Search

The selection of road network is a fuzzy decision-making process. Aiming at problems such as noise, conflict and human decision-making in the cartographic generalization algorithm based on general case reasoning, this paper proposes a road network selection method which uses ontology to organize cases and conduct knowledge reasoning. In this paper, we extract expert cases from series scale map and describe them formally according to ternary model. To realize knowledge reasoning, these cases are used to build a domain ontology repository, reorganizing their representation format. Then, ontology rule extension helps to identify and eliminate noise and conflict cases. Additionally, with methods of semantic mapping and graph query, the semantic and geometric features of the road network to be selected are associated with the ontology repository to realize road network selection automatically.

The traditional spatio-temporal prediction methods could hardly model the complex nonlinear relationship of spatio-temporal phenomenons, thus they lack the ability to consider the influence of spatial multi-scale characteristics into the prediction results. In order to overcome this deficiency, a novel model of space-time network (MST-Net) is proposed in this paper, which transforms the regression problem of volume prediction into a discriminant model with time-space characteristics. The spatial and temporal characteristics of spatio-temporal data are extracted by multi-scale parallel convolution and gate recurrent unit respectively. Thus the extracted features are fused with the attention mechanism introduced to capture the long-term features. Finally, the prediction results can be obtained by using the full connection layers. In order to prove the reliability and validity of the model, the model is tested on two challenging social media sign-in datasets. The results indicate that the proposed model outperformed other algorithms in two prediction results evaluation indexes, namely the root mean square errors (RMSE) and mean absolute percentage errors (MAPE), which illustrate that the proposed method could achieve higher prediction accuracy and could better fit the nonlinear relationship of the space-time problem. The proposed model is suitable to predict the flow of human activities.

The BeiDou Navigation Satellite System (BDS) is a constantly-updated huge system with tightly-coupled space and ground segments and long construction period. Faced with the multiple challenges of highly complicated key technical scheme, high-intensity satellite deployment and launching tasks, and high difficulty in system stable operation, BDS has very high demands for the system's test and verification. This paper analyzes the test and verification tasks of BDS including system design and verification, space and ground segment integration, and simulation operation, and proposes the system architecture of the ground test and verification bed which equivalent to the full-system, full-scale and full-elements BDS. The bed is featured with synchronized work of the software and hardware platforms, equivalent operation of the simulation and real systems, and remote connection between the test bed and system segments, and solves the problems of multi-system simulation modeling, high-dynamic simulation as well as software and hardware synchronized simulation; realizes the whole process verification of the BDS design, deployment and operation, and also realizes the test and verification of the BDS new technical schemes such as inter-satellite links, BDS satellite autonomous integrity monitoring, and precise orbit determination. The bed for BDS is the only test and verification system in the field of satellite navigation that can realize all-states system simulation, and synchronized iterative evolution and operation with the real system, and can serve as important reference for the verification work of other space missions.

This paper begins with a brief review of the current state of development of precise point positioning (PPP) in recent years, with a focus on summarizing the latest research progress of several hotspots such as real-time rapid estimation of high-rate satellite clocks, multi-GNSS PPP ambiguity resolution, multi-frequency GNSS PPP models and ambiguity resolution, rapid initialization of PPP and PPP-RTK. Then, the evaluation of positioning performance of single/multi-GNSS PPP with latest observation of GPS, GLONASS, Galileo and BDS, especially the positioning accuracy, convergence time and time to first fix of BDS-2+3, is given. The results show that:PPP performance of BeiDou Navigation Satellite System is comparable with other GNSS. The convergence time and time to first fix of BDS-2 and BDS-3 combined PPP are about 20~30 minutes, positioning accuracy of static PPP in east, north and up directions is at millimeter to centimeter level, while that of kinematic PPP reach 5 cm in horizontal and 7 cm in elevation direction. Multi-GNSS combination can improve positioning accuracy and shorten the convergence time and time to first fix significantly, the positioning accuracy of ambiguity-fixed solutions can be improved by 14.8, 12.0 and 12.8% in the east, north and up directions compared to float solutions, respectively. The convergence time and time to first fix of multi-GNSS PPP can be reduced by 36.5 and 40.4% compared to that of GPS PPP.

Aiming at the lack of high accurate antenna phase correction (PCC) model for the BDS-2 IGSO and MEO satellites, in the BDS-2 and BDS-3 satellites joint precision orbit determination (POD) and precise point positioning (PPP), in this paper, an improved method for estimating PCV and z-offset parameters is used to refine the B1I/B3I ionosphere-free linear combination PCC model for the BDS-2 IGSO and MEO satellites. Compared with the results obtained using the ground-calibrated PCO model published by BeiDou official, the STD of SLR residuals obtained by the POD using the refined PCC model is reduced by 0.6~2.4 cm, and the improvement percentage is about 8.6%~33.3%. The floating PPP solution results, based on the refined BDS-2 satellite PCC model in this paper and the existing BDS-3 PCC model, have been significantly improved by 9.5 mm (37.2%) in the elevation direction.

Distortions of GNSS signals result in pesudorange biases. The biases are different for different receivers depending on frontend bandwidth and employed filters of receivers. Pesudorange biases are treated as a new error sources in this paper. Firstly, two pesudorange bias calibration calculation methods namely collocated receivers-based method and differential code bias (DCB) based method are proposed to separate pesudorange biases and other error sources. Then pesudorange biases are calculated, estimated and separated from other error sources using collocated receivers-based method. The STD of BDS estimated pesudorange biases series is approximately 0.1 m. The BDS estimated pesudorange biases keep constant with time. Pesudorange biases are also irrelevant with geographical locations of receivers. At 1.5 G band, B1I pesudorange bias is the largest. The pesudorange biases of BDS new signal B1C improve a lot compared with pesudorange biases of BDS B1I and are also much better than the GPS L1C/A. At other bands, pesudorange biases of GPS L2C are slightly larger than BDS B3I, followed by GPS L5C, pesudorange biases of BDS B2a are the smallest. The influence of pesudorange biases is also analyzed with measurements. Pesudorange biases are highly-correlated with total group delay (TGD). If the user receivers are quite different with those used for TGD calculation, the positioning accuracy will get worse.

The BDS-3 navigation satellite system currently provides five-frequency observations, so theoretically it can perform multi-frequency carrier ambiguity resolution (MCAR). This paper systematically studies the basic theory and method of the MCAR, including three-frequency (TCAR), four-frequency (FCAR) and five-frequency (FiCAR) carrier ambiguity resolution. Firstly, from the perspective of linear combination, basic mathematical models including TCAR, FCAR and FiCAR methods are given. Secondly, the high-quality signals and the optimal linear combinations under different conditions of baseline lengths-are discussed. In addition, single-epoch ambiguity resolution using geometry-free model is analyzed. Finally, experiments were conducted using the real BDS-3 five-frequency data. The results showed that the MCAR can effectively fix the ambiguities in single epoch, and the increase of the frequency number can significantly improve the success rates of the ambiguity resolution.

