河源市cm级似大地水准面模型的确定

地理空间信息GEOSPATIAL INFORMATIONJan.,2020Vol.18,No.1河源市cm级似大地水准面模型的确定

洪　伟1，邱清清1，周建营1

（1.广东省国土资源测绘院，广东 广州 510500）

摘 要：综合利用地球重力场模型、数字高程模型、重力数据和高程异常数据，采用“移去-恢复”法和三次多项式拟合算法，确定了河源市cm级似大地水准面模型；研究了地形复杂的城市级似大地水准面模型计算的关键技术难点，并提出了优化方法。关键词：似大地水准面模型；重力异常；高程异常；重力场模型中图分类号：P223 文献标志码：B

文章编号：1672-4623（2020）01-0094-04

高精度局部似大地水准面为测绘学、地球物理、地球动力学以及海洋学等学科的发展和应用提供了重要的基础地球空间信息

[1]

为10.5 mGal，比EGM96模型的高程异常精度提高了3～5倍3；采用航天飞机雷达地形测绘任务（SRTM）3″×3″高分辨率DEM数据进行地形改正4。

[]

[]

。目前，高精度似大地水准

面的确定主要是综合利用高阶地球重力场模型、高分辨率数字地形模型和重力数据，先采用“移去-恢复”法确定重力似大地水准面模型，再利用GNSS/水准数据对重力似大地水准面进行拟合，获得与地区正常高系统定义一致的似大地水准面2。

河源市地处广东省东北部，总面积为1.58万km2，地形以山地、丘陵为主（约占%），辖区内包含面积约为1 600 km2、总体积约为139亿m3的巨大水体万绿湖，地形起伏大，GNSS、水准测量观测环境复杂，是研究复杂环境下高精度局部似大地水准面模型计算和优化的典型案例。

[]

2 河源市似大地水准面的确定

2.1 高精度GNSS控制网

项目完成了河源市GPS框架网以及基本网建设，建立了高精度的三维坐标基准。该网由框架网（B级网）和基本网（C级网）构成，共包括111个GPS站点。

基线解算采用GAMIT10.5进行，框架网共计算了12个同步时段，NRMS最大值为0.177 94；基本网共计算了29个同步时段，NRMS最大值为0.233 13。

平差采用GLOBK在CGCS2000坐标系下进行，框架网点坐标分量在X方向的平均精度为0.000 8 m，Y方向的平均精度为0.001 6 m，Z方向的平均精度为0.000 8 m；基本网点坐标分量在 X方向的平均精度为0.001 5 m，Y方向的平均精度为0.002 9 m，Z方向的平均精度为0.001 5 m5。2.2 二等水准测量

项目构建了全河源市范围的二等水准网，建立了高精度高程基准，共布设二等水准点190个，观测水准路线80条，总长度为3 048.2 km。平差结果为：每km中误差为±0.87 mm，最弱节点高程中误差为±6.90 mm。2.3 重力数据归算

项目共搜集河源市及其周边3 903个离散重力点，所有重力点数据都进行了坐标基准和重力基准的转 换

[6]

[]

1似大地水准面模型计算流程

似大地水准面模型计算流程如图1所示。

GPSⅠ󰛲󱵨㑽䛹󰟇󱐮󱍤󰣮㔯䛹󰟇󰱦Ὅ󰲷Ą⼧󰣧󰀎󱖎󰸹ą∁''5󱞬󱱛󰹆䶥󱐻󱠋󰤴䛹󰟇󰹓󰱜Ⅰ󰛲䲎DTM󱠋󰤴󰤺󰹓󰱜Ⅰ󰛲䲎Shepard󱠋󰤴󰞕҅󰹓󰱜Ⅰ󰛲䲎󱵨㑽󰞕҅󰹓󰱜Ⅰ󰛲䲎󰤌󰟌喋Ѩ喌󰹓󰱜Ⅰ󰛲䲎Ὅ󰲷图1　似大地水准面模型计算流程图

本文采用的地球重力场模型为EGM2008，其高程异常精度在我国整体约为20 cm，空间异常精度整体约

收稿日期：2019-04-12。

项目来源：广东省基础测绘“十三五”专项资助项目。

。对全部重力点进行空间重力异常和布格重力异

常的计算，结果如表1所示。

第18卷第1期洪　伟等：河源市cm级似大地水准面模型的确定·95·

表1　离散点重力异常统计结果/mGal

统计项最大值最小值平均值均方差标准差空间重力异常62.585-46.115-15.72621.110±14.082布格重力异常-14.931

-74.502

-42.516

43.926

±11.030

采用Shepard曲面拟合法将离散重力点的布格重力异常分别拟合为30″、1′、2′30″三种分辨率的格网[7]

