最新刊期
卷 21 ， 期 5 ， 2017
-
摘要：O3 is the key composition of stratosphere. The problem on ozone and the effect on ozone by anthropogenic actions have always been the hot issue in the world. During attenuation circle of ozone, Chlorine composition is the most important medium. HCl is the storage of active chlorine. Therefore, monitoring HCl is significant. The development of limb sounding can supply three-dimensional information of O3 and HCl. Simulation interface of THz limb sounding is developed based on the THz limb sounding radiative transfer software ARTS. This study analyzes the low-limit height within the limb sounding in THz bands between 100 GHz and 1000 GHz. The frequencies, 235.7098 GHz and 625.9188 GHz, were selected as the central frequencies for O3 and HCl, respectively, based on the frequency of international available THz sensor radiometer. Then, the brightness temperature of limb sounding was simulated under different atmospheric backgrounds (standard atmospheric background and enhanced concentration of O3 and HCl at 5%, 10%, and 15%, respectively), different temperatures and moistures (standard background temperature at 0.5 K, 1 K, and 1.5 K increase, standard water vapor concentration increased by 5%, 10%, and 15%), and different bands and different sensor characteristics (different HPBW and different spectral resolution). The sensitivity analysis module of the simulation interface shows that O3 and HCl sensitive factors are further studied by limb sounding. The results are as follows: (1) Measuring O3above the tangent height of 8 km at frequency of 235.7098 GHz is highly possible, whereas to measure O3 below 8 km is difficult. For the measurement of HCl, the threshold height is 10 km, above which profiles of HCl can be available and below which profiles is difficult to measure. (2) At THz bands, the variety of temperature and moisture has little effect on the measurements of O3 and HCl. Therefore, THz bands are superior to infrared bands, which are influenced greatly by temperature. (3) Half power beam width and spectral resolution have a certain effect on the measurements of O3 and HCl.关键词：THz;ARTS(Atmospheric Radiative Transfer Simulator);HCl;O3;limb sounding- 293
- |
- 106
- |
- 2
发布时间：2021-06-07 -
摘要：In the past 50 years, the Ebinur Lake had deteriorated rapidly. The dry lake became a rich source of dust and salt, causing the region to face serious threat of salt dust disaster. At present, studies on salt dust pollution focus on conditions for salt storms, wind erosion intensity of Lake Bottom, and dust fall flux. Few studies investigated the regional air pollution caused by salt dust storms and the optical properties of salt dust aerosols. The temporal variability, vertical distribution, and potential diffusion characteristics of salt dust in Ebinur region were investigated using the ground-based and satellite remote sensing and its potential diffusion characteristics were analyzed using the HYSPLIT model to learn more about aerosol optical properties. The aerosol product data provided by the MODIS were used to obtain the overall evolution trend of dust storms in Ebinur Lake Basin. CALIOP data provided the vertical distribution of aerosol including total attenuated backscatter coefficient, volume depolarization ratio, and color ratio. The AOD and Angstrom exponent obtained by Microtops II sun photometer observations at the Jinghe and Wusu sites were used to investigate the aerosol properties in salt dust storms. The HYSPLIT model was used to simulate the salt dust diffusion trajectory of Ebinur Lake in spring. The degradation of the lake was the main reason for the occurrence of soil salinization and salt dust storms in Ebinur Lake Basin. The influence range of salt dust disaster in Ebinur region was mainly concentrated in the southeast of Ebinur Lake. The most active period of salt dust release and transportation was in spring. Salt dusts were mainly settled in Jinghe area, which was close to the dry lake bottom. The particle size of the aerosol particles in the atmosphere decreases, and the particle size increases with the increase in the distance from the dry lake dust source area. Thus, the area, which was far from the dust source, was not only affected by the Ebinur Lake salt dusts, but was also affected by the local dust source intervention. In the spring, the atmospheric aerosols were mainly in the form of coarse mode, and the dust aerosols were the main factor. The wind speed was the main factor affecting the local aerosol. Salt dust aerosols were mainly concentrated in the range of 0—2 km. The depolarization ratio was 10%—20%, the cumulative frequency was 28.538%, the color ratio was 0.4—0.8, and the cumulative frequency was 51.378%. With increasing height, the irregularity and size of the aerosol decreased, and little possibility was observed for the dust particles to be transported in other areas at high altitude. The aerosol particles mainly showed hydrophobicity when the dust storm occurred. However, aerosol particles, which are affected by the local aerosols, also showed a certain degree of hydrophilicity. The potential diffusion path to the eastern regions mainly affects Jinghe, Wusu, Kuitun, Shawan, Shihezi, Changji, Urumqi, Hami as well as Gansu, Inner Mongolia, Mongolia, Hebei, Liaoning, and other regions. The combination of ground-based and satellite remote sensing techniques to estimate regional aerosol optical properties can provide a scientific reference for the comprehensive understanding of salt dust disasters.关键词：salt dust aerosol;spatial and temporal characteristics;Microtops II sun photometer;potential diffusion path- 363
- |
- 73
- |
- 4
发布时间：2021-06-07 -
