Table of Content

20 August 2020, Volume 49 Issue 8

Previous Issue    Next Issue

Geodesy and Navigation

A new generation of global bathymetry model BAT_WHU2020

HU Minzhang, ZHANG Shengjun, JIN Taoyong, WEN Hanjiang, CHU Yonghai, JIANG Weiping, LI Jiancheng

2020, 49(8):  939-954.  doi:10.11947/j.AGCS.2020.20190526

Asbtract ( )   HTML ( )   PDF (25461KB) ( )  

References | Related Articles | Metrics

In this paper, a 1'×1' bathymetry model BAT_WHU2020 in the range of 75°S—70°N is constructed by using the latest version of the global gravity anomaly model derived from the multi-source satellite altimetry data and the shipboard depths. The accuracy of the model is analyzed and evaluated based on ship depths, existing models and multibeam soundings. The standard deviation of the difference between the proposed model and the ship depths in China Sea and its adjacent areas (104°E—160°E, 0°N—50°N) is about 70 m, which is equivalent to the accuracy of SIO V19.1 model, superior to ETOPO1, DTU10, GEBCO_08 model, and about 30% higher than the accuracy of BAT_VGG model published before, which shows that the method in this paper is reliable, and the data processing is accurate and the accuracy is high. The standard deviation of the differences between BAT_WHU2020 model and ship depths is about 50~65 m globally, and the ratio of the difference within ±200 m is greater than 95%. It is showed that the accuracy of BAT_WHU2020 model is equivalent to SIO V19.1, better than ETOPO1, DTU10, GEBCO_08 model, and improved about 27%~36% from the BAT_VGG model. Comparing to SIO V19.1 model, the standard deviation of the model differences is about 90~110 m, about 90% of the grid differences is within 200 m, and about 95% is within 300 m . Finally, the effects of crustal isostasy and high order terms in the Parker's formula on the accuracy of the results, the accuracy of the model compared to multibeam soundings, and the spatial resolution are discussed. It indicates that the spatial resolution of BAT_WHU2020 model is about 10~18 km, and the relative accuracy is about 5%~6% around the Mariana trench and Macquarie ridge.

Conditional variance stationarity evaluation method for GNSS ambiguity decorrelation

LU Liguo, LIU Wanke, LU Tieding, MA Liye, WU Tangting, YANG Yuanxi

2020, 49(8):  955-964.  doi:10.11947/j.AGCS.2020.20190417

Asbtract ( )   HTML ( )   PDF (10111KB) ( )  

References | Related Articles | Metrics

GNSS ambiguity decorrelation is to optimize the permutation order of conditional variance by integer transformation, so as to improve the search efficiency. One of the key problems is how to evaluate the relationship between decorrelation and conditional variance. Aiming at this problem, this paper theoretically analyzes the numerical relationship between decorrelation and conditional variance after sorting. It is found that the decorrelation performance is related to the stationarity of the conditional variance sequence. The stronger the decorrelation performance, the more stable the conditional variance sequence. So based on this theoretical basis, the conditional variance stationarity is proposed as an index to evaluate the performance of decorrelation. The results are verified by both simulation and actual test experiments, and the conditional variance trend graph as well as search time are also used to qualitatively and quantitatively evaluate the performance of decorrelation, to determine the rationality of the conditional variance stationarity. The experimental results show that the conditional variance stationarity proposed in this paper can more accurately and intuitively measure the performance of ambiguity decorrelation. The index defined in this paper reveal the essence of GNSS ambiguity decorrelation.