，结果如表2所示；再利用格网对离散点进行插值，将插值结果与离散点的布格重力异常进行比对，结果如表3所示。30″×30″分辨率的布格重力异常格网等值线如图2所示。

表2　格网布格重力异常的统计结果/mGal

格网间隔

最大值最小值平均值均方差标准差

30″-15.023-74.485-42.263.810±11.5271′-15.023-73.074-42.27143.817±11.5392′30″

-16.304

-70.409

-42.313

43.866

±11.570

󰀎12

󰀎24

󰀎36

󰀎48

󰀎60

󰀎72

图2　30″×30″分辨率布格重力异常格网等值线图/mGal表3　格网布格重力异常与实测布格重力异常的比较结果/mGal格网间隔

最大值

最小值平均值均方差标准差30″3.348-2.914-0.0060.434±0.4341′5.924-4.2550.0020.876±0.8762′30″

10.057

-7.213

-0.024

1.668

±1.667

采用“移去-恢复”法计算格网空间重力异常，首先从离散重力点的重力异常中消除利用高阶地球重力场模型计算的中长波重力异常以及利用DTM计算的地形改正（短波重力异常），得到离散重力点的残差重力异常；然后对残差重力异常进行拟合，得到格网残差重力异常；最后在格网残差重力异常中恢复重力场模型和DTM的贡献，获得格网空间重力异常。

离散重力点的残差重力异常为：

δg=∆gFA−∆gGM+TC

（1）

式中，∆gFA为空间重力异常；∆gGM为模型重力异常；TC为地形改正。

hxy

TC(i,j)=−G∫∫

ρ(x,y,z)(hij−z)

Eh∫

ij

r3(xx,ydxdydz（ 2）

i−j−y,hij−z)

式中，G为地球引力常数；ρ(x,y,z)为流动点的地壳密度；hij为计算点(xi,yj)的高程；E为积分区域；r(x,y,z)=

x2+y2

+z

2[8]

。

首先基于30″分辨率的SRTM数据，采用2D-FFT和100%填零技术计算了30″分辨率的地形改正，并对30″分辨率的地形改正取平均分别得到30″、1′和2′30″分辨率的地形改正，统计结果如表4所示；然后采用加权平均法，通过3″分辨率的地形改正内插离散重力点上的地形改正；再利用式（1）计算离散重力点上的残差重力异常，统计结果如表5所示。

表4　格网地形改正的统计结果/mGal

格网间隔最大值最小值平均值均方差标准差3″

48.3550.0001.7062.5±1.91430″34.6670.0141.6992.558±1.9121′31.6750.0141.6962.552±1.9072′30″

113.400

0.000

1.6

5.741

±5.487

表5　重力点上残差重力异常的统计结果/mGal最大值最小值平均值均方差标准差69.690

-47.802

-6.711

13.345

±11.534

采用加权平均法，通过离散点上的残差重力异常分别拟合30″、1′和2′30″格网的残差重力异常，恢复相应分辨率的格网模型重力异常和地形改正，获得30″、1′和2′30″格网的空间重力异常。统计结果见表6。

表6　重力异常分量的统计结果/mGal

重力异常分量最大值最小值平均值均方差标准差残差重30″85.627-52.946 -6.63611.316±9.166力异常1′75.587-51.374-6.65611.966±9.9442′30″48.620-41.425-6.61811.861±9.843模型重30″49.041-35.315-6.29715.719±14.403力异常1′49.041-35.106-6.30715.718±14.3972′30″48.998-35.677-6.34416.145±14.847地形 30″34.6670.0141.6992.558±1.912改正1′31.6750.0141.6962.552±1.9072′30″113.400

0.0001.65.741±5.487空间重30″83.921-49.583-14.77720.126±13.663力异常

1′74.827-47.347-14.80520.248±13.813 2′30″

.6

-43.693

-14.793

20.398

±14.045

利用30″、1′和2′30″格网的空间重力异常分别内插离散点上的空间重力异常，并将内插值与实测值进行比较，统计结果如表7所示，可以看出，30″格网的结果明显优于1′和2′30″格网结果。

