2015 Vol. 40, No. 5
China Geodetic Coordinate System 2000(CGCS2000)is national new geocentric coordinatesystem of China,it has been adopted nearly 7years since July 1,2008,and it is being applied fully indifferent fields.The progress made in the construction CGCS2000framework was reviewed in this pa-per,the refinement of CGCS2000framework was introduced,and especially the plate motion modeland grid velocity field model which modeled in the CGCS2000frame maintaining were focused on de-scription,the update recommendation about CGCS2000frame was given at the end of this paper.
We propose a new image fuzzy classification method which based on the image relational de-gree.The correlation degree basis points are obtained in each category,and the mean and standarddeviationσof relational degree are calculated for each category.Then the correlation degree for eachimage and the mean difference of correlation degree for each categoryΔi are calculated.IfΔi≤2σ,theimage should“stay”in the class,otherwise,the image should be associated with a greater correlationdegree of the other two categories.Each image should have such inspection to achieve image re-divi-sion until the process is stable.Comparison results show that the quality of the quality of the imageclassification method has certain advantages.
A super-resolution method based on sparse representation and classified texture patches wasproposed,mainly using the priori knowledge and texture to reconstruct remote sensing images.First,extract image blocks for dictionary learning,the over-complete dictionary was learned from the highand low resolution remote sensing image blocks.Orthogonal match pursuit was used to calculate thesparse conefficients,then the coefficients were fixed,iterative method was used to update the diction-ary until the algorithm converges.Then,the training dictionary was used to reconstruct the remotesensing images.In this step,the image was divided into smooth patches and non-smooth patches,bicubic interpolation was used for smooth patches while sparse conefficients and over-complete diction-ary were used for non-smooth patches.Experiment shows that this method has a faster reconstructionspeed and can achieve satisfied super-resolution results in the visual effects and objective evaluation in-dicators.
Traditional dimensionality reduction methods in face recognition are methods that reshapetensor face into a vector,which may lose the structural characteristics of the original data,leading toa relatively low identification result.We present a dimensionality reduction method———multilineardiscriminant subspace projection(MDSP)based on tensor.Our algorithm aims to use tensor to de-scribe face data directly,and project the tensor data onto the vector discriminant subspace through anew kind of projection method———tensor to vector projection(TVP).To reach this target,the algo-rithm first finds out the projection vectors(PV)that make data in the discriminant subspace get themaximum between-class scatter as well as the minimum within-class scatter.Then with the help ofPV,tensor data can be projected into the low dimensional vector data.As long as proper constraintsare given,the vector data can be the most representative feature data.The feature data is then sent tothe KNN classifier for classification.Results in experiments on databases ORL confirm the veracity ofour algorithm.
The superiority of waveform data is hard to evaluated and compared with point clouds quan-titatively due to a high correlation between two data sources from the most commonly used LiDARsystems.Leica ALS60airborne LiDAR system can record both of discrete return data and waveformdata simultaneously and independently.The raw waveforms are decomposed into individual pulse re-turns and compared with the laser points from the hardware system quantitatively.Data from typicalforest and urban areas are picked to perform a digitized assessment of the capability of waveform.Theresults show that waveform data can increase the vertical information in different types of ground ob-jects and spatial resolution of point cloud at a certain level and that the increment in forest area is high-er than that of urban area and bare earth.It is concluded that the penetration capability of laser pulseis more evident in full waveform data than in point cloud from ALS system itself,and that full wave-form LiDAR has much more potential in constructing the vertical parameters in the forest areas thanoriginal LiDAR with only point clouds.
