2010 Vol. 35, No. 2
The state-of-the-art and achievements of digital Earth are analyzed.Smart sensor Web and ubiquitous sensor network(USN),a new infrastructure which appears with the development of IT,communication technology and network technology,is discussed.Platform framework and typical applications of USN based on all-IP technology are designed and prospect on the trend and beautiful future from digital Earth to smart Earth is given.
We present a data model for GeoVideo and a system framework for VideoGIS,including the data structure of GeoVideo data,the processing flow of Geovideo data,key modules of a VideoGIS,and the technical design and implementation issues.Several key technologies,such as Web map design,GeoVideo management,GeoVideo player design and system integration,are discussed in detail.Utilizing Google Maps API,ArcGIS 9.3 map services and Flash video standard,a prototype of VideoGIS is implemented.This VideoGIS provides flexible interfaces for GeoVideo query,index,visualization,navigation in the context of Web maps,and has verified the feasibility of the GeoVideo data model and VideoGIS design.
By adopting objected-oriented method,we begin with a profound analysis of the 3D vector models and possible operations from the micro perspective,and propose an integrated framework for 3D vector modeling together with its data organization.A series of 3D vector model components with powerful functions can rapidly be constructed in the form of parameter configuration with the help of this framework.By the way of model components + semantic attributes,it supports the geo-science modeling in practice as well as the analogy and analysis of geo-science objects in VGE or 3D GIS systems.A case study was given to verify the validity and feasibility of the integrated framework together with its model components.
An efficient hardware-accelerated view-dependent level-of-detail rendering technique for out-of-core visualization of large terrains is developed.To achieve realistic in visualization,the corresponding satellite or aerial imagery texture is applied over the terrains.Also,to display large collection of static objects such as buildings,trees,roads,and so on.Over the terrain while maintaining the real-time frame rates,efficient object handling methods are developed.A technique to use powerful performance and programmable vertex and fragment features of current GPU for advanced visual effects and increased realism in terrain visualization are proposed.The above algorithms have been successfully tested on different terrains and satellite images.Results show that large-scale terrain renders fast and with high realistic using GPU-based programmable hardware.
We present a novel spatio-temporal index for constraint networks,which is called NBR-tree(network-based R-tree).We focus on the background of the applications in urban traffic networks.The NBR-tree is an improvement on the previous index named MON-tree,with an analysis on the specific properties of the moving objects in urban traffic networks.We discuss the index structure,operating algorithms as well as the experiments of the NBR-tree in detail.The experimental results show that our proposed index is able to support NN queries,and is more efficient than the MON-tree in evaluating trajectory queries.
Both static and kinematic tests were investigated by using IGS 5 min,30 s and 5 s-interval precise satellite clock product in precise point positioning(PPP) solution.Test results show that the sample rate of IGS satellite clock has little effect on the static PPP solution.All the three types of sampling interval of precise satellite clock can satisfy mm-cm level positioning accuracy.A higher sampling rate has no significant improvement for PPP solution.While in kinematic PPP,sample rate of satellite clock has significant impact on the PPP solution,The higher the interval of satellite clock is,the better accuracy achieves.The accuracy of kinematic PPP achieved by using 30 s-interval precise satellite clock is improved by nearly 30%-50% with respect to the solution by using 5 min-interval precise satellite clock.While using 5 s and 30 s-interval satellite clock almost produce the same accuracy of kineamtic solution.
A new factor based on partial state discrepancy was developed.Compared with the standard Kalman filtering,both the adaptive factors constructed by the predicted residuals and partial state discrepancy can resist the influence of the dynamic model errors when no outliers exist in measurements.The precisions of their navigation are almost idential.But if outliers exist in measurements,the adaptive factor based on the predicted residuals can not identify the state model errors and the observation errors while the adaptive factor based on partial state discrepancy can resist the influence of the outliers.Hence,the latter navigation precision is prior to the former navigation precision.
An integrated MINS/GPS navigation system is presented as an airborne equipment for UAV.The heading's observability is very weak.In order to eliminate this divergent trend,two methods that Kalman filter with heading coupling and dead reckoning revising the heading of SINS are proposed.The simulation results show that these methods can effectively eliminate the divergent trend of heading,and have high positioning and orienting precision.The system can be used on UAVs and airships.It will have more and more bright future.
In order to process the rover and the reference stations GPS data with different sampling rates,a reasonable strategy is presented.When the observations of the rover are nonsynchronous with those of the reference station,the position of the rover will be estimated with the calculated precise "pseudo-ranges".Otherwise,the double difference(DD) model will be employed.The proposed method not only makes full use of the available data,but also saves the cost of data collection and transmission.An experiment was carried out,in which a rover receiver was mounted on an airplane while a reference station receiver was located on the ground.The rover positioning precision obtained by our presented method with nonsynchronous data was within several centimeters compared with those obtained by the DD model with synchronized observations.
