有关设计生成数字地面模型的几点建议英文.pdf

《有关设计生成数字地面模型的几点建议英文.pdf》由会员分享，可在线阅读，更多相关《有关设计生成数字地面模型的几点建议英文.pdf（5页珍藏版）》请在得力文库 - 分享文档赚钱的网站上搜索。

1、Guidelines for Conception of the DT M Projects有关设计生成数字地面模型的几点建议张伯兰,陈鹰N tyam e Jean Champlain,CHEN Ying(同济大学测量系,上海200092)Abstract:W ith regard to the effects that influent the precision of the DTM,a number of experiences hasbeen made in this sense and has given possibilities to assi m ilate all conto

2、urs of the crucial problem of theprecision of the DTM s perfectly.These experiences have been led by different centers and have succeededallmore or less to the si m ilar results.Its now recognized by unani m ity that the accuracy of the pri marydata,the density of the pri mary data,the characteristi

3、cs of the terrain,and the modeling approaches arethe principal factorsw hich affect greatly the accuracy of theDTM s.A lthough these factors act together onthe precision of theDTM,it isw hen w e isolate them and w e analyze them that the influence of each on theprecision of the DTM is exposed clearl

4、y.This paper analyses the results of the most typical experiences and outline some useful subjectionsthat can be used to prepare a reasonable project w hich w ill assure the efficiency of the generation ofDTM,of the point of econom ic view,ti me and quality.Keywords:DTM,pri mary data collection,M od

5、elingM ethods摘要:关于影响数字地面模型精度的因素问题,不同研究机构的科研人员进行过大量的实验研究,并成功获得了影响数字地面模型精度的因素,研究成果有着或多或少的相近之处,普遍认为原始数据点的三个属性(精度、密度和分布),地形特征以及建模方法是影响数字地面模型精度的主要因素。本文分析了一些典型研究的成果,并列出一些建设性的提议,以提高数字地面模型生成过程中经济上、时间上和质量上的效率。关键字:数字地面模型;原始数据点的采集;建模方法中图分类号:P2文献标识码:A文章编号:1000-3177(2002)67-041-051IntroductionD igital terrain mo

6、dels provide essential basicinformation for all types of engineering tasks andtherelativeplanningactivities.V ariousapplicationsfieldssuchasroadandrailw ayprojects,tasks of rural engineering,flood control,geology,m ilitary operations,m ining engineering,landscapearchitecture,andnumerousotheractiviti

7、es are using frequently the D igital TerrainM odels.Recently,w iththepopularityofgeographical information system s,D igital TerrainM odels have also become an i mportant part of anational spatial data infrastructure.By becom ing part of a national spatial datainfrastructure,DTM w ill not be generate

8、d only tosatisfy to some small areas covering some specificprojects as above mentioned,but also to cover aw holecountry.Forsuchalargeproject,photogrammetrist w ho are usually the responsibleof generation of the D igital Terrain M odel,musthave some guidelines for establishment of the plan收稿日期:2002-0

9、6修订日期:2002-07作者简介:张伯兰,(1970),男,航空摄影测量与遥感专业硕士,同济大学喀麦隆留学生。142002.3GIS技术遥感信息 1995-2005 Tsinghua Tongfang Optical Disc Co.,Ltd.All rights reserved.for DTM s production.The principal objective isthe generation of a high quality of D igital TerrainM odel in a li m ited ti me,w ith less cost and in themore

10、 easy conditions.The facility of production,the cost of the production and the rapidity ofproduction w ill bethe mosti mportant factorstoconsiderduringthephaseofprojectplanestablishment.Since D igital Terrain M odel came into use,numerous papershavebeenpublishedinthejournalandconferenceproceedingstr

11、eatingdifferent problem s of D igital TerrainM odel.M anyexperi mentaltestshasbeentestedby manypersons.The conclusion ofthose experi mentaltests has provided some considerable guidelines forthe specific DEM project.The first section of thispaper outlines the presentation of those differentexperi men

12、tal tests and their final conclusion.Thisw ill be follow ing by an intensive discussion w hichshould give to the establishment of some guidelinesnecessary to the production of the DEM s2Overview of D igital Terrain M odelD igital terrain model is the representation ofthe terrain by using georeferenc

13、e coordinates(X,Y,Z),Z is representing the height of the terrain inpoint w ith coordinates(X,Y).Compare to thesi mple map,DEM is the numerical representationof the map,it can be introduced into the computerandbeusedbynumberofsoftw aresforspecificpurposes.DEM w as created in 1956 by professorM iller.