Terrestrial photogrammetry differs from traditional aerial photogrammetry and close-rangephotogrammetry,for the former adopts different resolution photography,while the latter is similar totake same resolution photography.In the classical least squares adjustment,the coefficient matrix isconsidered without error.Unfortunately,variables in the coefficient matrix are inevitably error-con-taminated.Considering the random errors may exist both in the observation vector and the coefficientmatrix,the total least squares algorithm is studied in this paper.According the weight matrix of theobservations based on different resolution Stereoscopic images of terrestrial photogrammetry,theweighted total least squares in the relative orientation and Procrustes algorithm with absolute orienta-tion are formulated,then the procedure is described.At the end of the paper,the experiments of rela-tive orientation and absolute orientation are investigated to prove that using the weighted total leastsquares based on terrestrial photogrammetry is practicable,and the reliable and accurate solution canbe obtained.
The rapid and accurate large scene 3Dreconstruction technique from multi-view images canprovide important and reliable information for emergency response and disaster assessment.Againstthe low efficiency of Structure from Motion(SfM)algorithm,this paper develops an image Topologybased Structure from Motion(TSfM)algorithm referring to image topological conjunction.Genera-ting the image topological conjunction with the flight-control data acquired by unmanned aerial vehiclesystem(UAV),the searching range for matched images is reduced in the process of feature matching,and the time complexity of TSfM algorithm in the feature matching stage decreases from O(n2)toO(n)as compared with SfM algorithm.The experimental results show that TSfM algorithm makes itpossible for rapid large scene 3Dreconstruction with sequence images from UAV.Furthermore,it isreached that the relative error of the 3Dscene model reconstructed by TSfM algorithm is comparablewith that by SfM algorithm.The proposed TSfM algorithm is applied for Lushan earthquake disaster3Dreconstruction with UAV images,which can help to detect the seismic-induced landslides withmore information.
Apolarimetric SAR classification method is proposed in this paper.We can distinguish themorphology characteristic and do initial classification using anisotropy and orientation randomness.After that,Wishart distance is used to iteratively classify.E-SAR L band,SIR-C/X-SAR L band andSIR-C/X-SAR C band polarimetric data of Oberpfaffenhofen,Germany are employed to test the newmethod.The results demonstrate that the result of new method is better than the results of Freeman-Duren and Yamaguchi for vegetation area.
Spectral similarity measure is an important tool of hyperspectral remote sensing image clas-sification.By setting the threshold to judge the pixel spectrum and the reference spectra is similar ordissimilar.To overcome this problem,this paper proposes a multi-feature conversion adaptive classifi-cation of hyperspectral image,this is done through using spectral similarity measure value as similari-ty patterns.Experimental results show that the proposed methods are,compared with the traditionalSVM method in the overall accuracy of classification,increased by 6.25%and 8.72%,also it impliesthat using simple learnable measures outperforms complex and manually turned techniques used in tra-ditional classification.
We propose a new adaptive corner detector based on multi-scale chord-angle sharpness accu-mulation, which can reduce location error and detects fine accuracy on noisy images. Firstly, we use the canny detector to detect edges at low computational cost. Secondly, we devise support regions of the contour into three sections as scales and computes the chord-angel sharpness respectively, then accumulate the three scale sharpness as corner response function. Finally, we use an dynamic adaptive corner threshold to label corners. The results on fine and low quality images show that the proposed algorithm performs better than the other three algorithms in terms of both detection accuracy and location error.
We present a change detection method based on multi-scale geometric feature vector(MSG-FV).The change analysis standard in this novel method is converted from pixel spectral space to seg- ment scale space.Context information beyond multi-scale imagery segmentation is applied to performchange detection.Specifically,this approach specifies number and values of multiple segmentationscales at first.Secondly,two-date imageries are segmented respectively using multiple scales.Third-ly,multi-scale geometric feature vector of a detection unit are constructed in different dates.Finally, 第40卷第5期陆 苗等:利用多尺度几何特征向量的变化检测方法627the change intensity between two-date multi-scale geometric feature vectors is calculated as the changestandard.The study area of Weinan in Shannxi province is tested to analyze the land cover changefrom 2000to 2009.In this sample area,three geometric features are used in this approach.Then,theoptimized geometric feature is compared to other existing methods(CVA,Correlation).The MSGFVapproach is proved to outperform other methods.
