Fourier Transform Hyperspectral Imager (named HSI) aboard on the Chinese Huan Jing-1A (HJ-1A) satellite has a spectral coverage of 0.459-0.956μm with 115 spectral bands. In practical, periodical and directional stripe noise was found distributing in the HSI imagery, especially at the first twenty shortwave bands. To fully utilize all information contained in hyperspectral images, it is demanded to eliminate the stripe noise. This paper presents a new method to deal with this problem. Firstly, possible sources of HSI stripe noise are analyzed based on interference imaging mechanism. Traditional noises, e.g. device position changes due to launch, non-uniformity in the instrument itself and aging degradations, are directly recorded at the focal plane array and thereafter in the interferogram. After inverse Fourier transform exerted on the interferogram, the spatial dimension of the interference hyperspectral image is restored with complicated and untraceable stripe noises. Therefore traditional image processing methods based on spectral image will not be effective for removing the HSI stripe noise. In order to eliminate this effect, a stripe noise removal method based on interferogram correction is necessary. Then, the implementation process of the interferogram correction method is presented, which mainly contains three steps: 1) Establish relative radiometric correction model of the interferogram based on relatively homogeneous ground scenes as much as possible; 2) Correct the response inconsistency of CCD array by carrying out relative radiometric correction on the interferogram; 3) Convert the corrected interferogram to obtain the revised hyperspectral images. An experiment is conducted and the new method is compared with several traditional methods. The results show that the stripe noise of HSI image can be more effectively removed by the proposed method, and meanwhile the texture detail of original image and the correlation among different bands are finely reserved.

A new method to estimate SNR of remote sensing imagery

The signal-to-noise ratio (SNR) of a remote sensing image is one of the most important indicators to evaluate the quality of the image, and also can reflect the SNR performance of a remote sensing payload to a great extent. Meanwhile, the SNR determines the information precision of a remote sensing image by which researchers could use the spectral characteristics to identify the surface features. Optical remote sensing images are usually contaminated by Gaussian white noise. Surface features often interfere with each other when imaging, which increases the difficulty of SNR evaluation. For heterogeneous region, the interference between different features is stronger and could not be removed easily. For homogeneous region, same features present the same or similar characteristics, showing as similar digital number (DN) values, so the interference between same features could be removed in some way. One of the ways to remove the interference between same features is to do subtraction operation between the adjacent row DNs or column DNs in homogeneous region. And the residuals, due to subtraction, are more indicative to the noises. This paper presents a novel method for SNR estimation of optical remote sensing images. Firstly, calculating the column residuals between the same features in homogeneous region. Secondly, doing subtraction operation to calculate the row residuals between the same features in homogeneous region. Thirdly, integrating the column and row residuals to evaluate the SNR. In this paper, the new method and a traditional typical method are used to estimate the SNRs of measured images. By analyzing the results of the two methods, we can find the new one is more stable and accurate. This method provides a new way to evaluate the SNR performance of optical remote sensing payload onboard.

A new method based on Spatial Dimension Correlation and Fast Fourier Transform for SNR estimation in remote sensing images

For optical remote sensing images which are contaminated by white Gaussian noise, in general, uniform features indicate the same spectral characteristic. Uniform features will present the same or similar digital number (DN) value with a certain band in imaging. Therefore, the DNs of the uniform features are highly correlated [1]. When dividing an image into small blocks to estimate noise standard-deviations (SDs) and distributing SDs into a number of bins with equal width, within the minimum to the maximum SD, the statistical curve of numbers of SDs in bins theoretically meets Gaussian distribution [2]. Combining the two features, we develop a new method for SNR estimation. Results of tests indicate the new method performs better than other ones and overcome some disadvantages of some typical methods.

Comparison of the land surface temperature and emissivity separation method for hyperspectral thermal infrared measur ement

Land surface temperature and emissivity separation (TES) is a basic problem in various thermal infrared (TIR) imagery applications. However, TES is essentially a typical nondeterministic inverse problem. From the late 1990s, a number of TES algorithms for hyperspectral TIR data have been developed based on various additional constrains. In this paper, different hyperspectral TIR TES methods, including spectral smoothness methods, downwelling radiance residual index (DRRI) method and linear spectral emissivity constraint (LSEC) method, were first recalled. Subsequently, the simulated hyperspectral TIR data were used to evaluate the performance of the TES method. The results shows that spectral smoothness method and DRRI method perform similarly, while the LSEC method may have better performance to retrieve land surface temperature and emissivity.

Estimation of the directional reflectance in Middle Infra-Red channel from SVISSR/FY-2C data

This work addressed the estimation of the directional reflectance in Middle Infra-Red (MIR) channel from the data acquired by the Stretched Visible and Infrared Spin Scan Radiometer (SVISSR) onboard Chinese geostationary Meteorological satellite FengYun 2C (FY-2C). SVISSR/FY-2C sensor acquires image covering the whole disk with a temporal resolution of 30 minutes. The MIR directional reflectance retrieval procedure can be seen as follows. Firstly, the atmospheric profiles data provided by European Centre for Medium-Range Weather Forecasts (ECMWF) were used to correct atmospheric influence with the radiative transfer code (MODTRAN 4.0). Secondly, the bi-directional reflectance in SVISSR/FY-2C MIR channel 4 (3.8 micron) was estimated from the combined MIR and TIR channel with day-night SVISSR/FY-2C data. Finally, a BRDF model referred to as the RossThick-LiSparse-R model was used to estimate the directional reflectance in MIR channel from the time-series bi-directional reflectance data. The results have been demonstrated that the method can be applied well to estimate the directional reflectance in MIR channel of SVISSR/FY-2C sensor.

An improved physical method with linear spectral emissivity constraint to retrieve land surface temperature, emissivity and atmospheric profiles from satellite-based hyperspectral thermal infrared data

In this paper, an improved method is proposed to simultaneously retrieve land surface temperature (LST), emissivity (LSE) and atmospheric profiles. This method employed the linear spectral emissivity constraint to efficiently reduce the number of retrieved variables. The proposed method was validated with some simulations. The initial guesses were derived from a neural network model. This method could greatly improve the accuracies of LST, LSE and atmospheric profiles. The RMSE of LST was decreased from 5.12 K (the initial guesses) to 1.59 K (the physical retrieved). The retrieved emissivity spectrum was in good agreement with the actual spectrum. An improvement of 1K in the tropospheric temperature was also been found. Those results showed that the proposed method is capable of improving the retrieval accuracies of land surface and atmospheric parameters with the remotely sensed thermal infrared data.

Appropriate spatial resolution analysis based on land surface heterogeneity

In order to meet the various demands of remote sensing applications, it is still worth being discussed that whether higher resolution is necessarily better. Actually, the spatial resolution should be carefully chosen according to application demands, for optimizing the balance of effectiveness and cost. In this paper the concept of “appropriate spatial resolution” is used, which should be the spatial scale with least data volume and most interested information, while not the highest precision of information. Characterization of spatial heterogeneity is the basis of analyzing spatial structure of the image under the given spatial resolution, and it is also the basis of selection of appropriate spatial resolution. Based on the wavelet variance method to characterize spatial heterogeneity, appropriate spatial resolutions of albedo, NDVI and surface radiation temperature on the same research area are given. Results show that appropriate spatial resolutions to various retrieved physical properties are different due to different grades of spatial heterogeneity. Proper scale range will be obtained to observe certain geographic quantity, by operation of the appropriate resolution selection.