Table of Content

20 May 2022, Volume 51 Issue 5

Previous Issue    Next Issue

Geodesy and Navigation

The positioning accuracy of the Lunar surface sampling and packaging mission of the Chang'e-5 probe

ZHANG Shuo, CHEN Liping, LI Tieying, YAN Yongzhe, DENG Xiangjin, GU Zheng, ZHENG Yanhong, MA Youqing, QI Chen, LIU Shaochuang

2022, 51(5):  631-639.  doi:10.11947/j.AGCS.2022.20210022

Asbtract ( )   HTML ( )   PDF (4558KB) ( )  

References | Related Articles | Metrics

The Chang'e-5 is China's first unmanned lunar probe to conduct the lunar surface sampling and return. It is necessary to carry out the high precision integrated geometry calibration for the sampling and packaging mission in order to ensure the positioning accuracy of the sampling and packaging mission. The combined adjustment model with the stereo constraints and the robot arm motion constraints is proposed. It is proved by the field test that the proposed calibration method has the high precision. The average positioning error of the 5.118 mm is obtained in the simulated sampling test and the 4.745 mm in the lunar sampling mission. It ensures that the Chang'e-5 probe can complete the automatic lunar sampling and packaging mission accurately.

Accuracy analysis of LEO satellites orbit prediction for precise position service

YUAN Junjun, LI Kai, TANG Chengpan, ZHOU Shanshi, HU Xiaogong, CAO Jianfeng

2022, 51(5):  640-647.  doi:10.11947/j.AGCS.2022.20210473

Asbtract ( )   HTML ( )   PDF (5085KB) ( )  

References | Related Articles | Metrics

Precise orbit prediction of low earth orbit (LEO) satellites is one of the important technologies for LEO navigation enhancement system. In this paper, we use a variety of algorithms to realize orbit prediction under different mission requirements. For LEO orbit prediction on ground, algorithm 1 processes orbit determination and prediction simultaneously. In algorithm 2, the discrete orbit points are fitted dynamically and then orbit integral is extrapolated. The average predicted user range error (URE) accuracy of GRACE-C satellite in 5, 10 and 15 minutes is 5.25, 5.67, 6.25 cm, that of HY2A satellite is 7.83, 8.69, 9.66 cm, that of SWARM-A satellite is 8.88, 9.22, 9.63 cm, and that of SWARM-B satellite is 8.49, 8.98, 9.63 cm. For LEO orbit prediction on board with limited calculation conditions, an orbit integral extrapolation algorithm with a single orbit point and simple dynamic models is used. Because this method mainly considers the perturbation of the Earth's central gravity and non-spherical gravity, the order of the Earth's gravity field has a significant impact on the accuracy of orbit prediction. 60 order gravity field is selected for LEO satellite with an average height of 500 km and 30 order gravity field is selected for LEO satellite with an average height of 1000 km, which can realize the prediction accuracy of about 10 cm for 10 min arc length.

Sliding window single-frequency real time precise point positioning algorithm with epoch constraints

YANG Kaichun, LV Zhiping, LI Linyang, KUANG Yingcai, XU Wei, ZHENG Xi

2022, 51(5):  648-657.  doi:10.11947/j.AGCS.2022.20210207

Asbtract ( )   HTML ( )   PDF (5603KB) ( )  

References | Related Articles | Metrics

To solve the rank deficiency problem of single-frequency precise point positioning (PPP) caused by calculating ionospheric delay simultaneously, an algorithm using multi-epoch observations is proposed in this paper. Based on the characteristic that the ambiguity is invariable without cycle slip, only one ambiguity parameter is set in the multi-epoch joint data, so as to solve the rank deficiency without the constraint of external prior information. In addition, the time correlation between parameters and observations are considered, and the square root information filtering (SRIF) with additional constraints to some parameters between epochs is utilized to overcome the ill-conditioned nature and obtain more reliable results. 14 days observation data from 15 IGS stations are used in the experiment,the static positioning accuracy is better than 3 cm, and the simulated dynamic solution is 1.5 dm. Compared with the single-frequency PPP method which estimates the ionospheric delay, the convergence speed is increased by 24%, and approximately reaches the level of the dual-frequency ionospheric-free PPP algorithm. The positioning accuracy is improved by 30%, and especially a significant increase in the vertical component.