摘要：Inconducting direct assimilation experiments of microwave radiance data AMSU-A with the numerical weather prediction model T639-GSI 3DVAR system, the GSI monthly snow products cannot reflect the process of snow or snow melt in the middle and high latitudes of the northern hemisphere. The precision of the FY-3C snow daily products is higher than GSI snow monthly products. This study investigates the effect of different snow coverage values on the temperature field at different heights in the middle and high latitudes of the northern hemisphere. The numerical simulation results in assimilation process are based on the FY-3C SNC real-time snow daily products on January 20, 2016 and March 17, 2016, as well as the NCEP reanalysis data. In the assimilation experiment, we analyzed the distribution of winter snow over the northern hemisphere in January 2016 and that of spring snow over the northern hemisphere in March 2016. This study selects the largest snowpack of the FY-3C snow daily products in the two months as the test data and replaces the snow month products as the snow products of the underlying surface in the background field to reflect the process of snow or snow melt in the middle and high latitudes of the northern hemisphere. The 6-hour forecast results of the T639 model serve as the background of the assimilation experiment. Furthermore, the same assimilation data for all groups are used. The key differences are as follows. In group A, the winter snow month products are sold as the snow products of the underlying surface in the background field. In group B, the FY-3C snow daily products serve as the snow products of the underlying surface in the background field. Groups C and D are spring tests similar to groups A and B. In the assimilation experiment, we analyzed the distribution of winter snow over in the northern hemisphere in January 2016 and the distribution of spring snow over the northern hemisphere in March 2016. In order to reflect the process of snow or snow melt in the middle and high latitudes in the northern hemisphere, we selected the biggest snowpack of the FY-3C snow daily products in 2 months as the test data and replaced the snow month products as the snow products of the underlying surface in the background field. The 6 hour forecast result of the T639 model as the background of assimilation experiment. The same assimilation data were added for all groups. A key difference was as follows. In group A, the winter snow month products in business were the snow products of the underlying surface in the background field. In group B, Fy-3C snow daily products were the snow products of the underlying surface in the background field. Group C and group D are spring tests (similar to A and B). In terms of quality control, the regional differences of the temperature field using the GSI monthly products and FY-3C snow daily products evidently corresponded to the snow coverage differences between the two snow products. In winter and spring, we utilized the FY-3C snow daily products to replace the GSI snow month products as the snow products of the underlying surface in the background field. Quality control was also established in the assimilation system for the radiance data. The condition improved within 120 hours to a certain degree on the temperature field prediction at low to middle atmosphere levels from 1000 hPa to 600 hPa.关键词：quality control;GSI;AMSU-A data assimilation;FY-3C;VIRR snow daily products- 284
- |
- 74
- |
- 1
发布时间：2021-06-07 -
Detection and analysis of television frequency interference from an FY-3C microwave radiation imager
摘要：Observations of a satellite Microwave Radiation Imager (MWRI) are easily interfered by signals of active remote sensing because of their similar frequency. The primary source of interference over oceans is the reflected signal of geostationary television (TV) satellites by the ocean surface. The accurate detection of such interferences is largely important for the effective use of MWRI observations and an essential preprocessing step of the MWRI data onboard the FY-3C satellite. A Normalized Principal Component Analysis Method (NPCA) is used to detect the TV Frequency Interference (TFI) signals over oceans. High correlations exist among the different MWRI channels, but the RFI signal eliminates these correlations. Hence, NPCA can detect the TFI signal by employing this aspect. The TFI signals of the MWRI at 10.65 GHz horizontal polarization over oceans are distributed widely near the coastal areas of Europe, especially the English Channel and western parts of the Mediterranean Sea. The TFI signal definitively originates from the hot bird, and its intensity is related to the angle between the MWRI incidence and geostationary satellite TV signals reflection. The threshold of the TFI signals is defined to quantify the TFI intensity. The TFI signals at 18.7 GHz are observed over the offshore marine areas of North America. As expected, no RFI signal is detected near the coastal areas of China because the geostationary satellite TV frequency over China is different from those of MWRI channels. The NPCA method can detect TFI signals over oceans. The distribution of TFI signals are in Europe at 10.65 GHz and North America at 18.7 GHz. The intensity of the TFI signals is completely related to the Glint angle. TFI signals evidently affect the result of the retrieval production. Correcting and eliminating TFI signals will be pursued in future studies. Clear brightness temperature data will also be adopted in retrieval production and data assimilation.关键词：FY-3C;Microwave Radiation Imager (MWRI);ocean;television frequency interference;detection- 252
- |
- 71
- |
- 5
发布时间：2021-06-07 -