14、M any others persons have worked on that subjectand their work have greatly i mprove the quality ofthe DEM.M ultiple data acquisition methods suchasTerrestrialmeasurement,photogrammetrymeasurement,measurement from map digitize hasbeen the most used for the generation of DTMduring the century.W ithth

15、efastdevelopmentincomputerscience,photogrammetry measurement methodshasbeenw idelyi mproved.M anualphotogrammetric measurements has been replacedby automatic photogrammetry.By this reason thecrucial problem of DEM accuracy w ill concern onlyDEMproducedbyusingphotogrammetrymeasurement for data acquis

16、ition.2.1Data AcquisitionData acquisition for DEMis the first step tobe processed,it include the definition of the typeof data source from w here pri mary data have to becollected,and also the method to use for obtainingrequested data w ith suitable accuracy.For thisphase,various data sources can no

17、w adays be usedsuch as digital contour map,aerial photographicand solid ground.V arious measurements methodscanbe employedfor example fully automateddigital photogrammetry,partial automated digitalphotogrammetry,manually method by analyticalphotogrammetrysampling,manualphotogrammetry contouring and

18、cartographic anddigitization of contours maps.Each method hassome advantages andinconveniences.Some ofthempresenttheadvantageofbeenmoreautomatic,moreeasytomanipulatebytheoperator,then more quickly than others,but w itha poor precision for the pri mary data,and theothers can get pri mary data w ith h

19、igh precision,but are very ti me consum ing.2.2M odelingThe second stepis the modeling,its theestablishment of relation betw een data them selvesto obtain a 3D view of the terrain.Two methodsareusuallyused,thefirst oneisthedirectmodeling method for form ing a T I N.The closestthree points betw een t

20、he pri mary data w ill be usedto form a small triangular surface,and the D igitalTerrain M odel w ill be formed by all those smalltriangles.Thesecond modeling methodis theindirect modeling.It used the random-to-gridinterpolationtoformagridnetwork.Afewpri mary data points w ill be used for computing

21、anewpointcalledinterpolated point.Thenallinterpolated pointsw ill be putting in a grid form tofinally represent the DEM.24遥感信息GIS技术2002.3 1995-2005 Tsinghua Tongfang Optical Disc Co.,Ltd.All rights reserved.3Experiences and D iscussionsA ccuracyofDEMisbecamethemosti mportantproblemtoberesolveinphoto

22、grammetry.A great number of papers havebeen published in journals.Both theoretical andexperi mental.Themost famous experi mental textsare presented here.3.1The Research ofthe Un iversity ofStuttgart 1979A ckermann in Stuttgartw as probably the firsttomakeexperi mentalinvestigationsofhighresolution D

23、TM data sets w hich w ere qualified fordetailedcivilengineeringprojects.Hefoundapproxi mately a linear relationship betw een DTMaccuracy and the density of observed points andexpressed the results in the formula:M2z=B2+(A 3d)2 M z=r.m.s error of interpolated heightvaluesA=terrain characteristicd=mea

24、n point distance of observed pointsB=standard error of pri mary data and noiseof terrain surface3.2ISPRS Texts 1986A nISPRS comm ission headed by Tolegardinvestigated the accuracy problem s on 6 differenttext sites by comparative measurements.Two testsites w ere covered by small scale photographs,th

25、ree by medium scales and one by large scales.Only one examplereferstoa highresolutionDTM.He summarized the result of this test asfollow s:a)Thestandarderrorinphotogrammetrically derivedDTM s is in the rangeof 0.20.4of the flying height in flat ormoderately undulating terrain.In very hilly terrainthe

26、 standard error increase to approxi mately 12.The maxi mun error is about 48 ti mes thestandard error.b)The number of blunders in aDTM varies from 0%up to 3%w here 0.5%is amedian value.Blunders occur more frequently indifficult terrain,but the size of the blunder isindependent of the terrain type.3.

27、3An Investigation by L I 1992The purpose of L Is investigation w as to findsome relationship betw een DTM accuracy and thedensity of the data set for certain types of terrainusing certain data patterns.He used three of theISPRS data sets to investigate the relation betw eenaccuracy and point density

28、,both of regular gridpoints and composite data sets,and regular gridw ithsupplementary featurespecific points.Heanalysesthedatapri marilyaccordingtotheformula of A ckermann,assum ing in contradictionw ith A ckermann that the accuracy may especiallybe influenced by the slope of the terrain.He findsth

29、e modelofA ckermannconfirmedfortwosamples,but he states for the third data set thatthe standard error versus the grid w idth is rather apow er function of greater than 2.3.4The OEEPE TestInvestigated andReported by A.Flotron and O.Koelbl in 2000The investigation treating the data capture ofD igitalT

30、errainM odelsbyphotogrammetricstereoscopic measurements.Exclusively analyticalplotters have been used for the measurement.Themain part of the work is the presentation of themeasuring methods of the different centers and adetailed precisionanalysis.Intotal 7centersparticipated in the test,each center

31、 working in 6different terrains test and using its ow n specifictechniques also applied for their routine work.The follow ing methods have been applied for thearrangement ofthe mass points:a)A daptivesampling,b)Profilemeasurements,c)M easurement of a regular grid.W ith the exceptionof two test sets,

32、all centers obtained the requiredprecision of 20 cm.How ever,in certain zonesconsiderableindividual pointerrorshavebeendetected exceeding up 4 ti mes the rootmean squareerror.Furthermore,it show ed that one has to takeintoaccountaconsiderablesystematicerrorcomponent of 510 cm.Based on the practicali