Photogrammetry and Remote Sensing

The automatic determination method of the optimal segmentation result of high-spatial resolution remote sensing image

CHENG Jiehai, HUANG Zhongyi, WANG Jianru, HE Shi

2022, 51(5):  658-667.  doi:10.11947/j.AGCS.2022.20210423

Asbtract ( )   HTML ( )   PDF (12373KB) ( )  

References | Related Articles | Metrics

The existing methods cannot fully take into account the multi-band spectral information of remote sensing images, and ignore the multi-scale characteristics of geographical elements in remote sensing images. This study proposed an unsupervised evaluation method for automatically determining the optimal segmentation result of high-spatial resolution remote sensing image. This method generates the spectral information divergence based on information entropy, and uses the spectral information divergence to construct the indexes that can express the intra-segment homogeneity and inter-segment heterogeneity. Based on the constructed homogeneity and heterogeneity indexes, the strategy of "rough estimation + fine determination" is adopted to gradually obtain an optimal image segmentation result after multi-level optimization. The proposed method was carried out in three different underlying surface image areas. Experimental results demonstrate that the method can effectively automatically determine the optimal segmentation results of high-spatial resolution remote sensing images. Compared with existing methods, the optimal image segmentation results determined by the method have higher quality and are closer to the reference segmentation results.

Remote sensing image change detection fusion method integrating multi-scale feature attention

LIANG Zheheng, LI Xiao, DENG Peng, SHENG Sen, JIANG Fuquan

2022, 51(5):  668-676.  doi:10.11947/j.AGCS.2022.20200540

Asbtract ( )   HTML ( )   PDF (8937KB) ( )  

References | Related Articles | Metrics

Deep learning technology has become the mainstream method of remote sensing image change detection research. Existing change detection methods based on deep learning mainly obtain the change characteristics of a single scale. In the real scene, the scale of the change area is diverse. Therefore, we propose a change detection method of multi-scale feature attention fusion, which solves the multi-scale problem of change detection by focusing on multi-scale fusion strategy. We take advantage of the multi-scale characteristics of the feature pyramid network, the purpose is to enable the network to learn change features in different scales; meanwhile, in order to improve receptive field of network and exploit global information, atrous convolutional spatial pyramid module is introduced at the end of the feature extraction network; In addition, when different change features are fused, the change feature fusion module is used to control information flow to reduce the difference in feature fusion; Finally, the gating mechanism is utilized to perform weighted summation of the change feature maps predicted by different scales, and a high precision change feature map is generated. The proposed method can not only obtain multi-scale change features, but also use global information and precise spatial details to improve the spatial accuracy of the predicted feature maps. Experimental results show that our method has achieved competitive results on the change detection benchmark datasets CDD and LEVIR-CD, and the recall rate has increased by 6.58% and 5.26%, respectively.

Superpixel spectral features-based automatic fuzzy clustering segmentation for UAV image

TANG Xiaofang, ZHAN Zongqian, DING Jiujie, LIU Jiahui, XIONG Zirou

2022, 51(5):  677-690.  doi:10.11947/j.AGCS.2022.20210151

Asbtract ( )   HTML ( )   PDF (22604KB) ( )  

References | Related Articles | Metrics

Aiming at the problems of the existing popular fuzzy C-means clustering in image segmentation, such as the weak boundary attachment ability, the unstable segmentation process and the need to manually set the number of clusters, a super-pixel spectral features-based automatic fuzzy clustering segmentation for UAV image is proposed. Firstly, the watershed-based super-pixels algorithm with boundary advancing criterions are used to generate boundary adherent and compact super-pixels. Then extract the spectral features of super-pixels, and obtain the cluster number is automatically by rescaled density peak algorithm. Finally, an improved FCM method combining spectral features and hidden Markov random field is adopt to achieve high-precision super-pixels merging. Through qualitative analysis and quantitative evaluation, the results show that the proposed method can accurately locate the target boundary, obtain the optimal segmentation results and effectively improve the image segmentation accuracy.