摘要：Land Surface Temperature (LST) is a key parameter for hydrological, meteorological, climatological, and environmental research fields. Accurate regional and global LST products can be obtained from thermal infrared remote sensing data. Himawari-8 is the next-generation of Japan geostationary meteorological satellite, which carries a new optical sensor called Advanced Himawari Imager (AHI), with significantly higher temporal and spatial resolutions. AHI has 16 observation bands, with spatial resolutions of 0.5 or 1 km for visible and near-infrared bands and 2 km for infrared bands. AHI can provide full disk images every 10 minutes, and can provide high temporal and spatial resolution LST information for many studies. The bands 14 and 15 of AHI can be used for LST retrieval by using the Split-Window (SW) algorithm. Thus, the objective of this paper is to propose a practical SW algorithm to retrieve LST from AHI data. Land Surface Emissivity (LSE) is one of the essential parameters for SW algorithm. SW algorithm is extremely sensitive to emissivity errors, and the sensitivity is significantly higher for direr atmospheres. A 0.005 error in emissivity will result in a LST error of 1 K or more under drier conditions. The ASTER Global Emissivity Dataset (GED) version 4 was adopted to calculate the LSE in this paper to improve the accuracy of emissivity in barren surfaces. The refined Generalized Split-Window (GSW) algorithm developed for MODIS was adopted to retrieve LST from the brightness temperature of AHI bands 14 and 15. MODTRAN 5.2, TIGR 3 atmospheric profile database, and ASTER spectral library data were used to create a simulation database to obtain the coefficients of the GSW algorithm. The coefficients were determined based on view zenith angle and atmospheric Water Vapor (WV) sub-ranges to improve the accuracy, and the WV was directly calculated using a simple method based on the brightness temperature of AHI bands 14 and 15. Two kinds of emissivity products were used to calculate LSE for AHI bands 14 and 15. The first product is the ASTER GED version 4 monthly product. The second is the MODIS MOD11C3 version 6 monthly emissivity product. The spatial resolution of the two products is 0.05°. Ground measurements collected from three Aerosol Robotic NETwork (AERONET) sites were used to validate the WV result. The root mean square errors (RMSEs) were 1.16 g/cm2, 0.94 g/cm2, and 1.23 g/cm2 for Baotou, Beijing-CAMS, and Hong_Kong_PolyU sites, respectively. The 2079 daytime and 2983 nighttime scenes of AHI images between June 1, 2015 and December 29, 2015 were used for LST retrieval. Ground LST measurements collected from four Heihe Watershed Allied Telemetry Experimental Research (HiWATER) sites and MODIS LST product data of the central points of seven lakes were used to validate the LST. The results show that the proposed algorithm demonstrates a reasonable accuracy, with RMSE less than 3 K, which has a comparable accuracy of current remote sensing LST products, such as MODIS, VIIRS, and FY-3B VIRR LST products. For LSE, ASTER GED v4 provides more realistic values of surface emissivity than MOD11C3 v6 because the emissivities are in accordance with the seasonal variation on NDVI. The MOD11C3 v6 typically provides constant values of emissivity, which was overestimated over the JCHM site. Thus, the LST was underestimated due to the overestimation of the emissivity. A practical SW algorithm for estimating land surface temperature from Himawari 8 AHI data was proposed based on GSW algorithm. ASTER GED v4 product was introduced to estimate the LSE for GSW algorithm. The LST result was evaluated with ground LST measurements collected in four HiWATER sites and the MODIS LST products, with RMSE of less than 3 K. The results also show that ASTER GED v4 product has higher accuracy than MOD11C3 v6 product in our study sites; thus, is more suitable in generating high accuracy LST product.关键词：land surface temperature;split-window algorithm;AHI;ASTER GED;atmospheric water vapor- 362
- |
- 174
- |
- 8
发布时间：2021-06-07 -
摘要：The diffuse attenuation coefficient (Kd) is an important property related to the penetration and availability of light underwater, which is of fundamental interest in studies of primary production, physical and biological process. The remote sensing estimation of diffuse attenuation coefficient can reveal the change of underwater light field, which is one of the main ways to obtain the diffuse attenuation coefficient. Models developed in the recent decades were mainly based on theoretical and numerical(radiative transfer)simulations. In this paper, a Monte Carlo calculation procedure has been developed for modeling of the penetration of light into turbid inland waters found at Meiliang Bay of Taihu Lake. We assume the water optical homogeneously, inherent optical properties (IOPs) do not change with depth, the water surface as a mirror, We can use the Monte Carlo method to simulate the movement process of photons in the water, energy distribution statistics of a large number of photons at different depths, calculate the downward diffuse attenuation coefficient value. We use several phase function with different backscatter fraction and overall shape as comparison. The results indicate that in the same condition of sun zenith angle and absorb coefficient (a)、scatter coefficient (b), simulated Kd values appear quite different when choose different scattering phase function. Simulated Kd values based on phase function applied to ocean or slightly turbid waters are less than the measured values. Within 450 nm, 550 nm, 650 nm band, the relative error between simulated Kd value with phase function that have a backscatter fraction of 0.03 and measured Kd value is the minimum. The phase function that have a backscatter fraction of 0.03 is more appropriate to develop a semi-analytical model based on inland turbid lake water than particular normalized phase function used in ocean water. We use the MC model to discuss the relationship between variation of sun zenith angle and phase function