Torus approach in high degree gravity field model determination from GOCE satellite gradiometry observations

LIU Huanling, WEN Hanjiang, XU Xinyu, ZHAO Yongqi, CAI Jianqing

2020, 49(8):  965-973.  doi:10.11947/j.AGCS.2020.20200044

Asbtract ( )   HTML ( )   PDF (8352KB) ( )  

References | Related Articles | Metrics

Currently, space-wise method, time-wise method and direct-wise method are widely used to deal with the GOCE satellite real observations. Different from these conventional methods, this paper proposes one Torus approach. As one refinement of the reference model, the GOCE earth gravity field model complete to 200 d/o is recovered using 71-day GOCE satellite gradiometry observations with Torus approach. Firstly, a Butterworth filter with zero-phase combining with remove-restore is used to deal with the colored noise in GOCE satellite gravitational gradient observations. These observations are reduced from the real orbits to nominal orbits using Taylor series expansion and Kriging is used for gridding on the nominal orbits. Then, 2D-FFT and the block-diagonal least-square adjustment are performed to process the gridded observations, so that the GOCE earth gravity field model complete to 200 d/o named as GOCE_Torus can be realized efficiently. GOCE_Torus is assessed with the GPS/leveling data in China and USA, respectively. GOCE_Torus and the first generation model released by ESA have a similar accuracy. In USA, EGM2008 complete to 200 d/o reveals a similar accuracy with GOCE_Torus. But in China, GOCE_Torus is improved by 4.6 cm due to the contribution of GOCE gravity gradiometry data. With the application of Torus approach the satellite gravity field model can be recovered quickly and accurately, and therefore this approach is useful in mission design, error analysis and in-orbit evaluation of gravity gradient satellites.

Fuzzy clustering analysis method for optimal combinations of BDS triple-frequency signals

LI Yuzhao, YAN Haowen, WANG Shijie, YNAG Weifang

2020, 49(8):  974-982.  doi:10.11947/j.AGCS.2020.20190432

Asbtract ( )   HTML ( )   PDF (1866KB) ( )  

References | Related Articles | Metrics

Aiming at how to select the optimal carrier-phase linear combinations for triple-frequency BDS, a fuzzy clustering analysis method based on the theory of triple-frequency GNSS carrier phase combination is proposed. Firstly, the combination which satisfies the conditions of extra-wide-lane and narrow-lane combination and has different lane-number, ion-number and noise amplification factor is constructed, and the theoretical optimal combination with optimal parameters is assumed. Then, the extra-wide-lane combinations and narrow-lane combinations satisfying the conditions has been clustered and sorted by fuzzy clustering analysis method. Finally, according to the clustered sequence of each combination with the assumed theoretical optimal combination, the order of selection for triple-frequency BDS optimal combinations is determined. The results show that the fuzzy clustering analysis method achieves the selection of the optimal combined observations by clustering the combined observations.The results of geometry-free TCAR algorithm also verify the correctness of fuzzy clustering analysis results.

A wavelet neural network for optimal wavelet function to predict GPS satellite clock bias

WANG Xu, CHAI Hongzhou, WANG Chang, CHONG Yang

2020, 49(8):  983-992.  doi:10.11947/j.AGCS.2020.20190180

Asbtract ( )   HTML ( )   PDF (1347KB) ( )  

References | Related Articles | Metrics

To develop the accuracy for predicting SCB based on the the problem that the wavelet neural network (WNN) model fails to select the appropriate wavelet function according to the actual situation, an wavelet neural network for Optimal Wavelet Function based on Shannon entropy-energy ratio to predict SCB is proposed herein. The wavelet coefficients are obtained by carring on the continuous wavelet decomposition to the clock a once difference sequences. Then, the energy value and Shannon's entropy value of the wavelet coefficient are calculated respectively, and the “Shannon's entropy-energy ratio” (SEE) is taken as the evaluation index for the selection of the optimal wavelet function to induct select the most suitable wavelet function as the activation function of WNN model. Finally, the optimal WNN model is used to predict SCB, and the predicted results are compared and analyzed. The results show that the evaluation index can accurately guide WNN model to choose the appropriate wavelet function according to the actual situation of SCB, improve the prediction accuracy and applicability of WNN model, and enable the model to realize high accuracy SCB prediction.