It is a common and effective method in engineering science to use the "closed-loop detection" idea to detect and correct the cumulative error of system measurement. This paper indicates the favorable closed-loop conditions in the global inter-satellite link (ISL) of BeiDou-3 navigation satellite system (BDS-3), and proposes to detect and analyze the closed residuals. On this basis, a network adjustment model of the closed residuals is constructed by which the correction of inter-satellite clock offsets is realized. Calculations based on the on-orbit measured data indicate that the obvious constant or periodic non-zero closed residuals really exist in the global ISL of BDS-3. Attributed to the correction of the closed-loop residuals of the ISL, the non-closed inter-satellite clock corrections are basically eliminated, the random noise of the inter-satellite clock corrections is reduced by 30% to 50% and the accuracy of the inter-satellite clock corrections are effectively improved. It is helpful forthe service capability of BeiDou.

With the increasing number of high orbit spacecraft, the use of high orbit GNSS technology to a certain extent avoids the limitations of traditional ground-based measurement and orbit determination, reduces the operation burden and maintenance cost of ground-based measurement and communication, and improves the autonomous navigation ability and orbit measurement accuracy of the satellite platform. Therefore, GNSS technology has significant research significance and application value for high earth orbit spacecraft. In this paper, the technology development of GNSS used in high orbit spacecraft is reviewed. Compared with low orbit spacecraft, the characteristics of GNSS in high orbit space are analyzed. From the three aspects of high orbit GNSS availability, on orbit technology verification and high sensitive receiver technology, the characteristics and development of high orbit GNSS autonomous navigation technology are summarized. The new tasks of high orbit post precise orbit determination and formation satellite relative navigation are analyzed. From the perspective of the future development of high orbit spacecraft, the future research prospect of high orbit GNSS technology is given.

Typhoon storm surges cause great losses to coastal cities every year. In recent years, a ground-based remote sensing technology using GNSS reflected signal has been proved to be able to monitor tidal level, called global navigation satellite system interferometric reflectometry (GNSS-IR), which can supplement the data of tide gauge. However, storm surges have strong destructiveness and occur quickly, the time resolution of GPS cannot meet the requirement to monitor this marine disaster. In this paper, three coastal storm surge events were monitored using the multi-mode and multi-frequency GNSS data of HKQT site in Hong Kong, China and BHMA site in Bahamas Islands. Firstly, we analyzed the quality of multiple system data. Then, multi-mode and multi-frequency GNSS-IR was used for monitoring three storm surges caused by hurricane "Dorian" in 2019, typhoon "Mangkhut" in 2018 and typhoon "Hato" in 2017.The sliding window least squares method was used to correct the results and compared with tide gauge observations. The experimental results showed that the accuracy of "Dorian" is better than 14 cm, and the accuracy of "Hato" and "Mangkhut" are both better than 9 cm. Compared with GPS, it could improve the accuracy and the time resolution of the results, which is helpful to record the whole process of tide rising, peak and falling in storm surge. And it can play an important role in the study of marine disasters.

Chinese BeiDou navigation satellite system (BDS) and Galileo system can provide the services of quad-frequency observations. In this paper, we assess the BDS/Galileo quad-frequency precise point positioning (PPP) models, i.e., PPP model with two ionosphere-free combinations (QF1), PPP model with one ionosphere-free combination (QF2), undifferenced uncombined PPP model (QF3) and ionosphere-constrained undifferenced uncombined PPP model (QF4), by comparing the dual-frequency ionosphere-free PPP model (DF). The equivalence of the QF1, QF2 and QF3 models is theoretically demonstrated by the equivalence principle. The static and simulated kinematic PPP performances are evaluated and investigated with one-month period observations from the network stations and the kinematic PPP performances are verified with a kinematic experiment in the campus. The results show that the pseudorange noises of BDS-3 B1C and B2a signals are larger than the B1I and B3I signals and the pseudorange noises differences for the Galileo quad-frequency signals are not obvious. The performances of the QF1, QF2 and QF3 models are basically consistent for the static and simulated kinematic PPP. By adding the external ionospheric constraint, the quad-frequency PPP performances are affected. Compared to the QF1, QF2 and QF3 models, the mean convergence time of the static BDS (BDS-2+BDS-3) model are reduced by 4.4%, 4.4% and 5.4%, respectively. The mean convergence time of static Q4 model increases 16.8 minutes when compared to the QF3 model. Compared to the dual-frequency PPP, the quad-frequency kinematic PPP performances are obviously improved. The three-dimensional positioning accuracy of BDS and Galileo QF4 models are improved by 11.4% and 31.4%, respectively, when compared to the QF1 models. Furthermore, the BDS/Galileo kinematic PPP models perform better than the single-system solutions.

The attenuation of GNSS signals in soil is of great significance for the related research of using GNSS signals to measure soil moisture. In this paper, for the first time, the attenuation of BDS (BeiDou navigation satellite system) and GPS (global positioning system) signals in the soil was studied through experiments. In the experimental design, the GNSS antenna was placed into the soil, then the soil thickness and moisture above the antenna were continuously changed to collect the power attenuation data of the GNSS signal. Finally, these data were used to retrieve soil moisture in order to validate the GNSS signal attenuation model. Experimental results show that soil can significantly attenuate GNSS signals. The greater the soil moisture value and thickness value is, the more severe the attenuation is. In the case of clay type soil and soil moisture of 0.15~0.30 cm3/cm3, the GNSS signal power has been attenuated to be undetectable by the GNSS receiver when the soil thickness reaches 21 cm. Further retrieval of soil moisture based on the GNSS signal attenuation model was carried out, the results show that the attenuation model is more accurate when the soil thickness is larger than or equal to 10 cm and when the satellite elevation angle is larger than 50°. And under this situation, the root mean square error of soil moisture retrieval using BeiDou B1 signal and GPS L1 signal is less than 0.04 cm3/cm3 and 0.09 cm3/cm3, respectively.