and Kd value. It turns out under the condition of high values of b/a, phase function become the main factor to effect the change of Kd value. When b/a values are greater than 9, the incident zenith angle changes from 10° to 80°, the rate of change in Kd values is less than 10%, so the impact of the variation of the sun zenith angle on the change of Kd value is negligible. Using a single wavelength scattering phase function to simulate Kd values at different wavelength. It’s found that in more turbid water, when the wavelength are less than 550 nm, simulated values are greater than measured values, in contrast, when the wavelength are greater than 550 nm, simulated values are less than measured values, which indicate that with the increase of wavelength, the scattering properties of particulate matter in water appear a variation trend that it’s weaken in the forward scattering and heighten in the backward scattering. When we develop the remote sensing model of Kd value for inland lake water type. We should consider the difference of scattering properties of water. Under different concentration of suspended matter and different wavelength, we should consider the spectral variation of the phase functions and use the phase function with the correct backscatter fraction.关键词：Monte Carlo;numerical simulation;diffuse attenuation coefficient;zenith angle;scattering phase function- 225
- |
- 79
- |
- 0
发布时间：2021-06-07 -
摘要：The fine-grained classification of hyperspectral image with small training samples is a major challenge for all kinds of classifiers. The signal-to-noise ratio of hyperspectral image is usually difficult to improve, and the magnitude of noise has a direct impact on classification results. Thus, noise reduction is one of the most important pretreatment measures for hyperspectral image classification. Employing the strong relevance between adjacent bands of hyperspectral images and the relevance between adjacent pixels in the space, a novel hyperspectral image classification method based on multi-level denoising and filtering is proposed. One two-phase Sparse and Low Rank Matrix Decomposition (SLRMD) method is introduced to remove the noise with high energy. At the first phase, the hyperspectral image is segmented, and each patch will use the SLRMD method to perform noise reduction based on the spectral correlation between the pixels within the same patch. At the second phase, the pixels of all patches will be merged together for noise reduction based on the spectral correlation of the adjacent bands of the hyperspectral image. Secondly, then principal component analysis (PCA) is introduced to remove the noise with low energy. Thirdly, Support Vector Machine (SVM) is used to classify the de-noised and dimension reduced hyperspectral dataset. Finally, guided filter is introduced to remove the " salt and pepper noise” in the classification map. We use the Indian Pines hyperspectral dataset as an example to verify the noise reduction effect of sparse and low rank matrix decomposition methods. The effect of image noise reduction is very obvious, and the bands after noise reduction show very strong correlation. The pixel spectrum of the original image contains a lot of noise information, especially in the first few bands and the last few bands, whereas the pixel spectrum of the low rank image becomes very smooth. The Spectral and Spatial De-Correlation (SSDC) and Local Variance Estimation (LVE) methods were used to evaluate the change of image quality before and after noise reduction. The signal-to-noise ratio of hyperspectral images is significantly improved after low rank matrix decomposition, especially at both ends of the spectral range. Two hyperspectral images, i.e., Indian Pines and University of Pavia, and some related classification methods are used for comparative experiments.The results show that the classification accuracy of our method is 25.85% and 13.2% higher than that of the SVM method, and 6.04% and 5.79% higher than the best method respectively. The two-phase SLRMD method proposed in this paper has better strong noise removal effect than the conventional SLRMD method, and it is more helpful to improve the classification accuracy of hyperspectral image. Moreover, SLRMD, principal component analysis and guided filtering, these three noise reduction and dimension reduction methods are highly complementary, so they should be used together to improve the signal-to-noise ratio of hyperspectral image and make the classification results more natural and smooth.关键词：hyperspectral image classification;feature extraction;small training samples;sparse and low rank;matrix decomposition- 387
- |
- 113
- |
- 9
发布时间：2021-06-07 -
摘要：Anomaly detection is one of the most important issues in hyperspectral remote sensing. However, traditional anomaly detection algorithms cannot be used for onboard real-time processing due to heavy computational load caused by the dimensionality curse of hyperspectral data. To implement onboard real-time hyperspectral anomaly detection, the following must be performed: (1) conduct the process in a causal progressive manner without any future data relative to the pixel under test; (2) output the result for each sample right after collecting it; (3) process the data with constant theoretical computational complexity. The present widely used algorithms generally employ QR decomposition or Woodbury’s Identity for real-time anomaly detection. However, their computational load is still extremely high. Also, they suffered serious numerical instabilities led by runoff error in the matrix inverting process. The main objective of this study is to further accelerate the real-time detection process and avoid the matrix inversion module to maintain numerical stabilities. In this work, we proposed a novel real-time