表7　格网空间重力异常与实测空间重力异常的比较结果/mGal格网间隔最大值最小值平均值均方差标准差30″10.386-12.5830.1291.399±1.3931′14.711-28.10.2552.616±2.6042′30″

40.931

-67.917

0.148

7.096

±7.094

2.4 重力似大地水准面计算

在综合利用高阶重力场模型、局部重力数据和高分辨率DTM确定区域大地水准面的实际计算中， -恢复”法被证实是目前的标准算法，即

“移去·96·地理空间信息

残差大地水准面和DTM的间接影响。

第18卷第1期

N(φP,λP)=NGM(φP,λP)+NRES(φP,λP)+NT(φP,λP)（ 3）式中，NGM、NRES和NT分别为重力场模型大地水准面、

NRES由残差重力异常按Stokes积分公式计算，即

（4）

∆φ∆λRN−1

=NRES(φP,λP) ∑F{F{δg(φn)cosφn}F{S(φP,φn,∆λ}}

4πγn=0

地形对大地水准面的间接影响的计算公式为：

35

πGρ2Gρh3−hP3Gρh5−hP−h−dxdy+dxdy+ N(φ,λ)=TPPγP6γ∫∫3Er式中，h、hp分别为流动点和计算点的高程。

在实际计算和应用中，按上述方法计算的重力大地水准面应转换为重力似大地水准面，转换关系为：

=

ς−Ng−γ∆gBGγH≈−

γH

（6）式中，g、γ分别为平均重力和正常重力

[9-10]

。

利用上述方法计算的30″、1′和2′30″分辨率重力大地水准面各分量的统计结果见表8，并按照式（6）将重力大地水准面转换为重力似大地水准面。 30″×30″分辨率重力似大地水准面模型的曲面图如图3所示。

表8　重力大地水准面的统计结果/mGal

重力大地水准面分量

最大值

最小值平均值均方差标准差残差大地

30″

-0.317-0.876-0.5830.593±0.107水准面

1′

-0.325-0.877-0.5880.597±0.1082′30″-0.335-0.888-0.5940.603±0.109模型大地

30″

2.782-6.234-1.7262.621±1.972水准面

1′

2.781-6.234-1.7262.628±1.9822′30″2.780-6.239-1.7262.652±2.013地形直接

30″

0.2450.0910.1620.1±0.026影响

1′

0.2380.0920.1620.1±0.0262′30″0.2490.0950.10.166±0.027地形间接

30″

0.000-0.068-0.0070.011±0.008影响

1′

0.000-0.067-0.0070.011±0.0082′30″0.000-0.059-0.0070.011±0.008重力大地

30″

2.372-6.633-2.3163.050±1.984水准面

1′

2.357-6.0-2.3213.058±1.9922′30″

2.329

-6.659

-2.327

3.083

±2.022

20

󰀎2󰀎4󰀎624e44Ą

114e14Ą

23e57Ą

114e55Ą

115e35Ą

23e10Ą

图3　30″×30″分辨率重力似大地水准面模型曲面图/m

2.5 重力似大地水准面和高程异常控制点的拟合

采用全部137个高程异常控制点进行重力似大地

40γ∫∫r5 （5）

E

水准面插值，并将插值结果与高程异常值进行比对，结果如表9所示。

表9　高程异常控制点的高程异常与重力似大地水准面插值

比较结果/m格网间隔最大值最小值平均值均方差标准差30″-0.263-0.703-0.5080.514±0.0761′-0.282-0.714-0.5130.518±0.0752′30″

-0.302

-0.742

-0.522

0.528

±0.078

根据规范要求，城市级似大地水准面相邻检验点间距不超过30 km，检验点总数不少于20个。项目计算时，抽取25个（约总点数的20%）点作为检验点，检验点均匀分布测区范围，间距满足规范要求（实际平均间距约为25 km），且兼顾各种地形特征

[11]

。

利用高程异常控制点与已确定的重力大地水准面模型进行拟合，消除重力似大地水准面与似大地水准面之间的系统偏差；由似大地水准面模型计算的各检验点高程异常与实测高程异常不符值计算的中误差确定似大地水准面精度。

项目分别采用二次多项式、三次多项式、多面函数、双三次B样条等模型进行拟合，根据拟合精度以及拟合参数显著性结果，挑选最优的拟合方案，最终确定采用三次多项式拟合的结果作为项目最终似大地水准面成果

