2011 Vol. 36, No. 2
Aiming at the large-scale 3D city modeling and collaborative database applications,a high-performance 3D GIS database engine is designed and developed,which supports the multimode database management based on Oracle 11g DBMS.This paper presents a new 3D spatial database model concerned with semantics for the integrative management of outdoor & indoor and aboveground & underground 3D spatial data,then introduces related key technologies such as hierarchical 3D spatial index,multilevel caches,multithread dispatch and asynchronous communication and transmission.By means of the real 3D city models of Wuhan city,the experimental results prove the validity of this engine.
Based on integration of multi-agent system and GIS,the theory and methodology of simulating pedestrians flows in crowds' activity center were studied.The general framework and mathematical expressions of the model were put forward.As well,some expressions of moving decision of pedestrian agents and their implemented techniques were explored and designed.A modeling prototype was developed.As an example,a full scattering process of spectator agents in Shanghai stadium was simulated successfully.A successful simulating of a case of emergence when one or more exits emerge accidents will be very useful for managing and treating crowds' safety in a lot of assembling centers.
Aiming at the difficult problems of massive geological data load,the theoretical basis is spatial clustering and spatial interpolation.The spatial objects in cache are regarded as sample data,the hit-ratio of these objects are thought as interpolation weight,and the object information in spatial index are look upon as estimated data,and the RAM and the computing capabilities of CPU are also took into account.The self-adaptive pre-load method on massive 3D spatial data in geological space are designed,and the experimental result shows that our proposed method is correct and efficiency.
To alleviate the deficiency of excessive edge crossing brought by random layout of categories in categorical data visualization—parallel sets,we propose a heuristic layout algorithm based on median,which optimise the layout order of categories such that visual clutter is eased.And then we utilize the improved parallel sets to analyze the global terrorism database(GTD).The experimental results show that the improved parallel sets can clearly express the association among the multi-categories in GTD,thereby assist users in analyzing the implicit information of various terrorist organizations,such as the behavior characteristics.Furthermore,the median-based heuristic is simple and has high efficiency,which is suitable for large data sets with many categorical attributes.
Manifold learning,as the novel nonlinear dimensionality reduction algorithm,is applied to dimensionality reduction and feature extraction of hyperspectral remote sensing information.In order to address inherent nonlinear characteristics of hyperspectral image,Isometric mapping(Isomap),the most popular manifold learning algorithm,is employed to dimensionality reduction of hyperspectral image,and the experimental results show that it outperforms traditional MNF transform.In order to include spectral information into manifold learning,spectral angle(SA) and spectral information divergence(SID),instead of Euclidean distance,are applied to derive the neighborhood distances in Isomap algorithm,and the result is better than that using Euclidean distance in terms of residual variance and normalized spectral eigenvalue.It is concluded that manifold learning is effective to dimensionality reduction and feature extraction from hyperspectral remote sensing imagery.
By analyzing spectral characteristics and phenological characteristics of remote sensing images,we select multi-temporal NDVI data along with other useful bands for data processing.By performing an enhanced Lee filter and MNF transform on the data,the features of different land use types has been enhanced.Then,a morphological filtering is conducted on the data to extract dry land.Finally,we separate wetland and paddy field by a SOM neural network clustering.As a result,the accuracy of land use type classification has been improved.
In order to reduce the effect of fuzzy edges and tips of vein feature image in the template matching,we present a tri-value template fuzzy match algorithm,which segments the vein feature image into three areas: subject area,fuzzy area and background area;compute the average distance between the non-background areas as the similarity score between the matching templates.It is robust against the fuzzy edges and tips by increase the distance norm to the fuzzy area.The recognition accuracy rate,99.46%,to 456 near-infrared finger vein images,shows that our proposed method is feasible and practical.
The imaging mechanism,acquired angle and resolution between imagery acquired by diverse remote sensing sensors are different,which throws a big difficulty in the registration between them.The SIFT algorithm has had a successful application in aerial and video images presently;however,when applying it in registration between imagery acquired by different sensors or with big differences in resolution and angle,few matching couples are obtained.Aiming to this problem,a new algorithm named SIFT based on visual matching window is proposed.By constructing the visual matching window,the scale similarity and matching probability between SIFT features are increased,and then combined with least-square equation and bilinear interpolation method,the automatic registration is achieved.SPOT-5(P) and TM images,which have big differences in acquired angle and resolution,were tested,and the results show that the precision is lower than 1 pixel.