sample-wise hyperspectral anomaly detection algorithm based on Cholesky decomposition. The background sample correlation matrix and covariance matrix were symmetric positive definite, which can be factored into a lower triangular matrix and its transpose. We modified the background suppression process into a process that finds the solution to a lower-triangular linear system based on this characteristic. Furthermore, the real-time process is significantly accelerated by virtue of rank-1 updates to the Cholesky factor of the background statistical matrix. Moreover, the numerical stabilities were maintained. Finally, we performed a three dimensional ROC (3D-ROC) analysis to evaluate the performance of real-time anomaly detection in terms of background suppression, detection power, false alarm, and the relationship between each other. An experiment on an actual hyperspectral dataset collected by Field Imaging Spectrometer System (FISS) revealed the following. (1) The proposed algorithm significantly reduced the computing time to process the incoming data. The theoretical computational load demonstrated high efficiency of the technique developed in this work. The time consumption for each incoming sample was reduced to 0.4%—0.65% of the time consumed by traditional QR decomposition-based algorithms, and 27%—33% of the time corresponding to Woodbury’s identity-based techniques. (2) The sample varying background suppression provided an acceptable visual inspection of the anomalies. Real-time processing prevented these weak signals from being suppressed by later detected strong signals. Additionally, 3D-ROC analysis could effectively evaluate the detection performance of hyperspectral anomaly detectors. The real-time sample-wise hyperspectral anomaly detector developed in this study is not only computationally efficient but also numerically stable, significantly contributing to onboard implementations.关键词：hyperspectral;anomaly detection;real-time processing;Cholesky decomposition;rank-1 updating- 239
- |
- 84
- |
- 5
发布时间：2021-06-07 -
摘要：Time series remote sensing plays an important role in forest monitoring, and the performance of the fitting method of time series Normalized Difference Vegetation Index (NDVI) determines the precision of phenology parameter estimation. However, in cloudy and rainy areas, obtaining cloud free remote sensing data for time series application is greatly difficult, leading to the difficulty in precisely fitting the time series NDVI. We present a knowledge-guided fitting method for sparse time series remote sensing data, given that the NDVI is underestimated by noise. First, we use prior knowledge and time difference method to identify and remove the noise, and then the Gaussian second-order model is used to fit the original data. Third, the iterative fitting is carried out by updating the weights of original data according to the fitting residuals. Finally, this process is repeated until a stable result is obtained. We use NDVI derived from Landsat 8 OLI as the data source to fit the forest data in Hangzhou, Zhejiang province. MODIS data are used as the referenced data to test accuracy of fitting result. The correlation coefficient between the fitting result of sparse time series data and the MODIS data fitting results is 0.92, and the difference of the day of max NDVI is five days, which indicates the proposed can be used to estimate the growth intensity, growth period, and other biological parameters effectively, the result is closer to the MODIS data fitting result compared with the state of art method. Our research provides the knowledge-guided fitting method, which is more effective than the other conventional methods, especially for the sparse time series data. This method reduces the human subjective influence and the excessive disturbance caused by the noise in the original data, making the fitting result more close to the actual fitting curve. The advantages of this method include less input parameters, fast converge, and the stability of fitting result. However, the proposed method suppresses the noise in the fitting processing to obtain relatively smooth fitting curve, which also may lead to the loss of the curve information; further improvements can be performed in the future.关键词：sparse time series data;iterative weighted;data fitting;Landsat 8;Gaussian model- 255
- |
- 91
- |
- 1
发布时间：2021-06-07 -
摘要：The capability of remotely sensed data acquisition is constantly improved. Thus, enhancing the automation degree for land cover classification at a large scale by remote sensing has become a key problem. However, present manual methods of selecting samples are becoming the bottleneck of automatic land cover classification. Many global and national land cover datasets based on remote sensing have been produced in the past two decades for different international or national initiatives. However, the rich knowledge implied in these products has not been fully exploited. The overall objective of this study is to set up an automatic land cover classification approach at a large scale by remote sensing through an automatic method of collecting land cover samples based on multisource datasets. The practical goals are to improve automation degree of land cover classification and enhance the accuracy of land cover classification. Henan and Guizhou provinces were selected as the study areas based on their types of land covers. First, the national land use database of China at a scale of 1∶100000 (CHINALC) and global land cover data (GlobleLand30) at a resolution of 30 m were selected as the data sources for the sample collection. Second, the initial sample areas were collected based on the spatial consistency analysis and heterogeneity analysis. Third, invalid samples were removed from the initial samples through the