Unsupervised remote sensing image scene classification based on semi-supervised learning

BAI Kun, MU Xiaodong, CHEN Xuebing, ZHU Yongqing, YOU Xuanang

2022, 51(5):  691-702.  doi:10.11947/j.AGCS.2022.20210270

Asbtract ( )   HTML ( )   PDF (8791KB) ( )  

References | Related Articles | Metrics

Self-supervised learning can extract features from remote sensing images without relying on sample labels, but supervised method is still required to complete feature classification. To overcome the shortcomings of feature classification process and automatically classify remote sensing image features, an unsupervised semantic clustering method based on semi-supervised learning is proposed. First, the features of remote sensing images are extracted using self-supervised learning to abstract the high-level semantic information contained in the images. Then, each sample's closest neighbors are found based on the feature similarity, and a linear classifier is trained by clustering similar samples into one class using the online clustering method. Finally, based on the clustering results, pseudo labels are generated for samples with high confidence to construct a label sample set, and the model is fine-tuned using a semi-supervised method. Four public remote sensing image scene classification datasets, EuroSAT, GID, AID and NWPU-RESISC45, were validated, and the classification accuracy reached 94.84%, 63.55%, 76.42% and 86.24%, respectively. The method presented in this paper combines the advantages of online clustering and semi-supervised learning, alleviates the problems of error accumulation and insufficient sample utilization in existing methods. It makes full use of the self-supervised features to train the classification model and complete the scene classification of remote sensing images without using labeled samples at all. It achieves the classification effect close to supervised learning and has good application value.

Seamless spherical video generation for multi-head panoramic camera(MPC)

HUANG Mingyi, WU Jun, GAO Jiongli

2022, 51(5):  703-717.  doi:10.11947/j.AGCS.2022.20210020

Asbtract ( )   HTML ( )   PDF (12760KB) ( )  

References | Related Articles | Metrics

Multi-head panoramic camera(MPC) is a kind of integrated optical system with ultra-wide field of view and can image an entire hemisphere onto a flat image plane through image stitching technique. But for actual scenes, the spherical video outputted by MPC often generates undesirable visual artifacts. This paper presents a method to generate seamless spherical video suitable for any scene. It mainly includes two aspects:① Estimation of MPC spherical projection parameters. Taking the tie points(pixels) detected in overlapping area as the observation values, those parameters are automatically estimated though minimizing the angle between the spherical projection center and corresponding spherical space points with the least square technique. As a result, the MPC video stitching errors caused by inaccurate calibration parameters, depth change of scene content and the inconsistency between MPC are effectively reduced; ② Establishment of TPS model for MPC spherical video generation. This model is established using the spherical re-projection geometry of MPC video pixels as global transformation and estimated with the salient pixels along selected seam line. As a result, video pixels of MPC sub-cameras can be directly mapped to the stitched spherical video with minimal pixel error in the overlapping area and the visual artifacts are completely eliminated simply by linearly mixing the pixels around selected seam line. The experimental results on stimulated and real scene show that, proposed method can generate seamless spherical Panoramic video for MPC used in various scene. In addition, proposed method is implemented efficient and can meet the requirements of high frame rate video output of MPC, and has good application value.

Location Services and GeographicInformation

Indoor and outdoor integrated pedestrian network construction based on crowdsourced data

ZHOU Baoding, ZHANG Wenxiang, HUANG Jincai, LI Qingquan

2022, 51(5):  718-728.  doi:10.11947/j.AGCS.2022.20210276

Asbtract ( )   HTML ( )   PDF (6762KB) ( )  

References | Related Articles | Metrics

The integrity and accuracy of the pedestrian road network is the key to ensuring pedestrian navigation services. Most of the current pedestrian road networks are constructed based on outdoor road facilities, lacking data support for indoor walkable paths, and cannot provide accurate and true optimal path planning services in navigation applications. In view of this, this article proposes a method for constructing an integrated indoor and outdoor pedestrian road network based on crowd-sourced data. It uses crowd-sourced trajectories recorded by smartphone positioning sensors and inertial sensors. The missing or drifting indoor walking data is first filtered, and then used the improved pedestrian dead reckoning (PDR) method calculates accurate indoor trajectories, and then uses Morse theory to generate a complete pedestrian road network covering indoor and outdoor pedestrian paths. In the experimental analysis, the pedestrian road network was constructed on the collected 260 walking trajectory data, and the real road network data was collected by high-precision measurement equipment for comparative analysis, and the experimental results were comprehensively evaluated with the Open Street Map data. Experimental results show that the method in this paper can accurately and completely generate an integrated indoor and outdoor pedestrian road network.