BDS-3 is a global satellite navigation system independently built by China. Its performance level and performance comparison with other satellite navigation systems have an important impact on the follow-up promotion and application. In this paper, the signal in space range error (SISRE) is used as a key performance index of the system. Taking the multi-system precise orbit and clock offset provided by GFZ as the standard, the comparison and evaluation method of satellite orbit, satellite clock offset and SISRE is given. Based on the measured data of three months from January to March 2020, the accuracy improvement of BDS-3 relative to BDS-2 is verified, and the performance comparison between BDS-3, GPS, Galileo and GLONASS is analyzed emphatically. The results show that the accuracy level of BDS-3 is significantly higher than that of BDS-2 both in satellite orbit and in satellite clock offset. The orbit accuracy of BDS-3 in the R, T and N direction is 0.07 m, 0.30 m and 0.26 m respectively, which is at the optimal level among the four global systems. The satellite clock offset accuracy is 1.83 ns, which is basically the same as that of GPS, superior to GLONASS, but slightly worse than Galileo. In terms of the signal in space range error, if only orbit error is considered, BDS-3 SISRE(orb) is averagely 0.08 m. Next, Galileo SISRE(orb) is 0.26 m, GPS SISRE(orb) is 0.57 m, and GLONASS SISRE(orb) is 0.98 m. If the orbit and clock error are considered comprehensively, the average SISRE of BDS-3 is 0.50 m, which is slightly lower than 0.38 m of Galileo, better than 0.58 m of GPS, and significantly better than 2.35 m of GLONASS.

The BDS triple-frequency carrier phase ambiguity resolution is greatly affected by the residual of atmospheric error between long range stations, and GEO satellites are also unfavorable for carrier phase ambiguity resolution relative to geostationary. Taking advantage of the relatively stable signal path of GEO satellites and the influence of atmospheric delay which do not change with the space position of satellites, the constraint of atmospheric delay errors more in line with actual conditions for GEO satellites are carried out. Using the linear relationship of carrier phases ambiguities of GEO satellites B2 and B3 to reduce the effect of single-difference ionospheric delay error between stationson the ambiguity candidates, the candidates of carrier phases ambiguity of B2 and B3 areselected. The linear relationship of triple-frequency phase ambiguity which does not include the influence of errors makes evaluation of the ambiguity candidates, and the ambiguity search space can be constrained. The ambiguity candidates of the GEO satellites are used to determine the residual change of the ionospheric delay between adjacent epochs, and the parameter estimation of GEO satellites is more constrained in accordance with actual situation.The method of long range BDS triple-frequency carrier phase ambiguity resolution with actual atmospheric delay variation constraints and the integer ambiguity constraints of GEO satellite is studied.And determining the ionospheric delay constraint value by using the candidates of ambiguity between epochs is proposed, adjusting the constraint value of the random walk between GEO satellite according to the actual situation.The experimental results show that the method in this paper can improve the efficiency of the triple-frequency carrier phase ambiguity resolution and the accuracy of positioning.

The geodesy discipline has been evolving and constantly intersecting and merging with other disciplines in the last 50 years, due to the continuous progress of geodetic observation techniques and the continuous expansion of application fields. The background and roles of the geodesy development are described in this paper at first, then the formations of the geodetic disciplines are sorted out as well. Furthermore, the basic growth of the geodesy subject is described from the progress of observation techniques, sensors and platforms, the cross-discipline formation is analyzed from the expansion of application fields. And the future development trends are analyzed from national requirements and science developments. The sub-disciplines for geodetic subject are suggested at present stage, based on the requirements of The National Natural Science Foundation of China and the development status of the geodesy, which may be referenced for geodetic scientific research topic selection and fund application.

The TH-2 satellite system is China's first microwave surveying satellite system based on interferometric synthetic aperture radar technology, and is also China's first short-range formation satellite system. It is the second microwave interferometric surveying satellite system after the German's TanDEM-X system in the world. A method to solve the problem of absolute ambiguity in the interference phase by designing dual-frequency imaging is proposed for the first time in the world, which completely got rid of the dependence on ground control data. The system works in the X band, with the resolution of 3 m and a solar synchronous orbit of 500 km. It is composed of two equal satellites. It adopts a technical system of satellite formation in different orbits and a bistatic radar transceiver mode. It can measure the global digital surface models and acquire radar orthophotos in a short time. In this paper, through the selection of the interference baseline system, satellite formation configuration and radar transceiver mode, the technical system of TH-2 satellite is proposed. The engineering design is carried out from three aspects:system mission, main performance and composition. The implementation of the project is expounded from the three stages of the overall demonstration, the key technical research and verification, and project development. Finally, the on-orbit test and verification of the satellite system is introduced. Test results show that all indicators have reached the requirements of engineering design, products' precision is equivalent to TanDEM -X system and can meet the 1:50 000 scale mapping accuracy. These verify the TH-2 satellite' engineering design ideas correct and the method of project implementation reasonable and feasible.

BDS reference stations typically store and transmit raw observations directly and make them available to users, which can lead to two disadvantages:first, there is no hierarchical authorization mechanism for users. Any user can solve the precise spatial, time and atmospheric datum information of the reference station through the original observation measurement, which poses a potential threat to the safety of the reference station, second, the strong correlation between the original observation measurement leads to a large amount of data, in BDS receiver observation appears network, uninterrupted, high sampling and multi-band trend, its observation data is a massive explosion, to the data storage and transmission have brought great pressure. Therefore, this paper proposes to use state and residual to express the raw observation data of BDS reference station. Compared with the common data services based on international protocols such as RINEX, Compact RINEX and RTCM, the advantage of this method is that it cannot only provide hierarchical authorization service of datum information for users, technically guarantee the security of data services, but also significantly reduce the amount of transmission data and user-side computation. At the same time, the new expression is equivalent to the original data and does not reduce the accuracy of the final calculation results. This method can be used to establish BDS observation data storage and transmission protocol with independent intellectual property rights, and provide technical support for satellite-based broadcasting services of high-precision data.

In the past decade, independent component analysis (ICA) has been gradually introduced into the component extraction of Earth's time-varying gravity field. Because of the independence of the extracted components, ICA solves the key problem that the extracted components are only irrelevant when using principal component analysis (PCA). The FastICA algorithm based on negative entropy is a common ICA method. The algorithm converges faster than the gradient algorithm, and the robustness of negative entropy is better. In this paper, the above algorithm is used to explore the relationship between the water storage components and precipitation components of the Qinghai-Tibet Plateau. The research indicates that there is a 7-month delay effect on the changes in water reserves on the Plateau compared to precipitation. It shows that in addition to seasonal frozen soils and glaciers, and the precipitation from the monsoon climate, there are also objective delays caused by regional water exchange. In view of the characteristic signals, the changes in annual average water reserves between central Qinghai, western Tibet, and central and eastern Tibet have reversed. It is pointed out that in the context of accelerated melting of frozen soil and glaciers caused by global warming, the regional water reserve changes due to different climate and flow regions. In addition, the changes of water reserves in the above areas are also different due to the north and south plate tectonics of the Qinghai-Tibet Plateau. Therefore, through the above studies, the feasibility of using the FastICA algorithm to analyze changes in regional water reserve components and their related factors on a large scale has been proven.