On the ground of hypothesis that the change of marine terrain is continue and slow,a new method of detecting outliers of multi-beam data by Bayes estimate theory is presented.In order to verify the validities and rationalities of this method,a contrast between this method and the approach of select-weight iterative filter is done.The result shows that the approach can solve the problem of judge-standard reliability and could detect the outliers of Multi-Beam data valid and rational.
Space variation of simulation and practical data are analyzed.The conclusions show that the influence of fish's height change to surveying results is relate with complexity of horizontal gradient of magnetic anomaly.Based on foregoing analysis,threshold is proposed,and rationality and practicability are proved by test data.We suggest that magnetic survey data must be corrected to mean sea level when carry level 1,level 2 and level 3 marine magnetic survey.
We develop a new strategy for evaluating the GPS stochastic model in long baseline scenario.Firstly,we employ high frequency pass filtering of multiple differences in time domain to efficiently eliminate the systematic biases.Then,the mathematic models are proposed to assess the different aspects of GPS stochastic characteristics.Finally,the real GPS data is used to demonstrate the performance of the developed models.The findings can benefit the quality evaluation and integrity of CORS measurements and the quality control of positioning results in rover end.
Inversion of parameter based on total least squares(TLS) is investigated.Strain parameters TLS inversion method is deduced,and the errors of distance changes and azimuthal angles are considered.The method of precision evaluation is given.Through the examples,the strain parameters based on total least squares are more reasonable.If there are constant columns in the coefficient matrix,they must be removalled.Otherwise,the obtained parameters are wrong using the TLS.Because the errors are distributed to the constant columns.
The method combining homotopy functions with filled functions to solve nonlinear least squares adjustment is presented.We first solve the well-posed nonlinear equations with homotopy method to obtain local optimal solutions,and reasonable filled functions are generated according to the local optimal solutions.It obtains better local optimal solutions than current solutions by the filled functions.Then,the homotopy functions and filled functions will be restructured.Finally,we can find the optimal solutions by limited loop-iteration method.The results show that the method can find optimal solutions effectively.
The p-norm distribution is an extensive distribution family of measuring errors.For each concrete measuring data,we can selected a suitable value for p to make the theoretical model of the error distribution,which was used in data processing be more close to the real one of the error than the normal distribution.We discussed the relation between p value and the efficiency of Lp estimation,then bring up a fast parameter estimation method of p-norm distribution.Finally,we verify the method by simulation.
A stochastic model of DEM linear interpolation is derived based on the stochastic process model,as well as an uncertainty description of its basic element——the irregular random spatial triangle.Error propagation of the TIN points in the linear interpolation is discussed.Through theoretical deduction and practical examples,we obtained the point-seat variance of the TIN DEM linear interpolation points,analytical expressions for three semi-axises of the error ellipsoid and the coordinate of the point in the DEM linear interpolation with the highest precision.The results has nothing to do with the shape of the triangle.Moreover,theoretical proof makes on the inevitability of the dramatically decrease of the precision caused by the DEM linear extrapolation.An analytical expression of the average point-seat variance of the TIN DEM linear interpolation is obtained.This model could uniformly describe the uncertainty of the points on the triangle and its sides.Through theoretical deduction and numerical test,we obtained the coordinate of the point in the DEM linear interpolation with the highest precision.
In order to overcome the limitations of RMSEs,we developed a scheme for DEM error surface construction based on conditional simulation(CS).A DEM error surface of Dongzhi tableland locate in Gansu province were comparatively constructed based on CS and ordinary Kriging(OK).Results indicate that OK has an obvious smooth effect,whereas CS can accurately represent the spatial fluctuation of DEM errors.DEM error has a more serious effect on slope accuracy in flat area than that in complex area;Results of water and soil loss level determination show that about 70.2% of grids are influenced by DEM errors.
The theory of quaternions is introduced into the field of photogrammetry.A new method which uses quaternions to describe the position and attitude of line scanner CCD image is presented.Firstly the quaternions are used to describe the rotation matrix in the algorithm.Then,the strict collinear equation is linearized,and at the same time the iteration by correcting characteristic value can effectively overcome the strong interrelationship among exterior elements.The numerical experiments were done.Results show the correctness and reliability of the method
The urban street is a primary component of the city as one's intuitionistic impression on the city is coming from it.Hence,it is very significant for 3D city modeling(3DCM) to rapidly realize 3D visualizations of the street sight.Accordingly,the research on rapidly realizing 3D visualizations for the street sight combined the vehicle-based image sequence with 2D vector map is presented.The algorithms proposed were verified by experiments on real data sets.