33、nvestigations and some theoretical considerationsan attempt w as made to establish a functionalrelation betw een the measured point density,theprecision of terrain approxi mation and the specificterrain type.A ccording to these considerations,the relative error of interpolated points can be342002.3G

34、IS技术遥感信息 1995-2005 Tsinghua Tongfang Optical Disc Co.,Ltd.All rights reserved.esti mated according to the formula:R=A 3d;R=coefficient depending on the supportingpoints,d=the mean point distance ofthesupporting points andA=Terrain characteristics.3.5An Investigation by Jianya Gong,ZhilinL i,Qing Zhu

35、,Ha igang Sui,and Y i Zhou in 2000The reported in this paper ai med to investigatethe variation of DTMaccuracy w ithsamplinginterval,sourcedata(accuracy,density,anddistribution),terrain type,and modeling type.A ccordingly,five plans for an experi mental testingof DTM accuracyw ere established in eac

36、h ofw hichonly one of the five factors as mentioned in aboveis used as the independent variable and the otherfour w ere used as controlled variables giving issueof the follow ing three testing;test reporting thevariation of DTM accuracy w ith Source Data andTerrain Type,test reporting variation of D

37、TMaccuracy w ith Sampling Interval and test reportingvariation of DTM w ith M odelingM ethods.A systematic analyzes of the above mentionedexperiences are made for the effects that influencethe DEM accuracy,and the follow ing observationsare obtained:(1)Effect from Source Data.In above experience,the

38、 source data w eremeasuredbydifferent methodssuchasfullyautomateddigitalphotogrammetry,partiallyautomateddigitalphotogrammetry,analyticalphotogrammetry,manuelphotogrammetrycontouringandcartographicscanninganddigitization of contour maps,and have resulting indifferent accuracies.The DEM directly meas

39、uredfromthe stereo models on an analytical plotterhave highest accuracy.A ccording to RM SE,DEMobtained by this method is in average tw ice ti mesbetterthanthatobtainedbyusingothermeasurement methods and the DEM generated bythe automated photogrammetric system have thepoorest accuracy,because automa

40、ted measurementusing i mage matching techniques could generatesystematic errors This is true for any type ofterrain.W hen using automated photogrammetricsystem,editing by experienced operator should beconsidered.(2)Effect of Density.The RM SE value of DEM s increases w henincrease the sampling inter

41、val.It appears that theincrease is quite linear in the hilly areas.How ever,the change of RM SE become very small w hen theterrain is very flat.For data collected fromthecontours line,the RM SE of DEM s increase w henincrease in contour interval and the increase in theRM SE is quite linearly proport

42、ional to the increasein the contour interval.(3)Effect of M odelingM ethods.D irect modeling from pri mary measured datato form triangular network(T I N)w ill yield betterresults than indirect modeling using a random-to-grid interpolation to form grid network.This isbecause the interpolated points c

43、omputation causecertainly aloss of accuracyinthe DEMThedifference could be more significant if the terrain ismostly rough.4ConclusionsA sithasbeensaidinabove,V ariousapplications fields are using frequently the D igitalTerrain M odels.W ith regard to the quality of theprecision of the DEM,every doma

44、in of applicationhas particularrequirements.Someapplicationsfieldw illratherpresentsomerequirementsrigorous,w hereas others w ill only be very flexibleof it.Depending of the specific DEMaccuracy,reliefandothersnaturalsparameters,thephotogrammetrist w ill be therefore calls to do aconception of the p

45、roject that must answ er theneeds of quality,econom ic and of facilitates.W hatdata source to be use?W hat measurement methodw ill be preferable and w hat modeling process w illthe most appropriate to reach the objectivementionhigher?Some fewsubjection have been givinghere:It seem s clear that the h

46、ighest accuracy ofDEM w ill be generated by directly measuring allnecessary pri mary data,include feature pointsfrom stereo models on an analytical plotter,andthen modeling to formtriangular network.Thisdesign w ill certainly has a high accuracy,because44遥感信息GIS技术2002.3 1995-2005 Tsinghua Tongfang O

47、ptical Disc Co.,Ltd.All rights reserved.ofthe use ofthe most accurate measurementmethod and alsothe computing by directly byT I N.But measurement on a stereo plotter is goingpoint by point.The ti me and cost of such designproject w ill then be greatly increased in greatsurface.The use of an automate

48、d photogrammetricsystem such asV irtuozo N t w ill greatly resolve thequestion of ti me by automatically measured a greatnumber of data according to the quality of therelief.This operation can be finishinginfewseconds,but the density and the distribution of thedata may not respond to the requirement

49、 of therelief,some specific points so called feature pointsw hich can more naturally describe the terrain mayhave been m issed during the automation.In thecase of urban and dense vegetation areas,somepoints maybe automatically appeared ontop ofbuilding or on top of trees,but not on the solidground.I

50、n these cases editing by an experiencedoperator w ill include some points and w ill deleteotherspointstothepreviousobtainedautomaticallybyusingmanuallymeasuringtechnique.W hen terrain relief is very rough.Because of agreatnumberofdatatobemeasured,themeasurementbasedonastereoplotterw illcertainlybeve