The classical three-stage forest height geomatical inversion method is easily affected by the assumption of the amplitude ratio of ground-to-volume scattering (GVR) and terrain slope. To address these problems, from the perspective of survey adjustment, the S-RVoG (slope-random volume over ground) based nonlinear complex least squares forest height inversion method is proposed in this paper. On the one hand, it does not need to hold the GVR assumption. On the other hand, it can take into account the terrain slope effect by adopting the S-RVoG model as the adjustment model. Three scenes of P-band PolInSAR data acquired from ESA BioSAR2008 campaign are used to construct two groups of single baseline tests for forest height inversion. The results show the RVoG-based nonlinear complex least squares method can obtain better forest height results than the three-stage geometrical method in a single baseline configuration. The proposed S-RVoG based nonlinear complex least squares method can further improve the accuracy. The improvement reaches a stand-level mean of 18.48% for slopes greater than 10°.

The increasing demands and diverse missions in earth observation lead to the complexity of satellite area coverage analysis, and satellite area coverage algorithms or applications cannot provide real-time response and online services that can handle high concurrent requests. Current analysis results cannot be visualized in real time, which increases the difficulty of data cognitive. To solve these problems, this paper proposes an efficient real-time analysis method and visualization service for satellite area coverage, which is combined with flow processing technology. First, a multi-mode judgment strategy is adopted to calculate the satellite area coverage parameters efficiently. Then, the continuous computing tasks are segmented to implement real-time data processing. Finally, based on streaming processing technology, this paper proposes an online service for real-time calculation and visualization. By using the data which contains three satellites and two ground areas to verify the algorithm, the experimental result shows that the method in this paper can not only ensure the accuracy under different sensor orientations, but can also greatly improve the computational efficiency. It can also provide real-time visualization of the analysis results in a timely and efficient manner.

For hyperspectral image classification problem of small sample, this paper proposes a depth of less sample learning algorithm, this algorithm through the simulation of the small sample classification in the process of training is to train the depth 3D convolution neural network feature extraction, the extraction of characteristic with smaller class span and large spacing between classes, more suitable for small sample classification problem, and can be used for different hyperspectral data, has better generalization ability. The trained model is used to extract the features of the target data set, and then the nearest neighbor classifier and support vector machine classifier are combined for supervised classification. Three groups of hyperspectral image data of Pavia university, Indian Pines and Salinas were used in the classification experiment. The experimental results showed that the algorithm could achieve a better classification accuracy than the traditional semi-supervised classification method under the condition of fewer training samples (only 5 marked samples were selected for each type of feature as training samples).

Cloud and snow detection is an important part of satellite remote sensing image processing, and also a key step for its following analysis and interpretation. In this paper, a simultaneous cloud and snow detection method for satellite remote sensing images based on ResNet and DeepLabV3+ fully convolutional neural network is proposed. The ResNet50 backbone is used, and the DeepLabV3+ network structure is optimized and improved according to the characteristics of cloud and snow on TH-1 remote sensing image. The ELU activation function, Adam gradient descent method and Focal Loss function are used to speed up convergence and improve segmentation accuracy. The network is trained and tested with the cloud and snow image dataset of TH-1 satellite. The experimental results show that it has strong robustness compared with Otsu method, and the detection accuracy of proposed method exceeds FCN-8s and original DeepLabV3+ network, meanwhile the detection speed of proposed method is faster than original DeepLabV3+, which can be applied to a variety of different remote sensing images through some adjustment and has favorable application prospects.

The pattern structure of buildings in the map generalization is a key factor to be considered. When there are spatial graphic conflicts among subgroups of buildings in different patterns, the generalization of these buildings becomes more complex. The generalization of straight linear pattern buildings is dominated by typification on large scale topographic maps. This paper proposes a progressive typification method for building subgroups of straight linear patterns with potential spatial graphic conflicts. The proposed method can progressively deal with the spatial graphic conflict between building subgroups in multiple straight linear patterns, and retain the important straight linear pattern structure of building groups. The position, shape, size and orientation of the typical building can be determined automatically with the restriction of visual graphical constraints expressed by the building. This paper also studies the automatic identification method of straight linear pattern of building subgroups based on the spatial proximity graph of building groups, and analyzes the adjacent relationship and intersection relationship between these straight linear patterns. Finally, the buildings on the 1:5000 map as a test data were generalized into 1:25 000 map, which verifies the availability and effectiveness of the algorithm proposed in this article.

Data storage model is an important part of database model. This paper aims at bottleneck problems in developing relational spatial database and using HBase to manage spatial vector data. Based on the analysis of relational spatial database storage model, the rules fortransforming the relational database storage mode into HBase storage mode are applied to spatial vector data management, a method of transforming spatial vector data relational storage schema into HBase storage schema is proposed, and a HBase storage model is designed for spatial vector data. The model is optimized and improved by using HBase's features such as entity nesting, denormalization and schemalessness.The experimental results show that the query efficiency of the storage model designed in this paper is better than that of the HBase-based spatial vector data storage model in the absence of auxiliary indexes.

The international GNSS monitoring and assessment system (iGMAS) and the multi-GNSS experiments (MGEX) receive BeiDou navigation signals on different frequencies. In this paper, orbit determination and clock estimation based on two ionosphere-free combinations of observation (B1/B3 and B1/B2) from BDS triple-frequency data is proposed. In this method, data of both networks from BeiDou 2nd generation (BDS-2), BeiDou 3rd generation test system (BDS-3e) and BeiDou 3rd generation global system (BDS-3g) can be processed uniformly in one program run, which can effectively improve data usage at one run, thereby improving parameter estimation accuracy. The GPS and BDS data of iGMAS and MGEX were collected for several days for testing. The results show that for the BDS-3e+BDS-2+GPS joint orbit determination, the orbit overlap RMS of the BDS MEO is 9.5 cm after using the triple-frequency two-combination method, which is 29% higher than the accuracy of traditional dual-frequency processing. The new method introduces a code bias parameter related to three frequencies of each involving BDS-2 satellite. The result shows values of the parameter are quite stable for continuous one-day solutions, which proves that the model is reliable. The new method is used for BDS-3g+BDS-3e+BDS-2+GPS joint orbit determination. The 6 BDS-3g MEO satellite orbit overlap RMS is 14.5 cm, and the clock error overlap RMS is 0.43 ns, which is equivalent to 15.1 cm and 0.49 ns of BDS-3e. The BDS precise point positioning (PPP) test was carried out. The results show that after adding the precise orbit and clock difference of the six MEOs of BDS-3g, the horizontal positioning accuracy of the station is 39.6 mm, and the zenith accuracy is 37.8 mm, which increases positioning accuracy by 11.1% than BDS-2 and BDS-3e satellites only.