Recently,variational level set method is widely used in image segmentation,but its energyfunctional is non-convex,which can easily get stuck in local minima.Firstly,we propose a locally(下转第637页) 第40卷第5期段 平等:顾及异向性的局部径向基函数三维空间插值637 (3D),local radial basis function(RBF)interpolation based on exploring spatial anisotropy with vario-gram analysis was proposed.Firstly,three axes of the data was solved by constructing covariance ma-trix of the sampling point data and then the data was transformed into the new coordinate system byrotating transformation;the range of each direction was calculated using geostatistical variograms;thethree values of range was set as three axes of the ellipsoid;at last,node RBF at each sample point wasbuilt.The attribute values of interpolation were solved by linear combination of node RBF.Experi-mental results show that the proposed method is a feasible method for 3Dspatial interpolation consid-ering anisotropy with high accuracy and reliable interpolation result.
Aiming at the searching range of interpolation points in the process of three-dimensional段 (3D),local radial basis function(RBF)interpolation based on exploring spatial anisotropy with vario-gram analysis was proposed.Firstly,three axes of the data was solved by constructing covariance ma-trix of the sampling point data and then the data was transformed into the new coordinate system byrotating transformation;the range of each direction was calculated using geostatistical variograms;thethree values of range was set as three axes of the ellipsoid;at last,node RBF at each sample point wasbuilt.The attribute values of interpolation were solved by linear combination of node RBF.Experi-mental results show that the proposed method is a feasible method for 3Dspatial interpolation consid-ering anisotropy with high accuracy and reliable interpolation result.
In Geographic Information Systems(GIS),the exploration of the metric descriptions for to-pological spatial relations has been an active area of research.Construction processing of a metric de-scription is directly influenced by spatial data model.Vector and raster data models are the two typesof basic spatial data models.These two data models have complimentary advantages in terms of de-scribing spatial relations between objects.The integrative data model of vector and raster stems fromthe integration of the advantages of vector and raster data model.Firstly,this paper defines qualita-tive topology relations by using the 9-intersection model.Secondly,the ratio of the grid number of in-tersection to the two objects,is used to determine the intersect component.Thirdly,the maximumand minimum distances are used to determine the closeness component.Finally,a triple group inclu-ding qualitative topology relations,intersect component and closeness component,is proposed to de-scribe topology spatial relation.Because of two advantages of integrative data model of vector and ras-ter,the metric description of topology between different type objects can be realized more effectivelyin this paper.
Matching and updating rapidly changing road networks attracts much attention.Traditionalalgorithms for road network matching mostly focus on scattered segments matching,leading to poor matching integrity and low matching accuracy.After analyzing road network matching relationships,human cognition of the objective world from global to local was applied to road network matching.The algorithm puts scattered segments together with a Stroke model,and then detects the candidatematching sets by using the relative position of lines and regions,selects matching candidates and guar-antees global matching.In order to solve the uncertainty problem when setting weights in differentsimilar indexes,we analyzed the similar indexes for road network matching using qualitative and quan-titative analysis methods.We assigned weights automatically,based on the analysis of the relationshipbetween different indexes using the decision and thinking as a analytical hierarchy process.Automaticmatching was achieved by the global similarity.Results show that matching efficiency and accuracy aswell as the automation level can be improved by using an analytical hierarchy process.