technology of sample purification. Finally, the Jeffries–Matusita distance was used to measure the classification feature separability of the samples between the different land cover types to prove the feasibility of the proposed method. The accuracy of the land cover product by the proposed method of sample collection was assessed and compared with the globe land cover product MCD12Q1. Experimental results show that the following: (a) The overall accuracy of the classification product through the proposed automatic method of sample collection based on multisource datasets was higher than that of the global land cover product MCD12Q1, which was classified based on the manual method of sample selection. The overall accuracy of the land cover classification product based on the proposed method was 78% in Henan province and 57% in Guizhou province, whereas that of MCD12Q1 was 74% and 55%, respectively. The Kappa coefficients of the former were 0.54 and 0.25, respectively, whereas those of the latter were 0.42 and 0.15, respectively. (b) Compared with the method of sample collection based on a single source land cover dataset, the proposed automatic method of sample collection based on multisource datasets had better classification stability and higher classification feature separability. The standard deviation of the Kappa coefficient and accuracy of products by 8-time experiments were less than 0.004. The classified results were also more stable. Unlike the method of sample collection based on a single source land cover dataset, the proposed automatic method of land cover sample collection based on multisource datasets not only improves the automation degree of land cover classification, but also enhances the accuracy of land cover classification.关键词：automation;samples collection;land-cover/land-use;classification;MODIS- 292
- |
- 260
- |
- 3
发布时间：2021-06-07 -
摘要：Image segmentation is a very common application in remote sensing, in which the number of classes is always given by users. To segment remote sensing images automatically, a sampling method which can transmit observed data into sample data is proposed based on the characteristics in Riemannian space. Therefore, this paper presents an unsupervised image segmentation algorithm which can automatically segment remote sensing images by sampling the detected data into samples. First, model the initial samples obtained by block sampling or artificial sampling through Gaussian probability distribution function (pdf). Second, to take the neighborhood system of the detected image into consideration, Gaussian pdf is also employed to depict the features of the pixel and its correlation between neighbor pixels. Then both the samples and the detected image are mapped to the Riemannian space. In the Riemannian space, the similarity between the points expressing the detected image and the points standing for samples are measured by geodesic, which is the least distance on the curve surface of a manifold. The nearest points standing for the detected image to each sample are transmitted to samples and then the models of the samples are updated according to the new ones. By continually sampling, the models of the samples are tending to their real models, which represents the real segmentation through sampling the detected data. In each sample process, only the nearest detected data are transformed into sample data to make sure the presented algorithm can distinguish different classes with different number of pixels in it. Geodesic employed in this paper evaluates the differences between the detected model and the sample model to improve the accuracy of sampling. The proposed algorithm is carried out on synthetic and real remote sensing images. Experiments on synthetic image shows the changes of samples both in the image and in feature space. Display of the sampling process demonstrate that the models characterizing each class trends to the real ones and the samples tends to the real data of each class in the feature space. Analysis on segmentation results of HMRF-CSA, GRM-FCM, neural network and the proposed algorithms on real remote sensing images validate the effectiveness of the proposed algorithm. The overall accuracy of the proposed algorithm can even reach to 98.9%, which is much higher than those of the compared algorithms. Experiments on synthetic image show the models of samples can tends to real ones, which validate that the proposed sampling process is able to fit the real distributions of each class. Experimental results on real remote sensing images demonstrate that the effectiveness of sampling operation can distinguish classes with different number of pixels. Quantitative and qualitative analysis on the segmentation results show that the presented algorithm can segment remote sensing images accurately and quickly. Besides, the proposed algorithm can be used both as an unsupervised image segmentation algorithm to realize the segmentation of remote sensing images or as sampling process in supervised image segmentation algorithm to provide reliable sample results.关键词：remote sensing image segmentation;unsupervised;sampling;mapping;Riemannian space- 215
- |
- 65
- |
- 1
发布时间：2021-06-07 -