The carrier phase integer ambiguity resolution between reference stations is very important in BDS network RTK. The algorithm of determining the BDS triple-frequency carrier phase ambiguity between reference stations at single-epoch is studied in this paper. Firstly, the linear relations of triple-frequency carrier phase ambiguity between the reference stations are deduced. Then the ambiguity candidates of B1 and B2 carrier phase are selected by ambiguity candidates of B1 carrier phase ambiguity and the linear relationship of dual-frequency carrier phase ambiguities, and the same as the ambiguity candidates of B1 and B3 carrier phase. The ambiguity candidates of triple-frequency carrier phase are controlled and fixed by the linear relationship between triple-frequency carrier phase ambiguities without errors. In order to reduce the influence of tropospheric delay on the constraint ability of linear relations between multi-frequency carrier phase ambiguities, the method of selecting a reference satellite with altitude and azimuth is used by the spatial correlation of atmospheric errors. The results of test indicate that the carrier phase integer ambiguities of triple-frequency between reference stations can be fixed at single-epoch, and the algorithm is simple and efficient.

Due to large volume of data, uneven spatial distribution of point densities, and complex scene, it is difficult to automatically extract and delineate road boundaries from mobile laser scanning (MLS) point clouds. This paper proposed a method for extracting road boundaries from MLS point clouds based on Snake. Different from the traditional Snake model defined within 2D image, we modified the Snake model directly based on 3D point clouds. Firstly, we developed a way of initializing Snake curves using pseudo-trajectory data to match multi-type road boundaries. Then, the Snake’s energy function was designed with internal, external and constrained energy terms derived from Snake curve points with their neighborhood points. Finally, the road boundaries were precisely extracted by minimizing the Snake’s energy function. Experiments were undertaken to evaluate the validities of the proposed algorithm with three different urban scene datasets acquired by different MLS system. Quantitative evaluations on the selected MLS datasets indicate that the Precision, Recall and F1-Measure of road boundaries extraction results were over 97.62%, 98.04%, 97.83%, respectively, which validate that the proposed method proved a promising and robust solution for regular and irregular road boundaries extraction in complex urban environments.

In 3D reconstruction, mesh refinement is generally applied to deal with noise and lack of details in triangular mesh built from dense point cloud. The existing variational refinement methods optimize the photo-consistency of the initial mesh by utilizing all the image data, but ignore the redundancy of image information and the impact of view quality on mesh refinement to some extent. In this regard, this paper proposes the strategies of master view selection and slave view selection, so as to improve the efficiency and quality of mesh refinement. Firstly, Markov random field is constructed by combining image gradient magnitude and contour detection to select master view for each triangular facet, and then slave view is selected according to the corresponding observation condition for each master view. Afterwards, we calculate the norm-weighted photo-consistency between master view and slave view, and finally surface energy function is minimized by using gradient decent method to obtain the refined mesh. The experiments show that the proposed method can recover more fine-scale details, meanwhile shorten time and increase accuracy of refinement, which confirms the validity of the proposed method qualitatively and quantitatively.

Using deep convolutional neural network (CNN) to intelligently extract buildings from remote sensing images is of great significance for digital city construction, disaster detection and land management. The color difference between multi-temporal remote sensing images will lead to the decrease of generalization ability of building semantic segmentation model. In view of this, this paper proposes the attention-guided color consistency adversarial network (ACGAN). The algorithm takes the reference color style images and the images to be corrected in the same area and different phases as the training set and adopts the consistency adversarial network with the U-shaped attention mechanism to train the color consistency model. In the prediction stage, this model converts the hue of the images to that of the reference color style image, which is based on the reasoning ability of the deep learning model, instead of the corresponding reference color style image. This model transforms the hue of the images to be corrected into that of the reference color style images. This stage is based on the reasoning ability of the deep learning model, and the corresponding reference color style image is no longer needed. In order to verify the effectiveness of the algorithm, firstly, we compare the algorithm of this paper with the traditional image processing algorithm and other consistency adversarial network. The results show that the images after ACGAN color consistency processing are more similar to that of the reference color style images. Secondly, we carried out the building semantic segmentation experiment on the images processed by the above different color consistency algorithms, which proved that the method in this paper is more conducive to the impro-vement of the generalization ability of multi-temporal remote sensing image semantic segmentation model.

In this paper, we propose an evaluation model of geographic information data classification based on relative exponential entropy. The internal relative exponential entropy of class and external exponential entropy among classes are designed to respectively quantify the agglomeration level of the data within each class and the discrete level among classes. With these two indexes, the relative exponential entropy of geographic information data classification is constructed. We implement the classification of the geographic information data and relative exponential entropy calculation of the classification in the Python platform, based on which the experiments are conducted. Six datasets of 5 classical distributions and a census dataset are classified with 5 frequently-used classification methods and the relative exponential entropies of the classifications are calculated. The experimental results demonstrate that when facing different distributed data the relative exponential entropy index could well indicate the optimal classification method. Meanwhile, the relative exponential entropy could reflect the tiny differences of different classification methods. The relative exponential entropy proves to be valid for the evaluation of geographic information data classification.

Firstly, the importance of stochastic model in precise positioning is demonstrated from the perspectives of parameter estimation, accuracy evaluation and quality control. Then, based on the single-difference functional model, the rigorous variance component estimation (VCE) method is used to allow the estimation of satellite-specific variances, cross correlations between two arbitrary frequencies, as well as the time correlations for phase and c+ode observations per frequency. The influence of the stochastic model on baseline precisions and the overall statistics was subsequently analyzed. The results show that the observation precisions of the BeiDou user receiver are overall elevation-dependent for phase and code of all frequencies. It is recommended to use the elevation-dependent exponent weighting model; there are different degrees of correlation between the three frequency phase observations, and the cross correlation between other types of observations is not obvious. The time correlation between phase and code observations of different frequencies is obvious, and attention should be paid to high-precision positioning. Furthermore,the baseline precisions that used the correct stochastic model match the theoretical ones very well for the three baseline components. The paper provides support for users to correctly understand the three types of satellite observation of the BeiDou system and correctly apply the BeiDou system.

Aiming at the problems of nonlinear gray difference, speckle noise and different imaging viewpoints in SAR and optical image registration, this paper presents a SAR and optical images registration using uniform distribution and structure description-based ASIFT. In the proposed algorithm, firstly, the guided scale space is constructed by guided filter to achieve noise suppression and edge preservation. In the feature extraction stage, the phase congruency is utilized due to the nonlinear gray difference, and combined with scale space gridding to extract from images for the uniform feature points. In the feature description stage, the consistency gradient magnitude and orientation of SAR and optical image are calculated by extended phase congruency method, which improves the accuracy of the main orientation and descriptor. At last, Optimal-RANSAC is used to establish feature descriptor matching to achieve effective registration. The simulation experiment and analysis on four pairs of real images show that the proposed algorithm has more accurate registration accuracy than SAR-SIFT and traditional ASIFT.