An indoor navigation network considering walking habits and its generation algorithm

HAN Litao, ZHOU Lijuan, GONG Cheng, ZHANG Aiguo

2022, 51(5):  729-738.  doi:10.11947/j.AGCS.2022.20210065

Asbtract ( )   HTML ( )   PDF (7063KB) ( )  

References | Related Articles | Metrics

An indoor navigation network is the basis of indoor path planning and navigation. The existing indoor navigation networks have some shortcomings such as unreasonable topological connection structure and unnatural geometric shape of generated paths. Accordingly, a novel indoor navigation network and its automatic generation algorithm are proposed in this paper according to the "short cut" behavioral characteristics formed by human beings for a long time and the security need of collision avoidance. The navigation network divides the indoor passable space into ordinary rooms and corridors. The network in one ordinary room is mapped as straight lines connecting the room node and the door nodes, and the corridor space with complex shape is subdivided into narrow corridor spaces and open corridor spaces according to spatial scale and convexity. The central axis of each narrow corridor space is taken as its corresponding route, and a complete graph is formed by connecting all door nodes and hatchway nodes in each open corridor space. Finally, door nodes are connected to the corridor route to form a complete indoor navigation network. The experimental results show that the proposed network model and its generating algorithm can reasonably partition the complex passable space according to indoor space scale and generate the navigation network structure consistent with spatial characteristics, which makes the shape of planned shortest paths more in line with human being's walking characteristics.

POI recommendation based on LBSN and multi-graph fusion

FANG Jinfeng, MENG Xiangfu

2022, 51(5):  739-749.  doi:10.11947/j.AGCS.2022.20210156

Asbtract ( )   HTML ( )   PDF (2178KB) ( )  

References | Related Articles | Metrics

As an important branch of the recommendation field, point of interest (POI) recommendation has always been favored by researchers. This paper proposes a POI recommendation algorithm based on location-based social network (LBSN) and multi-graph fusion, GraphPOI. It comprehensively analyzes the internal factors and external representations of users and POIs. First, it learns from the user-POI rating matrix to obtain the internal latent vector of users and POIs. Then, it constructs a user-POI interaction diagram according to the rating matrix, and obtains the representation vector of the POI in the user space and the representation vector of the user in the POI space. Next,it clusters the POIs according to their geographic locations to obtain the representation vector of the POI in the location space, and combines the representation vector of the POI in the user space to obtain the POI's external representation vector. At the same time, it models the information diffusion phenomenon in the user's social graph, captures the user's friendship to obtain the user's representation vector in the social space, and combines the user's representation vector in the POI space to obtain the user's external representation vector. Last, the internal latent vector and external representation vector of the user and the POI are combined to obtain the final vector representation of the user and the POI, which is input into the multi-layer neural network model for scoring prediction. The proposed model is verified on the Yelp dataset, and the results demonstrate that the method proposed in this paper can effectively improve the accuracy of POI recommendation.

Marine Survey

Automatic sea-land waveform classification method for single-wavelength airborne LiDAR bathymetry

WANG Dandi, XING Shuai, XU Qing, LIN Yuzhun, LI Pengcheng

2022, 51(5):  750-761.  doi:10.11947/j.AGCS.2022.20200314

Asbtract ( )   HTML ( )   PDF (11394KB) ( )  

References | Related Articles | Metrics

Sea-land waveform classification is a preprocessing procedure for airborne LiDAR bathymetry (ALB), which is related to the accuracy of subsequent signal detection and point cloud generation. However, the existing methods are not applicable for single-wavelength ALB systems and have low degree of automation. Thus, an automatic sea-land waveform classification method for single-wavelength ALB is proposed. The point cloud elevation features are firstly obtained by the detection of the first and last signals and the calculation of the point coordinates. With the mean water level elevation approximated by elevation histogram fitting, most of the sea and land waveforms are classified based on the elevation features. The remaining undefined waveforms are processed as single-signal waveforms with only the strongest signal retained. The signal features and energy distribution features are extracted from the waveforms, and a train sample set is automatically generated using the similarity of point cloud elevation features. The sea-land labels of the undefined waveforms are finally determined by a support vector machine classifier. The field data collected by a domestic ALB system (Mapper5000) are used to test the proposed method. The experimental results show that the initial classification based on the point cloud elevation features can quickly and accurately classify the waveforms away from the sea-land boundary, and the undefined waveform classification based on waveform features can automatically classify the waveforms closed to the sea-land boundary with the support of the self-established train sample set. Compared to the traditional methods, the proposed method can achieve high accuracy sea-land classification without the assistance of near-infrared channel and manual samples, and the overall accuracy and the accuracy of the areas closed to sea-land boundary reach 99.82% and 91.59%, respectively.