摘要：The radiation quality of images is severely affected by variations in topography especially in complex mountain areas. Thus, topographic correction is a necessary pre-processing step for remote sensing image radiative correction. However, over-correction problems were easily found in the commonly used topographic correction method in steep mountain areas. We proposed a semi-empirical topographic correction method for steep mountain terrain to overcome the over-correction problems; it is the simple topographic correction method using estimation of diffuse light (SCEDIL). The SCEDIL model estimates the diffuse fraction through finding the horizontal sunlight and shadow pixels based on images and Digital Elevation Model (DEM) data. The three radiation components (direct, diffuse, and terrain irradiance) focused on the impact of the solar irradiance, reaching an inclined surface to recover the actual land surface reflectance. We consider that the solar direct radiation is the main source of incident radiation for horizontal sunlight pixels, and the incident radiation of horizontal shadow pixels come from diffuse and terrain irradiance. Therefore, the diffuse radiation can be estimated by the mean reflectance of horizontal sunlit and shadow pixels. The estimation of the surface actual reflectance is achieved using a two-step procedure. First, the atmospheric correction model (such as 6S or MODTRAN model) was used to calculate the diffuse fraction and irradiance components of horizontal surfaces. Second, SCEDIL was used to remove terrain effects. The GF-1 and Landsat ETM+ images were used to evaluate different topographic corrections (SCEDIL, C, and SCS+C models). Four validation methods were employed to assess the performance under different illumination conditions, as below: (a) visual comparison for different terrain correction models, (b) changes in the correlation coefficient between the incidence angle and the spectral bands, (c) changes in the Standard Deviation (SD) andmean of the reflectance of each spectral band, and (d) land cover classification accuracy. The results show that: (1) SCEDIL, C, and SCS+C corrections performed effectively for the slightly rough mountain area (average slopes<26°, average elevation<600 m). However, SCEDIL-correction produced better and smoother corrected images than SCS+C and C methods for the steep mountain area; the average slope is larger than 26°, and the average elevation is more than 700 m; (2) the SCEDIL method produced the lowest correlation coefficient between the incidence angle and the spectral bands for steep mountain, (3)SCEDIL-correction maintained more closer mean and lesser reduction in SD to original image bands than C-correction and SCS+C-correction; (4) SCEDIL-corrected images have the highest overall accuracy of classification and highest homogeneity within each land cover class using the Support Vector Machine classification method. Therefore, SCEDIL-correction was robust in terms of different terrains, and especially the steep region images in this study.关键词：topographic correction;SCEDIL-correction;C-correction;SCS+C-correction;diffuse fraction- 335
- |
- 149
- |
- 7
发布时间：2021-06-07 -
摘要：Automatically extracting skeleton lines from river features with high resolution remotely-sensed images can be of extremely significance for applications such as hydrological monitoring, land and ship transportation planning. The fractures of river skeleton lines cause a main concern about accurately extracting them from relative smaller river features in an integrated manner. Many scholars have investigated different methods to solve the fracture problems in road extraction, but rivers are wider and more irregular, so those methods are hard to use to connect fractures of river skeleton lines. By combining mathematical morphology and topology theories, this paper proposes a novel method to overcome the issue. The method works as follows. First, rivers are extracted by using a Normalized Difference Water Index (NDWI) combining with the closing operation of mathematical morphology to eliminate fractures caused by bridges and holes caused by ships and small islands as much as possible. Second, the Initial Skeleton Lines (ISLs) are extracted by Rosenfeld Algorithm. Third, the broken points of these ISLs are obtained by leveraging the binary-image of the ISLs, and then the Broken Point Pairs (BPPs) are connected by using an adaptive dilation-thinning algorithm. Finally, the fake connections are removed by using a fake connection removing algorithm, which uses the topology relationship between BPPs and the skeleton lines of rivers. The results show that the proposed method can be applicable to extract integrate skeleton lines for single rivers that are usually extracted with fractures. To a single river, especially in flat terrain, the skeleton lines extracted by the proposed method are more consistent to the practical situation than the river network extracted from DEM by Arc Hydro Tool using of shapes and positions. The fracture problem can be solved by adaptive dilation-thinning algorithm, which can find the most adaptive dilation factor to images. The accuracy of fracture connection was up to 85%. Skeleton lines of rivers extracted by the proposed method are highly consistent to the practical situation. To solve the fracture problem, an adaptive dilation-thinning algorithm was proposed, which overcomes the shortcoming of the min pair of breakpoints method that the operating rate is low while the error rate is high when dealing with large images, and also overcomes the shortcoming of traditional dilation-thinning method that deformation happens when dealing with long fractures. Removing the fake connections operation according to topology constraints greatly reduces the connection error rate. The method will provide a new idea for extracting natural linear features from remotely-sensed images.关键词：mathematical morphology;topology constraints;river skeleton;fracture connection- 282
- |
- 82
- |
- 2
发布时间：2021-06-07 -