The positioning accuracy of the ground spot footprints is highly dependent on the positioning accuracy of the laser pointing angle. The centroid extraction and its variation law of space-borne laser spot are of important significance to the analysis of laser pointing angle. Firstly, based on the laser profile array(LPA) image data of ICESat/GLAS, this paper uses the gray-scale first-order matrix method to extract the centroid of LPA, with an accuracy better than 0.3 pixels, and the relative positioning accuracy of LPA is better than 0.11 pixels . next, Fourier transform and Fourier series fitting are used for periodic detection and modeling analysis of LPA centroid coordinates changes. The results show that the centroid coordinates of the laser spot image have four obvious periodic changes: 1.83×10-4 Hz,3.36×10-4 Hz,5.19×10-4 Hz and 6.71×10-4 Hz. The correlation R2 of fitting results reached 0.86, and the fitting accuracy was up to 0.4″, which was better than 0.13 pixels, and a better result was obtained. It can provide reference for the data processing of GF-7 and the subsequent satellite laser altimeter.

Influenced by the performance of imaging spectrometer and the distribution of complex ground objects, hyperspectral images have a large number of mixed pixels. Traditional learning-based unmixing methods are shallow models, or lack of comprehensive use of spatial and spectral information. This paper proposes a collaborative deep model with multi-dimensional convolutional network. Using multi-dimensional convolutional network can make full use of multi-dimensional semantic information, which is better to estimate hyperspectral mixed pixel abundance with small samples. The method augments training data, constructs three kinds of convolutional neural networks: spectral dimension, spatial dimension and cube dimension; the method designs fusion layer to concatenate features with three kinds of convolutional neural networks, and to “end-to-end” estimate of mixed pixel abundance; the model uses batch normalization, pooling and dropout to avoid over fitting phenomenon. The experimental results indicate that the introduction of our proposed method can extract spatial-spectral feature information more effectively. Compared with other convolutional network unmixing models, the accuracy of the estimated mixed pixel abundance is significantly improved.

Compared with spaceborne and airborne SAR, DEM extraction using vehicle-mounted SAR data can effectively shorten the update cycle of DEM data and reduce the cost. Due to the variable structure of vehicle-mounted SAR platform, interferometric calibration needs to be repeated, this paper proposes a DEM extraction method of vehicle-mounted interferometric SAR with dual antennas based on a single control point. This method corrects the phase, slope distance and elevation based on the precise coordinate information of a control point and realizes the iterative solution of elevation. It does not need to set up many control points for interferometric calibration, effectively simplifying the field work. The simulation results show that high accuracy of DEM extraction can be guaranteed when the estimation of baseline length reaches the accuracy of mm and the estimation of baseline inclination also reaches a high accuracy. Using vehicle-mounted interferometric SAR data with dual antennas obtained in wuhan, hubei province in 2018,the comparative test of DEM extraction was carried out using this method and multiple control point method based on interferometric calibration. The results show that the elevation error of this method at the checkpoints is 0.301 8 m, and multiple control point method is 0.258 4 m, the DEM results of two methods are highly consistent in the areas with high coherence.

Electric taxis have been demonstrated with the promotion of electric vehicles. Compared with internal combustion engine vehicles, electric taxis spend more time in recharging, which reduces the taxi drivers’ intention to use. Reinforcement learning is applicable to the sequential decision-making process of taxis driver. This paper presents the double deep Q-learning network (DDQN) model to simulate the operation of electric taxis. According to the real-time state of taxis, DDQN will choose the optimal actions to execute. After training, we obtain a global optimal electric taxi service strategy, and finally optimize the taxi service. Using real-world taxi travel data, an experiment is conducted in Manhattan Island in New York City, USA. Results show that, comparing with the baseline methods, DDQN reduces the waiting time for charging and the rejection rate by 70% and 53%, respectively. Taxi drives’ income are finally increased by about 7%. Moreover, the results of model parameter sensitivity analysis indicate that the charge speed and the number of vehicles have greater impact on drives’ income than the battery capacity. When the charging rate reaches 120 kW, electric taxis achieve the best performance. The government should build more fast charging station to improve the revenue of electric taxis.

Since the 21st century, with the rapid development of cloud computing, big data, internet of things, machine learning and other information technology fields, human beings have entered a new era of artificial intelligence. The subject of photogrammetry has also followed the tide of the new round of scientific and technological revolution and developed rapidly into the brand-new generalized photogrammetry and entered the era of integrated intelligent photogrammetry. Its carrier platform, instruments and data processing theories as well as application fields have also changed significantly. The multi-sensor and multi-level integrated stereo observation technologies from spaceborne, airborne and terrestrial platforms have been greatly developed. In this paper, the novel concept of generalized photogrammetry is first put forward, and its subject connotation, development characteristics and some key technologies and applications are discussed in details. Under the brand-new generalized photogrammetry framework, data acquisition presents the characteristics of multi-angle imaging, multi-modal collaboration, multi-time integration, multi-scale linkage, while data processing presents the trends of multi-feature coupling, multi-control constraints, multi architecture processing, and multi-disciplinary intersection. The all-round development and intelligent service of the general photogrammetry still need to make greater breakthroughs in the aspects of spaceborne, airborne and terrestrial multi perspective or multi-modal image processing, intelligent information extraction and monitoring, combined 3D modeling with point cloud and image, autonomous control of unmanned system, visual inspection of intelligent manufacturing system, etc. Finally, new theories and technologies from real-time or quasi real-time intelligent geometric processing of multi-source remote sensing datasets to information extraction and intelligent service need to be established, which will make a well foundation to meet the new eara of intelligent surveying and mapping.

The lack of an effective accuracy assessment method of geoid is a key problem in the modernization of height datum and its application. In this paper, according to the error characteristics of GNSS-leveling height anomaly and gravity field in frequency domain, the technical requirements for the fusion of GNSS-leveling and gravity height anomaly are studied. Furthermore, a method of error expression and accuracy evaluation of geoid is proposed. Through the example test and analysis, the main conclusions are as follows: ①The accuracy of the fused height anomaly should be expressed by the error curve of height anomaly difference varying with the distance. ②For accuracy assessment of quasigeoid, two error indicators and two error curves are recommended, namely, the error of gravity height anomaly difference, the internal error of the fused height anomaly, the error curve of the fused height anomaly difference and the error curve of GNSS-levelling height anomaly difference. ③When the distance between two terrestrial points is close to or less than the average distance between GNSS leveling points, GNSS-levelling height anomaly plays a major role in the contribution of the fused height anomaly. ④The accuracy of the fused height anomaly in large scale is mainly controlled by gravity height anomaly.