A method is proposed for filtering.KD-tree is employed to eliminate the blunders,only last returns and singular echoes are retained for filtering after making multi-returns analysis,and morphological reconstruction is used to achieve the DEM from the DSM.Compared to the existing methods,it only needs one human input,and the input has a very simple indication.The experimental results show that our method can retain the ground points and eliminate the non-ground point as much as possible.That is the type I error,type II error,total error are kept within 9.93%,7.27% and 9.76%.
According to SVM theory and the separability measure of hyperspectral data,we put forward a novel binary tree multi-class SVM classifier based on separability between different classes,constructed different multi-class SVM classifiers and tested their accuracy by experimented the hyperspectral image with the 64 bands OMISII data and Hyperion hyperspectral data.The experimental results show that the novel binary tree classifier has the highest accuracy than the other multi-class SVM classifiers and some traditional classifiers(spectral angle mapping classification and minimum distance classification).Use of the novel binary tree multi-class SVM classifier based on separability measure is a novel approach which improves the accuracy of hyperspectral image classification and expands the possibilities for scientific interpretation and application.
A simple and efficient linear 3D deformable model is proposed to reconstruct individuation 3D human face.First,a standardized 3D face database based on resampling of key feature region is built to realize the dense correspondence among faces.Then,the faces of which singular value vector is most similar to the target face are selected as the basic space to choose components of morphable model dynamically.Finally,the original morphable model is replaced to a feature-point based sparse morphable model to reduce the difficulty of solving linear combination coefficients.The experimental results show that this method has the short modeling time and low complexity,and can generate a realistic 3D faces from a few landmarks of real-world photos.
We propose a method for refining the RPCs.That is,refining the RPCs through eliminating systematic errors in the virtual grid.The empirical results show that the planimetric accuracy in image-space for QuickBird imagery reaches ±2.4 pixels after refining the RPCs by the proposed method.For SPOT-5 stereo imagery,the planimetric and height accuracy of the object stereo positioning are ±5.892 m and ±4.020 m,respectively.
An approach based on cross ratio to distance measurement for a single uncalibrated image is presented.Without any prior knowledge of the camera internal parameters(such as focal length and aspect ratio) or its pose(position and orientation),we put foreword and employ nested cross ratio to compute the length of a line segment on plane surface,while only the length(s) of the certain segment(s) is/are known.Based on previous steps one conclusion can be acquired: Once the lengths of rectangle two sides on one plane surface are known and the corresponding vanishing line of the plane is identified,then the length of any line segment on the plane surface can be measured.The extension of our conclusion is also discussed.
The perpendicular drought index(PDI) and modified perpendicular drought index(MPDI) were applied and compared over the vegetation area of Ming'an town,Inner Mongolia,based on TM imagery.The results show that the determination coefficient of PDI,MPDI with the measured soil moisture are 0.37,0.535 5 respectively.PDI and MPDI are both feasible for drought monitoring in our study area,and the monitoring precision of MPDI is higher than PDI.In addition,the comparison of the drought mapping of the two indexes,and the analysis of the variation trend of the two indexes with the soil moisture both indicated that monitoring results of the two indexes are similar in the whole test area,but in vegetation area,the precision of drought monitoring of MPDI is superior to PDI,this is mainly due to the vegetation impact considered in MPDI.
Matching line of MVLL algorithm was investigated for line-array CCD images,and it can be concluded that the matching line was the line between top feature and its corresponding bottom feature and can be regarded as a straight line.For MVLL algorithms,the difficulties of discontinuity feature that usually matched to the bottom were analyzed,and the algorithm of self-adaptive adjustment of correlation parameters was improved.After the process of self-adaptive window extension algorithm was given.The matching experiments were done on discontinuity features using two ADS40 line-array image datasets.The results show that,after applying the self-adaptive window extension algorithm,discontinuity features could be successfully matched and the self-adaptive function of MVLL algorithm was thus enhanced.
The water vapor transmittances in the 940 nm band are computed using MODTRAN for 34 sites in China.Parameters a and b for all the atmospheric models in the empirical water vapor transmittance function are fitted.It is found that the values of a and b is the function of atmospheric water vapor amount.The empirical functions are established for the relationships between the values of a and b and atmospheric model's perciptible water.Considering the parameter of the surface altitude H(or surface pressure p0),the empirical coefficients related to H are parameterized.Finally,the uncertainty in retrieved PWV caused by parameter a and b is analyzed.