The mining of co-location patterns is a hot issue in the field of spatial data mining.Howev-er,a little attention has been paid to the co-location patterns between network spatial phenomena.This paper extends an existing approach to mining the co-location patterns between network spatialphenomena.The approach consists of two core stages:①ｄｅｆｉｎｉｎｇ ａ ｃｏ－ｌｏｃａｔｉｏｎ ｍｏｄｅｌ ｔｏ ｈａｖｅ ｃｏ－ｏｃ－ｃｕｒｒｅｎｃｅ ｒｅｌａｔｉｏｎｓ ｂｙ ｐａｒｔｉｔｉｏｎｉｎｇ ｔｈｅ ｎｅｔｗｏｒｋ；②ｃｏｍｐｕｔｉｎｇ ｔｈｅ ｓｔａｔｉｓｔｉｃａｌ ｄｉａｇｎｏｓｔｉｃｓ ｆｏｒ ｔｈｅｓｅ ｃｏ－ｏｃ－ｃｕｒｒｅｎｃｅ ｒｅｌａｔｉｏｎｓ．Ｔｈｅ ａｐｐｒｏａｃｈ ｈａｓ ｂｅｅｎ ａｐｐｌｉｅｄ ｔｏ ａ ｃａｓｅ ｓｔｕｄｙ，ｗｈｉｃｈ ｄｅａｌｔ ｗｉｔｈ ｔｈｅ ｍｉｎｉｎｇ ｏｆ ｔｈｅｃｏ－ｌｏｃａｔｉｏｎ ｐａｔｔｅｒｎｓ ｏｆ ｍａｎｕｆａｃｔｕｒｉｎｇ ｆｉｒｍｓ ｉｎ Ｓｈｅｎｚｈｅｎ Ｃｉｔｙ，Ｃｈｉｎａ．Ｔｈｅ ｃｏ－ｌｏｃａｔｉｏｎ ｐａｔｔｅｒｎｓ ｈａｖｅｂｅｅｎ ａｎａｌｙｚｅｄ ｑｕａｌｉｔａｔｉｖｅｌｙ ａｃｃｏｒｄｉｎｇ ｔｏ ｔｈｅ ｔｈｒｅｅ ｍｅｃｈａｎｉｓｍｓ ｄｅｒｉｖｅｄ ｆｒｏｍ ａｇｇｌｏｍｅｒａｔｉｏｎ ｅｃｏｎｏｍｉｃｓ．Ｔｈｅ ｖａｌｉｄａｔｉｏｎ ｏｆ ｔｈｅ ａｐｐｒｏａｃｈ ｈａｓ ｂｅｅｎ ｖｅｒｉｆｉｅｄ ｂｙ ｔｈｅ ｃｏｍｐａｒｉｓｏｎ ｗｉｔｈ ｔｈｅ ｅｘｉｓｔｉｎｇ ｍｅｔｈｏｄ ａｎｄ ｔｈｅｎｅｔｗｏｒｋ ｃｒｏｓｓ Ｋ－ｆｕｎｃｔｉｏｎ．
One of the important problem in the field of pipeline 3Dmodeling is how to calculate the co-ordinates of vertices on the surface of pipelines,especially surface processing of piping connections.According to the connectivity topology of pipeline networks,using the“Sweep”modeling method cangenerate the whole surface of the pipe and establish the pipe model automatically without having todeal with the issue of joining patches.Through recursion after filleting and subdividing the pipe cen-terline to establish continuous frames at sub-points,vertices on surface corresponding to sub-pointscan be calculated,and then use the boolean union method to generate pipeline joints.The results showthat the coordinates of vertices on the surface is correct,and the entire surface is continuous andsmooth.The sweeping method reduces the steps of establishing the pipe model and has a better per-formance,and it achieves a good effect in the practical applications.
To overcome the singularity of traditional formulae of Gauss projection on the sphere in po-lar regions,the traditional formulae were improved and lastly nonsingular formulae of Gauss projec-tion in polar regions were carried out with introduction of colatitude.Based on the nonsingular formu-lae,equations of meridian and parallel were got.Compared with gnomonic projection,analysis oflength deformation and meridian deviation angle of Gauss projection were done.It’s showed thatwhen the colatitude was small,Gauss projection was similar to gnomonic projection,in other words,graticules of the two projections were approximate.Length deformation of Gauss projection was smal-ler than gnomonic projection’s,and the largest meridian deviation angle of Gauss projection from gno-monic projection is 2.468 8°in polar circles,while the largest value above the line of latitude=80°is0.438 6°.The nonsingular formulae in polar regions could satisfy the need for continuous polar projec-tion and provide theoretical basis for polar charting.