[12]

。因篇幅原因，本文仅列出三次多项

式拟合结果比对。采用三次多项式模型进行拟合的内符合精度见表10。将25个检验点作插值，进行外部检核，统计结果见表11、图4。最终确定的河源市似大地水准面模型的格网分辨率为30″×30″，精度优

于1.6 cm。30″×30″分辨率似大地水准面模型如图5所示。

表10　三次多项式模型拟合结果内符合精度统计/m格网间隔最大值最小值平均值均方差标准差30″0.008-0.00500.002±0.0021′0.012-0.02300.004±0.0042′30″

0.026

-0.037

0

0.010

±0.010

表11　三次多项式模型拟合结果检验点插值精度统计/m格网间隔最大值最小值平均值均方差标准差30″0.033-0.030-0.0010.016±0.0161′0.030-0.038-0.0020.018 0±0.018 02′30″

0.043

-0.037

-0.002

0.024

±0.024

第18卷第1期

0.1洪　伟等：河源市cm级似大地水准面模型的确定

󰀔󰀑ą󱵨㑽󰀒Ą󱵨㑽󰀓Ą󰀔󰀑ą󱵨㑽·97·

行精度分析、坐标基准和高程基准转换以及空间重力异常、布格重力异常等的归算。

3）格网分辨率的设置。采用多种分辨率格网进行试算，最终确定最优方案。通常城市级似大地水准面

䒯󱌚󰀐m0模型采用30″（或1 km）左右格网能达到最优效果。

4）外部检核高程异常控制点的选取。根据规范要求，确保参与拟合的高程异常控制点数量、边距、对不同类型地形的兼顾等能满足需要。

图4　三次多项式模型拟合结果检验点精度/m

󰀓󰀏󰀕

󰀎0.15）拟合算法的选择。在实际计算中，需采用多种拟合算法进行计算和比较，综合检核点精度、参数显著性、模型合理性等多种因素确定最终方案。

󰀒󰀏󰀓

󰀑󰀏󰀑

参考文献

[1] 宁津生,罗志才,李建成.我国省市级大地水准面精化的现状

及技术模式[J].大地测量与地球动力学, 2004,24(1):4-8

[2] 管泽霖,管铮,黄谟涛,等.局部重力场逼近理论和方法[M].

北京:测绘出版社, 1997

[3] 章传银,郭春喜,陈俊勇,等. EGM2008地球重力场模型在中

国适用性分析[J].测绘学报, 2009,38(4):283-2

[4] 张兴福,刘成.综合EGM2008模型和SRTM/DTM2006.0剩余

地形模型的GPS高程转换方法[J].测绘学报,2012,41(1):25-32[5] 鄂栋臣,詹必伟,姜卫平,等.应用GAMIT/GLOBK软件进行

高精度GPS数据处理[J].极地研究,2005,17(3):173-182

[6] 曾安敏,孙凤华,郭玉良. 2000国家重力基本网与1985国家

重力基本网的转换关系研究[J].海洋测绘,2006,26(3):7-9[7] 徐遵义,姜玉祥,赵亮,等.改进的Shepard算法及其在重

力异常插值中的应用[J].武汉大学学报(信息科学版), 2010,35(4):477-480

[8] 罗志才,陈永奇,宁津生.地形对确定高精度局部大地水准面

的影响[J].武汉大学学报(信息科学版),2003,28(3):340-344[9] 李建成,陈俊勇,宁津生,等.地球重力场逼近理论与中国

2000似大地水准面的确定[M].武汉:武汉大学出版社,2003[10] 宁津生,罗志才,杨沾吉,等.深圳市1 km高分辨率厘米级高

精度大地水准面的确定[J].测绘学报,2003,32(2):102-107[11] 区域似大地水准面精化基本技术规定:GB/T 23709-2009[S].[12] 高原,张恒璟,赵春江.多项式曲面模型在GPS高程拟合中

的应用[J].测绘科学,2011,36(3):179-181第一作者简介：洪伟，高级工程师，主要从事大地测量技术应用。

󰀎1.2

󰀎2.4

󰀎3.6

󰀎4.8

󰀎6.0

图5　30″×30″分辨率似大地水准面模型/m

3结　语

本文通过确定河源市似大地水准面模型，研究了地形

复杂的城市级似大地水准面模型建立过程中的关键技术问题。对于该类项目，确保其精度的关键因素主要包括：

1）提高观测精度，优化GNSS控制网和水准路线设计。项目中框架网、基本网设计精度分别为GPS-B和C级，实际观测时将基本网观测时间延长到23 h，以确保GNSS控制点精度，尤其是大地高精度。同时优化网型和路线设计，确保在地形起伏剧烈或有巨大水体的区域周边有足够数量高程异常控制点。