Prediction of the satellite clock bias based on MEA-BP neural network

LÜ Dong, OU Jikun, YU Shengwen

2020, 49(8):  993-1003.  doi:10.11947/j.AGCS.2020.20200002

Asbtract ( )   HTML ( )   PDF (6041KB) ( )  

References | Related Articles | Metrics

The satellite clock bias is one of the important factors that affect the accuracy of navigation and positioning, so establishing a high-precision clock bias prediction model is of great significance to high-precision positioning. Aiming at the problem that satellite clock bias error accumulates by common models over time in short-term prediction, and the easy overfitting and instability of the traditional BP neural network, this paper proposed a model and algorithm of clock bias prediction based on BP neural network optimized by the mind evolutionary algorithm(MEA). First, original clock bias data made once difference to obtain the corresponding once difference sequences. Then, the initial weights and thresholds of the BP neural network were optimized by the mind evolutionary algorithm, the specific steps of using this model for the clock bias prediction were given. The multi-day GPS precision clock bias product data provided by the IGS station is used for experimental analysis. The article used the GPS data for the first 12 h of the day for modeling were listed, and made short-term clock bias prediction within 2, 3, 6 and 12 h. The results showed that the above four periods of prediction precision obtained by using the MEA-BP model were better than 0.36, 0.38, 0.62 and 1.56 ns, respectively. The fluctuation of the prediction error curve was small, and the prediction performance of the new model was better than the three traditional models, which showed the new model is better in practicability and stability in the short-term prediction of clock bias.

A simplex search algorithm for the optimal weight of common point of 3D coordinate transformation

GUO Yinggang, LI Zongchun, HE Hua, WANG Zhiying

2020, 49(8):  1004-1013.  doi:10.11947/j.AGCS.2020.20190409

Asbtract ( )   HTML ( )   PDF (2390KB) ( )  

References | Related Articles | Metrics

In order to improve the calculation quality of 3D coordinate transformation parameters, a robust method for weighted-common-point coordinate transformation method is proposed based on optimization algorithm. The minimum sum of weighted squared coordinate residual, which is the coordinate difference from the transferred coordinates to the known coordinates of common points, is taken as the objective function, and the Nelder-Mead simplex direct search algorithm is utilized to search the optimal weights combination of common points coordinates automatically in calculating coordinate transformation parameters. Taking the alignment and installation of particle accelerator magnets as a typical application scenario, simulated data and measured data are used to verify the proposed method. The results show that the algorithm can effectively reduce the weight of gross errors and poor-quality observations. Compared with the least square method and robust method, the sum of weighted squared coordinate residual of the proposed method is smaller, and the quality of coordinate transformation parameters is better. The proposed method can improve the solution quality of 3D coordinate transformation parameters, and is especially applicable to the situation that the priori precision is unknown and the quality of observation is poor.

Photogrammetry and Remote Sensing

Geometric registration of close-range optical image and terrestrial laser point cloud constrained by nearest surface

LI Cailin, WANG Zhiyong, YU Lulu, GUO Baoyun

2020, 49(8):  1014-1022.  doi:10.11947/j.AGCS.2020.20190146

Asbtract ( )   HTML ( )   PDF (6608KB) ( )  

References | Related Articles | Metrics

In this paper, a high-precision registration method based on the nearest surface for close-range optical image and terrestrial laser point cloud is proposed.Three-dimensional sparse point cloud is generated from optical images. Constrained by the local surface fitted by the laser points adjacent to the 3D sparse point of the image, a transformation model between the three-dimensional sparse point cloud and the three-dimensional laser point cloud is constructed by using collinear conditional equation. The high precision geometric registration of optical images and the laser point cloud is completed by iterative calculation. The method only needs initial registration parameter and does not need to perform feature extraction and segmentation on the laser point cloud data. In addition, the problem that it is difficult to accurately determine the correspondence points between two sets of points is solved effectively based on the surface constraint. The results of two sets of experimental data show that this method can effectively improve the accuracy of rigid registration algorithm, and can achieve higher registration accuracy.