Using dual-frequency satellite-borne GPS observations from May 24 to May 30, 2015, the linear combination of Melbourne-Wübbena and ionosphere-free linear combination, based on the basis of precise point positioning technology, the batch least squares estimation method is used to precisely determine the Swarm undifferenced kinematic orbits with different orbital attitudes. The orbit accuracy is assessed using three methods, which include satellite-borne GPS phase observation residuals analysis, comparison with ESA reduced-dynamic orbits and external evaluation with SLR measurements. The results indicate: ①Swarm satellite-borne GPS phase observation residual RMS is in the range of 6~7 mm level. ②Comparisons with reduced-dynamic orbits computed by European Space Agency (ESA), radial, along-track and cross-track orbit difference RMS are in the range of 2~4 cm level. ③Comparisons with kinematic orbits computed by ESA, Radial, along-track and cross-track orbit difference RMS are in the range of 1~2 cm level. ④The external validation with satellite laser ranging (SLR) measurement shows RMS errors are in the range of 3~4 cm level. Therefore, it is feasible to use the undifferenced kinematic orbit determination method and the orbit determination strategy provided in this paper to carry out the precise orbit determination of swarm series satellites. The orbit determination accuracy is centimeter level.

Seafloor classification is of great significance for the development and utilization of marine resources and marine scientific research. At present, multibeam detection is one of the effective methods to achieve large-scale seafloor classification. Seafloor classification is usually based on the angular response (AR) features and backscatter image features extracted by using multibeam backscatter. Because the feature source is relatively single and classifier structure is simple, the classification accuracy is often not high. This paper proposes a seafloor classification method based on convolutional neural networks (CNN). In addition to backscatter features, bathymetry features are also used to classify. The feature vectors are converted into waveform maps, and then input to the convolutional neural network for training and classification. The experiment compares different feature combination models and four conventional classifiers: BP network, support vector machine (SVM), k-nearest neighbor (KNN), and random forest (RF). The overall classification accuracy of the experiment reaches 94.86%, the kappa coefficient up to 0.93, and it takes 1 min 25 s. The accuracy has obvious advantages and the efficiency is relatively high. This method can effectively obtain the seafloor information in two different data types, give full play to the characteristics of convolutional neural network weight sharing, high efficiency, and achieve high-resolution seafloor classification. This paper provides a reference for the seafloor classification based on multibeam.

Cropland is a scarce land resource in hilly and mountainous areas, which has the characteristics of complex topographic conditions and diverse planting structures, leading to the difficulty of rapid and accurate acquisition of cropland information in mountainous areas. Therefore, it is difficult to extract the cropland information in mountainous areas quickly and automatically based on the traditional remote sensing data and remote sensing monitoring methods. Aiming at this problem, this paper takes Xifeng County of Guizhou Province in the southwest mountainous area as the experimental area. According to the heterogeneity of geospatial space, this paper proposes the idea of cropland morphological information extraction by geographical division control and stratification extraction, and constructs a method for extracting cropland morphological information based on geographical division control and stratification extraction under the constraints of geomorphic unit. Firstly, according to the geomorphology-vegetation characteristics, the experimental area is divided into three geographical zones: flatland, hillside area and forest. Then, on the basis of each type of partition, the cropland is divided into different types according to the visual characteristics presented by the cropland, and different deep learning models are designed for hierarchical extraction of different types of cropland. The experimental results show that this method has a good suppression effect on the background noise of complex terrain in mountainous areas, and the extracted cropland plot information is more consistent with the actual distribution pattern of the actual cropland compared with the traditional method, which effectively reduces the rate of missing extraction and wrong extraction.

The spatial relationship in the paper means the spatial adjacency relation between the objects in word space, and the paper proposes a novel image matching method based on analyzing the spatial adjacency relation between the object contained in the image pair. In the method, the feature extraction network is first trained based on comparison mechanism, and the well-trained model could produce the deep features for the objects, which could effectively match the consistent objects across images. Then the priori images are employed to deduce the spatial adjacency information between different objects, which is further represented by the spatial adjacency graph. Finally, the spatial relation matching between the image pair is conducted by calculating the spatial adjacency score based on the spatial adjacency graph. The experimental results demonstrate that the spatial adjacency scores of the non-matched image pairs generally equal to 0, and those of the matched pairs are generally greater than 0. As several objects are involved, the proposed spatial relation matching method could achieve high robustness, it outperforms other methods in the comparison experiments, and it could effectively complete the image matching task with high efficiency.

The spaceborne atomic clock is the core equipment of Beidou Satellite Navigation System (BDS). Clock error model is an important part of BDS broadcast messages, and its accuracy directly affects the service performance of BDS. In this paper, the accuracy of the BDS clock model is evaluated using the TWSTF clock error measurement as a reference. The results show that the accuracy of the BDS clock model is better than 2 ns (data age less than 12 h). Aiming at the problem of calculating the clock error parameters of BDS, combined with the clock error model solution strategy of the BDS ground segment, this paper proposes a variety of strategies to further improve the accuracy. For the short-term forecast of 1 h duration, this paper proposes a clock error forecast method of weighted linear model and mixed interval linear model. The accuracy of short-term forecast can be increased by more than 20%. For medium and long-term (>6 h), clock error time series is analyzed by spectral method and prediction model is constructed according to the spectral analysis results. Compared with simple quadratic polynomial model, the prediction accuracy of the clock error model is improved by 13% in 6 h and 21% in 12 h on average. For IGSO/MEO satellite, this paper proposes a strategy of using TWSTF equipment delay as parameters to estimate, which improves the accuracy of clock error model by 18% after equipment switching. Finally, using the improvement strategy proposed in this paper, the clock error data from January to June 2017 is reprocessed to obtain a new broadcast clock error model time series. Using the new broadcast clock error model to locate the BDS monitoring receiver, the results show that, the direction accuracy of N, E and U is increased by 14.22%, 29.39% and 14.91% respectively, which significantly improves the service performance of BDS broadcast clock error model.

Extracting levee features (e.g. central lines, levee crowns, levee slopes and sections) has great significance for detecting demaged levees and security assessment. Airborne LiDAR can be utilized to produce continuous surface data DTM.Thus, it has its advantage of lighter workload and higher precision when extracting levee features. In this paper, it is proposed a method to extract levee features with airborne LiDAR-derived DTM. Firstly, central line of levees is generated by ring detection. Compared to conventional methods, ring detection is tolerant to deformated or broken central lines caued by levee erosion or collapsion. Secondly, on this basis, levee crown and slope data is producted according to the standard of levee slope classification. Lastly, equidistant section lines of levees are extracted from ring intersections, and automatically generated the levee section. An empirical research of our method was conducted on 120 km of levees in Gongshuangcha detention basin in the Dongting Lake area, China.This method performed well in accurate extraction of levee engineering information, and has great application potential.