Construction and verification of improved error model of MBES

LI Fan, JIN Shaohua, BIAN Gang, CUI Yang, TANG Yulin, ZHANG Yonghou

2022, 51(5):  762-771.  doi:10.11947/j.AGCS.2022.20200498

Asbtract ( )   HTML ( )   PDF (8996KB) ( )  

References | Related Articles | Metrics

To evaluate the quality of MBES's data more accurately, an improved error model is derived that combines effects of echo detection methods and draft and heave correction errors, based on the constant gradient sound velocity tracking model and Rob Hare's classic error classification. The correctness of the improved model is verified through the decomposition and comparison of error sources and the comparison of the errors in the intersections; the example calculations show that the model can reduce the errors in the intersections of the CUBE surface by 2% compared with the Rob Hare model. The improved model considers more error factors, describes the error distribution law of measuring points more finely, and the error composition is more perfect. The improved model is helpful to find out the problem and strip the error source, analyze the magnitude of the error caused by each error source more accurately, and the research results provide theoretical basis and technical support for data quality control and data verification.

Offshore towed-streamer seismic positioning based on polynomial curve fitting

YU Wenkun, WU Peida, ZHANG Haonan, HU Guanghao, RUAN Fuming, DAI Wujiao, KUANG Cuilin

2022, 51(5):  772-780.  doi:10.11947/j.AGCS.2022.20200588

Asbtract ( )   HTML ( )   PDF (4157KB) ( )  

References | Related Articles | Metrics

In offshore towed-streamer seismic exploration, the streamers are generally smooth curves in shape. Traditional streamer positioning approaches often recursively compute the coordinates by numerical integration, in which observations such as acoustic distance and compass azimuth cannot be rigorously modeled in theory. To make full use of these observations, a rigorous adjustment positioning model based on polynomial curve fitting is proposed. The experimental results based on survey datasets demonstrate that the positioning accuracy of the seismic point estimated by the new algorithm is about 4.5 m, and the maximum accuracy improvement rate is 58.1% compared with the traditional algorithm. Therefore, the new algorithm can better ensure the progress of the high-precision offshore towed-streamer seismic exploration.

Summary of PhD Thesis

Study of annual mass balance estimation in the Tibetan plateau glaciers based on remote sensing albedo

ZHANG Zhimin

2022, 51(5):  781-781.  doi:10.11947/j.AGCS.2022.20200568

Asbtract ( )   HTML ( )   PDF (720KB) ( )  

Related Articles | Metrics

GNSS/SINS/Vision multi-sensors integration for precise positioning and orientation determination

ZHU Feng

2022, 51(5):  782-782.  doi:10.11947/j.AGCS.2022.20200569

Asbtract ( )   HTML ( )   PDF (726KB) ( )  

Related Articles | Metrics

Theory and method of spatiotemporal analysis, modeling and inversion of vertical GNSS coordinate time series based on independent component analysis

LIU Bin

2022, 51(5):  783-783.  doi:10.11947/j.AGCS.2022.20200573

Asbtract ( )   HTML ( )   PDF (727KB) ( )  

Related Articles | Metrics

Risk assessment of power transmission corridors in forestry area based on multi-source data

ZHANG Ruizhuo

2022, 51(5):  784-784.  doi:10.11947/j.AGCS.2022.20200577

Asbtract ( )   HTML ( )   PDF (712KB) ( )  

Related Articles | Metrics

End-to-end dense stereo matching based on full convolutional neural network

KANG Junhua

2022, 51(5):  785-785.  doi:10.11947/j.AGCS.2022.20200583

Asbtract ( )   HTML ( )   PDF (712KB) ( )  

Related Articles | Metrics

Research on automatic classification method of mobile laser point cloud data based on deep learning

HUANG Gang

2022, 51(5):  786-786.  doi:10.11947/j.AGCS.2022.20200599

Asbtract ( )   HTML ( )   PDF (718KB) ( )  

Related Articles | Metrics

Study on the theory and method in integrated navigation of multi-constellation GNSS/INS

CHAI Dashuai

2022, 51(5):  787-787.  doi:10.11947/j.AGCS.2022.20200608

Asbtract ( )   HTML ( )   PDF (727KB) ( )  

Related Articles | Metrics

The effect of the urban spatial structure on the spatio-temporal patterns of the urban thermal environment

YANG Chaobin

2022, 51(5):  788-788.  doi:10.11947/j.AGCS.2022.20200612

Asbtract ( )   HTML ( )   PDF (713KB) ( )  

Related Articles | Metrics