Recursive estimation method of cubature Kalman filtering local polynomial coefficients for phase unwrapping

XIE Xianming, SUN Yuzheng, LIANG Xiaoxing, ZENG Qingning, ZHENG Zhanheng

2020, 49(8):  1023-1031.  doi:10.11947/j.AGCS.2020.20190385

Asbtract ( )   HTML ( )   PDF (4304KB) ( )  

References | Related Articles | Metrics

Recursive estimation method of cubature kalman filtering(CKF) local polynomial coefficients for phase unwrapping is proposed to retrieve unambiguous unwrapping phase from noisy wrapped phase. First, phase gradient information required is obtained using amended matrix pencil model (AMPM), and then the initial state estimation value of the polynomial coefficients is obtained. Finally, the polynomial coefficients are recursively estimated to obtain the unambiguous unwrapping phase by using the cubature Kalman filter. According to the density of the fringes of the interferograms and the signal-to-noise ratio (SNR) of the interferograms, the row-by-row (or column-by-column) scanning modes or the quality-guide strategy applied in traditional algorithms can be used to guide the cubature Kalman filter to unwrap the wrapped pixels along the suitable paths. The results with the simulated data and the measured data demonstrate that the algorithm in this paper can obtain robust solutions from noisy interferograms, with respect to some other similar algorithms.

Sparse hyperspectral unmixing algorithm supported by sparse difference prior information

ZHANG Zuoyu, LIAO Shouyi, SUN Dawei, ZHANG Hexin, WANG Shicheng

2020, 49(8):  1032-1041.  doi:10.11947/j.AGCS.2020.20190205

Asbtract ( )   HTML ( )   PDF (7155KB) ( )  

References | Related Articles | Metrics

Spectral library-based hyperspectral sparse unmixing technology has received attention in recent years, which uses spectral samples in the spectral library as endmembers and transforms the unmixing problem into a sparse representation problem. However, due to differences in the measurement environment, the actual endmembers of the hyperspectral image to be unmixed tend to differ from the corresponding spectral signatures in the spectral library. In this paper, an unmixing algorithm named spectral difference sparse constrained collaborative sparse regression is proposed. Firstly, we assume that the spectral differences have sparse property, and a spectral library correction model is established, which can make the spectral library be adaptively adjusted during the unmixing process; Then, the spectral library correction model is combined with the collaborative sparse regression unmixing model to establish a sparse unmixing model considering spectral differences; Finally, an iterative optimization solution based on the alternating direction method of multipliers is given. Synthetic and real hyperspectral data are used to verify the performance of different algorithms. The results show that the proposed algorithm is more effective than the compared algorithms in the presence of spectral library mismatches.

Power tower detection in remote sensing imagery based on deformable network and transfer learning

ZHENG Xin, PAN Bin, ZHANG Jian

2020, 49(8):  1042-1050.  doi:10.11947/j.AGCS.2020.20190356

Asbtract ( )   HTML ( )   PDF (8222KB) ( )  

References | Related Articles | Metrics

Power towers are important parts of power infrastructure, and it is indispensable to detect them. In view of the low precision and poor result of detection algorithms for power towers in remote sensing imagery, this study improves Faster R-CNN based on deformable network and transfer learning. And then we propose a new detection framework for power tower in remote sensing imagery. The framework includes a feature extraction sub-network and an object detection sub-network. The feature extraction sub-network uses deformable network model, which reconstructs the convolutional layer, to improve the model's feature extraction ability of the power towers with geometric deformation. The model parameters obtained from the feature extraction sub-network training are transferred to object detection sub-network, which accurately obtains position of power towers through RPN network,deformable area pooling and nms algorithms. Finally, the object detection sub-network is finely tuned and achieve high-precision detection for power towers in remote sensing image. The results show that in the test datasets AP_0.5, AP_0.6 and ACC are 0.886 1, 0.839 6, 0.894 8 which are at least higher 0.2 than SSD YOLOv3, Faster R-CNN. It can be seen from the comparative experiment that this method for power towers detection has great application potential.