摘要：Roads are greatly essential in a transportation system. The quality of road pavements has a crucial impact on the driving safety, comfort, and the cost of the roads. Therefore, timely monitoring the pavement conditions to guarantee the secure operation of the traffic system is significant. Currently, time-consuming field investigations and manual measurements are the conventional and main methods to detect and evaluate the pavement conditions. However, many of these methods are destructive to the road surface. Many forms of remote sensing data without destructive effect on pavement were introduced with the support of computer technology and remote sensing to detect pavement conditions, such as digital images, LiDAR, and Radar. This paper reviewed the current research status and problems in pavement condition detection based on remote sensing technologies. Remote sensing, as a non-intrusive method, has the advantages of wide spatial coverage and objective and repeatable data with high temporal resolution that can be analyzed by computer conveniently. Generally, the methods can be divided into four types, namely, the multi-/hyper-remote sensing, the thermal remote sensing, the microwave remote sensing, and three dimensional reconstructionbased on the categories of the sensors used in pavement monitoring. The multi-/hyper- remote sensing primarily utilizes the reflective spectral information of the pavement, which contains three specific methods, namely, image processing method, the method based on the variation of pavement brightness, and spectral modeling. The thermal remote sensing is based on the radiation characteristics of pavement, which contains two methods based on the variation of pavement temperature and emissivity information. A microwave device called Ground Penetrating Radar (GPR) has been widely used to detect the pavement defects by road departments. The three dimensional information of the road surface can be acquired using photogrammetry and LiDAR systems to measure the elevation information of the deteriorated pavement. An experiment using the Multiple Endmember Spectral Mixing Analysis (MESMA) and WorldView-2 satellite image was conducted to evaluate the potential of remote sensing in aging pavement monitoring. Results show that all methods are greatly potential in pavement condition monitoring. However, these methods still have some limitations in terms of complexity, accuracy, and robustness. High resolution remote sensing technologies, such as sub-meter satellite data and low-altitude UAV systems, can achieve large-range and rapid detection of the road pavement conditions in question. The results of the case study in Liangxiang area of Beijing show that the road pavement aging conditions could be detected appropriately and mapped with higher accordance with the practical pavement conditions. Therefore, some ground remote sensing methods have been applied successfully in pavement condition detection, such as the Pavement Management Systems (PMS), GPR, and so on. However, spaceborne and airborne remote sensing methods still encounter problems (poor robustness and low accuracy etc.)on pavement condition detection. For instance, the applicability of pavement spectral analysis and airborne or satellite data in pavement condition detection applications still require additional future works. Finally, a case study using satellite multispectral data to monitor asphalt pavement aging conditions is presented to demonstrate the usefulness of remote sensing technology in road pavement monitoring and assessment.- 303
- |
- 169
- |
- 6
发布时间：2021-06-07 -
摘要：The effect of forest biomass on carbon cycles has long been recognized.Therefore, an accurate assessment of forest biomassis required to understand ecosystem changes. This research uses vegetation indices and textural indices based on high spatial resolution Worldview-2 multispectral imagery to establish their relationship with forest AGB (Above Ground Biomass) and assess the accuracy of the estimation model. This research also explores the capability of spectral and textural information for AGB assessment at the Liang Shui National Nature Reserve, Northeast China. Remote sensing vegetation and texture indices were derived from high spatial resolution Worldview-2 multispectral data. We applied three different algorithms to extract the texture indices from the Worldview-2 data, including Gray Level Co-occurrence Matrix, Gray Level Difference Vector, and Sum and Difference Histograms. Six vegetation indices, namely, RVI, DVI, NDVI, EVI, SAVI, and MSAVI, were computed. The relationship among the above mentioned indices and 74 field measurements was established.However, the over fitting problems for the training regression model could occur due to the many input independent variables (i.e., vegetation indices and texture indices), which could decrease the robustness of the regression model. The random forest algorithm could avoid overfitting through the training process, so it was utilized to perform feature selection. Several optimal variables were selected to conduct the regression analysis.The support vector regression method was implemented to train and validate the AGB models. Results show that variables selection could better interpret forest AGB and obtain accurate predicted results. Comparisons between the two estimation models were made. The first model only applied vegetation indices, whereas the other model integrated vegetation and texture indices. The results also show that the accuracy of the vegetation indices model was lower than the vegetation+textural indices model (integrated vegetation indices with texture indices) at R2=0.69, RMSE=61.13 t/ha and R2=0.85, RMSE=42.30 t/ha, respectively. This research confirms that textural information could improve the accuracy of forest AGB estimation to a certain extent.关键词：aboveground biomass;texture indices;WorldView-2;high spatial resolution remote sensing image;temperate forest;Random Forest;SVR- 343
- |
- 184
- |
- 20
发布时间：2021-06-07
- Postal code：100190
- Tel：010-58887052 Email：nrsb@aircas.ac.cn
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备20021838号-9
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
0