The result of point cloud semantic segmentation includes the recognition of multiple objects in the scene, which is an important part of 3D scene information extraction. It also plays a key role in many fields such as smart cities. Since the large amount of data and high scene complexity, however, most existing methods can only extract a limited type of objects with a relatively low recognition rate from laser scanning data. This paper proposes a hierarchical multiple object automatic extraction framework combining residual learning and Markov random field (MRF) optimization in a 3D LiDAR point cloud scene. The framework first filters raw data into ground points and non-ground points; buildings are extracted from non-ground points to reduce scene complexity; then the residual learning network is adopted to pointwise classify the remaining point clouds; finally, a MRF-based optimization is used for post-processing and improving the accuracy of point cloud classification. In order to evaluate the effectiveness and robustness of the proposed method, experiments were performed on three outdoor large-scale point cloud scenarios. Experimental results show that the proposed method can perform effective semantic segmentation on various types of point cloud scenes, with (94.6%, 96.8%, 95.7%), (88.5%, 90.5%, 89.2%) and (95.3%, 95.2%, 95.3%), respectively. In addition, compared with existing advanced methods, it is shown that our method significantly improves the performance of semantic segmentation.

Instruments, surrounding environment and human operation often cause a lot of noise in the LiDAR, resulting in low model regression accuracy. RANSAC algorithm is widely used to solve model regression problems by virtue of its simple implementation and robustness. However, for different scenarios, RANSAC algorithm needs to constantly adjust the parameters to estimate the optimal model solution. Considering the RANSAC algorithm and its family existing shortcomings, according to the difference of density distribution between inliers and noise. This paper firstly optimizes the initial hypothesis model by using density weighted guided sampling, and then proposes a spatial density function to evaluate the optimal model and to calculate the number of iterations by using the spatial density function. The whole process does not need any prior knowledge. The method proposed in this paper can solve the model regression problem where the inliers ratio is more than 10%. In addition, compared with the existing methods, the method proposed in this paper can achieve high accuracy and robustness without prior information.

It has been a research hotspot to apply deep learning to remove cloud on satellite images. In this paper, we propose a cloud removal method based on DCLS-GAN for the plateau image of TH-1 satellite. The generator is constructed with the structure of encoder-decoder, and two types of fixed and removable cloud masks are used in training. The least squarereconstruction loss and cross-entropy adversarial loss are used to generate the prediction image of cloud coverage area, whilel east square loss is also used in the discriminator to identify the authenticity of the generated image. Joint optimization of generator and discriminator is achieved by continuous iteration, after which, bilinear interpolation is used to improve the restoration accuracy of cloud coverage area, and Poisson editing is used to smooth the prediction boundary and reduce the influence of artifacts. The experimental results on the testing dataset show that the cloud removal effect of proposed method exceeds classical methods and the original Context Encoder in peak signal-to-noise ratio and structure similarity, and experiments on images with real cloud area also show that proposed method has lower indicators under blind image quality assessment. Finally,the speed is faster than classical methods and equals Context Encoder, thus it has a better practical application prospect.

The Side-scan sonar image automatic recognition is an important part of verification for underwater obstacle and wreck search and rescue, in view of the traditional artificial interpretation of side-scan sonar image is inefficient, time consuming and resource consumption and strong subjective uncertainty and excessive reliance on experience. This paper attempts to introduce the method of convolutional neural network, considering that the side-scan sonar shipwreck image belongs to a small sample data set, and an automatic recognition method of side-scan sonar shipwreck image based on transfer learning is proposed.The sample data were expanded by means of normalization and image enhancement, the training set and testing set were divided into 4∶1 proportions, and an improved model was designed according to the characteristics of the side-scan sonar wreck data set by referring to the classical VGG-16 model, then, the improved model trained on the ImageNet image data set is used to learn and experiment on the small sample side-scan sonar shipwreck data set using two transfer learning methods: freeze and train and fine-tuning, and compared with new learning. The results show that the accuracy of the three methods for the recognition of side-scan sonar shipwreck images is 93.71%, 84.49% and 90.58%, respectively. The first transfer learning method has the highest accuracy rate, the fastest model convergence speed, and the highest AP value 92.45%, which is 8.06% and 3.06% higher than the second transfer learning and the new learning method, respectively,and has a better effect in improving the model’s recognition ability and training efficiency. which verifies the effectiveness and feasibility of this method and has certain practical guiding significance.

Denoising fitting of airborne laser bathymetry data is a key step in extracting the bottom terrain. The algorithm effects of wavelet adaptive threshold denoising, empirical model denoising (EMD) and joint denoising are compared in this paper, and then multivariate Gaussian fitting is used to test the denoising effect. The optimal denoising algorithm and parameter selection are obtained by comparison, and it is realized that the high-precision extraction of seabed features. This study has shown that: when the sounding data is denoised by wavelet threshold, the fixed threshold wavelet denoising effect is superior to other denoising effects, and the denoising decomposition level is more than 6 layers, which tends to be stable. The average accuracy of the algorithm reaches 8.218 2 after the fifth-order Gaussian fitting of the denoising data. The algorithm has strong robustness, it can meet the technical requirements of blue-green laser practical application, and provides a reference for accurately extracting seafloor feature information.

To overcome the phenomenon of the traditional tomography technique that cannot use the data of GNSS stations outside the research area, a novel three-dimensional water vapor tomography modeling considering the data of GNSS stations outside the research area is proposed. The initial water vapor field of tomographic area in each voxel was calculated based on the GPT2w model. The proportional coefficient is introduced and combined with the initial value of water vapor density to determine the proportional coefficient expression. The water vapor content of the signals in the tomography area from the stations outside this area is estimated. Finally, the water vapor observation equation considering the data from the GNSS stations outside the tomographic region is established. The proposed method can be used to utilize the GNSS observation efficiently and improve the accuracy of tomographic result, but it only validated to the GNSS stations located in a certain range outside the tomographic area. Twenty-four GNSS stations and one radiosonde station in Zhejiang CORS network were used to verify the proposed tomography model in this paper. The experimental results show that compared with the traditional chromatography method, the utilization rate of GNSS signals and the voxel number with signals passing through are increased by 26.8% and 14.9%, respectively. Taking the radiosonde data as reference, it is found that the accuracy of integrated water vapor and water vapor density profiles calculated by the proposed method are better than those calculated by the traditional method.