A remote sensing image classification procedure based on multilevel attention fusion U-Net

LI Daoji, GUO Haitao, LU Jun, ZHAO Chuan, LIN Yuzhun, YU Donghang

2020, 49(8):  1051-1064.  doi:10.11947/j.AGCS.2020.20190407

Asbtract ( )   HTML ( )   PDF (22044KB) ( )  

References | Related Articles | Metrics

Traditional convolutional neural network almost cannot obtain satisfactory classification results of the remote sensing images due to the large differences in the size and spectral characteristics of the objects. In addition, the complex background environment will also bring interference to the classification. Aiming at this problem, the multilevel attention fusion U-Net (MAFU-Net) is presented. To enhance the correlations between different pixels and channels, the attention module is applied to extract and process semantic information at different levels, which further improves the classification performance of the network under complex background. In order to verify the effect of the proposed network in the classification of remote sensing images, the experiments were carried out on Vaihingen dataset of ISPRS, Beijing and Henan dataset of GF 2, respectively, and several different semantic segmentation networks are used for comparison. The experimental results show that the proposed network has fewer parameters and lower computational complexity, but can achieve higher classification accuracy in the least time, which means the network is highly practical.In addition, the feature visualization was fully utilized to analyze the classification performance of MAFU-Net and other networks, and the results also show that most deep learning network models are difficult to be deduced according to the accurate mathematical principles. It is also difficult to explain why a particular network fails in a particular dataset. Therefore, the further study or more advanced visualization and quantification criteria are required to analyze and evaluate specific deep learning models and network performance, then the more advanced model structure can be designed.

Summary of PhD Thesis

Study on wetlands with multi-scale based on multi-source remote sensing data in Dongting Lake basin

ZHANG Meng

2020, 49(8):  1065-1065.  doi:10.11947/j.AGCS.2020.20190350

Asbtract ( )   HTML ( )   PDF (662KB) ( )  

Related Articles | Metrics

Seamless modelling of global multi-scale terrain based on multi-resolution half-edges structure

HOU Shaoyang

2020, 49(8):  1066-1066.  doi:10.11947/j.AGCS.2020.20190399

Asbtract ( )   HTML ( )   PDF (665KB) ( )  

Related Articles | Metrics

Research on detection of spatiotemporal object inconsistency for land cover update

KANG Shun

2020, 49(8):  1067-1067.  doi:10.11947/j.AGCS.2020.20190418

Asbtract ( )   HTML ( )   PDF (671KB) ( )  

Related Articles | Metrics

Multi-GNSS real-time precise point positioning and undifferenced ambiguity resolution

CAO Xinyun

2020, 49(8):  1068-1068.  doi:10.11947/j.AGCS.2020.20200035

Asbtract ( )   HTML ( )   PDF (665KB) ( )  

Related Articles | Metrics

Constraint, prediection, combination and update in transformation problem of terrestrial reference frame

LIN Peng

2020, 49(8):  1069-1069.  doi:10.11947/j.AGCS.2020.20200067

Asbtract ( )   HTML ( )   PDF (665KB) ( )  

Related Articles | Metrics

Study on the extended theories of generalized total least squares and their applications in surveying data processing

WANG Bin

2020, 49(8):  1070-1070.  doi:10.11947/j.AGCS.2020.20200074

Asbtract ( )   HTML ( )   PDF (666KB) ( )  

Related Articles | Metrics

Study on remote sensing image preprocessing method and landslide feature identification of UAV in northeast Yunnan mountain area

XI Wenfei

2020, 49(8):  1071-1071.  doi:10.11947/j.AGCS.2020.20200081

Asbtract ( )   HTML ( )   PDF (665KB) ( )  

Related Articles | Metrics

Research on regional CORS applications in geoscience using undifferenced precise point positioning

SHEN Fei

2020, 49(8):  1072-1072.  doi:10.11947/j.AGCS.2020.20200152

Asbtract ( )   HTML ( )   PDF (684KB) ( )  

Related Articles | Metrics