The simulation method was employed for studying the impact of the antenna pitch angle and semi major axis on mirror reflection points.The spatial and temporal resolutions of GNSS-R ocean remote sensing satellite constellation which was composed of four and six satellites have been figured out.Constellation formed by 6 satellites has good coverage performance for global 5°×5° ocean grid on the eight hours time.
We describe the technology of the orbit determination for the Lunar satellite,analyze and solve the key problems.The system of the precision orbit determination is verified and testified depending on the observation data of the SMART-1 Lunar satellite.The orbit determination can be obtained with an accuracy of 100 meters for three-day-arc data.During the CE-1 mission,the result of the orbit determination is accurate.The satellite inject the mission orbit successfully.There is only one middle course correction during the transfer orbit phase although three times of correction are prepared.The fuel is saved for the satellite.Compared with the ephemeris of CE-1 satellite from GEODYN II,the accuracy of the orbit determination is about 100 meters level for whole course and tens of meters for orbiting Lunar phase.It is shown that the technology described in the paper is correct,and the key problem is solved effectively.The system can satisfy the requirement of CE-1 mission.
With phase smoothing pseudo-range observed value,a regional ionospheric delay model is established by sequential least squares surface fitting with restrictive conditions.As a result,the average inner precision of the model is 6.9 cm,while the average outer precision is 8.8 cm.We also compare with the electron content calculated by global grid ionospheric delay model.Single frequency-PPP test of the two models have also done,the results show that the precision of regional ionospheric delay model is obviously superior to that of global grid ionospheric delay model.
In order to filter the series,the traditional algorithms require an explicit system model which is a difficult problem in nonlinear-system.We present an effective method based on the online support vector machine and Unscented Kalman filter method to filter a precipitation series.The experimental results show that our proposed method is more efficient to get accurate result than the traditional Kalman filter method and the support vector machine method.
The iterated closest contour point(ICCP) matching algorithm based on Fibonacci number series was proposed.The improved ICCP matching algorithm was also presented because of its confinement to the real track near the INS track.The INS track was corrected based on correlative extremum at first,then ICCP matching was made in order to further improve the accuracy.By analyzing the data within experimental area under different simulative conditions,the results show that all navigation errors were corrected effectively.
To determine an orthometric height using GPS,it is necessary to know the geoid/ quasigeoid undulation.Many methods have been studied for the interpolation of discrete GPS/leveling data to get the geoid/quasigeoid undulation unknown.A new half free-positioned point masses model is proposed based on equivalent sources theory.Corresponding iterative algorithm to construct the point masses is also presented considering the relationship of the nearest GPS/leveling points.The results in Hong Kong and Shenzhen show that the point masses model is efficient for the interpolation of geoid/quasigeoid undulation.
Joint working mode,which applying both Leica Axyz theodolite industrial measuring system and georobot TCA2003,is applied in pose measurment of mine shaft elevator equipment.In order to achieve a unified coordinate system measurement results,contact measurement and conversion between mine control survey coordinate system and measuring station coordinate system which established by theodolite industrial measuring system must be done firstly.And the mine shaft elevator equipment's space location and pose will be detected with high precision.At last,space pose analysis of each equipment can be done by virtue of Axyz software.Engineering practice shows that such technique can effectively solve many difficult problems such as feature points invisible,control points missing,etc.
Map symbols have the same origination with geographic data.The advantages and disadvantages of data structures on marker symbols,such as TrueType,Primitive,and CAD block,are analyzed for aim of building and sharing.A data structure on path is proposed to assimilate symbols from TrueType,Primitive and CAD block.The validation tests show that it has strong ability of description and high scalability.
To meet the requirements of complex calculations with large amounts of data quickly,using the advantages of flexible operation of data sets by SQL language.Presents a calculation based on SQL data computing.Combined with the secondary development of the SQL and statistical computing modeling features of dam monitoring system,change the software computing by SQl computing directly,ultimately achieve the integration of data calculation.The results show that the application of new methods improve the system operational performance and computing speed,simplify the preparation of the algorithmand achieve the optimal development of the system.
A method using time-frequency analysis and dummy frequency agility to suppress ionospheric clutter based on ionospheric clutter characteristics in high frequency ground wave radar(HFGWR) is proposed.Making use of eigen decomposition cooperating in the element-pulse domain and in the element-doppler domain to suppress ionospheric clutter by analyzing the time-frequency images where the swept pulse section and the Doppler section were exactly determined in the method.Then,the ionospheric clutter suppressed by the method of adaptive sidelobe cancellation using the structural features of the array.The experimental results show that the method can effectively suppress ionospheric clutter using the data recorded of the radar.