We present an efficient and low-cost method for automatically detecting and tracking of the space debris objects from astronomical images,by using a combination of active contours and shape feature similarities.An object detection algorithm is firstly implemented following some image preprocessing steps,in order to locate all the major objects in each image.Next,an object tracking method based on Greedy Snake algorithm is proposed so that all the detected objects are able to be correctly tracked.Finally,the space debris in the image sequence is identified by applying a shape feature similarity matching operation.Experimental results are also illustrated to demonstrate the capabilities of the proposed method to perform correct space debris tracking.
The utility of target shadows for automatic target recognition(ATR) in synthetic aperture radar(SAR) imagery is investigated.A method for synthetic aperture radar images target recognition using hidden Markov models and chain code is presented.Shadow shape of SAR image is described by chain code which flects shape characteristic well and computes effectively.The chain codes of target shadow are utilized for hidden Markov modeling.Targets are classified using HMM statistical model.Image samples of targets in MSTAR database are used to verify the method.The results show that the proposed method can enhance the target recognition rate evidently and is an effective method for synthetic aperture radar images target recognition.
The objective of the present research was to identify and select the spectral wavelength with the partial least square(PLS) method for the study of rice.We use PLS method to select the wavelengths which have the largest weightings for rice leaf biochemical concentrations respectively for leaf nitrogen concentration and chloroghyll-a concentration.In order to evaluate the wavelength selection results,we construct eight kinds of wavelength(bands) combinations for rice leaf nitrogen concentration inversion analysis by PLS.The most appropriate wavelengths are 552 nm,675 nm,752 nm,776 nm,respectively.
In order to facilitate follow-up filtering of the airborne LIDAR points cloud,we presented two preprocessing methods.One is the mean-tolerance method for preprocessing the forested areas with dense vegetation and steep slope;the other one is the angle-tolerance method for preprocessing the urban areas with dense laser footprints reflected from walls of buildings.Two datasets with different characteristics were used in experiments to test the two presented preprocessing methods.The experimental results were analyzed qualitatively and quantitatively.The results indicate that two preprocessing methods are effective,two types of filtering errors are reduced when filtering with preprocessing.
This paper proposes an improvement of Naive Bayesian classifier-selective Naive Baysian classifier together with two enhancement of Naive Bayesian classifier-tree Augmented Nave Bayes and Bayes Augmented Nave Bayes.We constructed these classifiers for remote sensing images based on the mutual information between bands,and compared their performance with the NBC.
This paper proposes a multi-sensor image registration algorithm based on line features and sift points.Firstly,line features from reference image and image to be rectified are matched,and virtual corners are constructed on the basis of matched lines.Next,for the lack of traditional sift algorithm in multi-sensor image registration,sift control points are extracted by line constraining primitive matched sift points.Finally,virtual corners and SIFT control points are combined to construct TIN for tiny facet primitive rectifying.Experimental results show that the proposed algorithm has high registration accuracy.
A novel color recovering method for black-white facial image based on image warping and color-space transformation is presented.The basic idea of the method is referring a colorized facial image and then adding its tint into the black-white facial image.This method is simple to understand and can easily be implemented.Experimental results show that the method is very effective and can naturally and livelily realize color recovering of different black-white facial images including images of the males,the females,the elders,and the youngers.The facial details in the color recovered image obtained by this method is very realistic and convincing.
There are many factors in land grading.It is very significative to determine the weights of factors.The traditional methods include the subjective and objective judgment.This paper proposes a genetic algorithm at the basic of concluding the research achievement of our predecessors.It over comes limiting factor of other methods,finds relationships and sequence of weighted by optimizing to fit the data sets based on setting up different targets.Genetic algorithm can dig up optimum weight sets of one field and find trends of one factor with the change of land-value in order to gain the classification of grading factors.This paper presents implementation steps of the method,verifies the feasible and scientific significance by examples,and achieves good results.
The selection criterions which are normally used for map projection are veviewed.This paper tries to employ shape ratio' as one of the basic factors in the assessment process.Analytic Hierarchy Process(AHP) is also used to maintain the weights of the related factors.Experiment has been done to verify the feasibilities and rationalities.
The particle swarm optimization(PSO) has the character of the high implementing speed.It is little affected by dimension.The support vector machines(SVM) is based on the theory of structural risk minimization.Combining the binary particle swarm optimization(BPSO) and the SVM,the feature subset selection for driving forces of cultivated land is constructed.This method uses the features defined in the feature subset to train the SVM which is evaluated by fitness fuction.The result of fitness fuction is utilized to direct the BPSO to search.The case study shows that this method has the excellent efficiency to select the feature subset.