2）重力数据的归算和选择。对搜集的重力数据进（上接第80页）

[5] Kanuganti S, Sarkar A K, Pratap S A. Quantifying Accessibility to Health Care Using Two-step Floating Catchment Area Method (2SFCA): a Case Study in Rajasthan[J]. Transportation Research Procedia,2016(17):391-399

[6] Matthew R M. Spatial Accessibility of Primary Health Care

Utilising the Two-step Floating Catchment Area Method: an Assessment of Recent Improvements[J]. International Journal of Health Geographics,2012,11(1):50

[7] Nakamura T, Nakamura A, Mukuda K, et al. Potential

Accessibility Scores for Hospital Care in a Province of Japan: GIS-based Ecological Study of the Two-step Floating Catchment Area Method and the Number of Neighborhood Hospitals[J]. BMC Health Services Research,2017,17(1):438

[8] 程顺祺,祁新华,林涵,等.移动搜索法在教育公共服务可达性

评价中的改进与应用:以福州城区幼儿园为例[J].人文地理, 2017,32(3):53-60

[9] 岳邦佳,林爱文,孙铖.基于2SFCA的武汉市低收入者公园

绿地可达性分析[J].现代城市研究,2017(8):99-107

[10] 沈怡然,杜清运,李浪姣.改进移动搜索算法的教育资源可达

性分析[J].测绘科学,2016,41(3):122-126

[11] 柳泽,杨宏宇,熊维康,等.基于改进两步移动搜索法的县域医

疗卫生服务空间可达性研究[J].地理科学,2017,37(5):728-737[12] 丁燕杰,关丽,陈思,等.基于空间大数据的教育设施服务能

力现状评估研究:以北京市为例[J].北京测绘,2018,32(2): 196-199第一作者简介：刘楠，主要研究方向为地理信息系统与遥感。

·VI·GEOSPATIAL INFORMATIONIndex(Vol.18,No.1)

Zhuhai City has applied the detection model into the video surveillance system around the island, and integrates with GIS maps and unmanned aerial vehicle images, which can realize the real-time detection of ships around the island.

Key wordsmap and video image YOLOv2, ship detection, cluster, video surveillance, integration of (Page:85)Implementation and Optimization of Networked Photogrammetric System Based on MapMatrixGrid Abstractdisadvantages in collaborative production, project management and mass The traditional single-machine photogrammetric system has some by DA Xingimage data scheduling. The above problems can be effectively solved by means of the networked cluster calculation mode and the ability of multi-source geographical data processing in MapMatrixGrid. In this paper, we discussed the key technologies and advantages of the networked photogrammetric system construction, and optimized the production process and quality inspection method for the main aerial survey products, which could provide some ideas for the improvement of large-scale collaborative production and management efficiency of aerial photogrammetry.

Key wordsinspection (Page: photogrammetric system, networked, MapMatrixGrid, quality 90)

Determination of the Centimeter Scale Quasi-geoid Model in Heyuan City by Abstracthigh precision gravity data and elevation anomaly data, we adopted In this paper, utilizing EGM2008 gravity field model, SRTM DEM, HONG WeiRestore” method and cubic fitting method to determine the centimeter scale “Remove-quasi-geoid model in Heyuan City. And then, we researched the key technical problems of the determination of urban level quasi-geoid model with complex topography, and proposed the optimization method.Key wordsmodel (Page: quasi-geoid model, gravity anomaly, elevation anomaly, gravity field 94)Application of Segmental Quadratic Curve Fitting Method in the Networking Highway Measurement Abstractby FU Cunpingconstruction and its parameter solving, improved the parametric equation of In this paper, we analyzed the existing methods of curve model the quadratic curve, and established a segmental curve fitting model. And then, we used the least squares algorithm to solve the model parameters, which could effectively avoid the ill condition problem. We used the fixed-step approximation algorithm to calculate the nearest distance between the sampling point and the fitting curve, which was as the residual value of the sampling point and the fitting curve, and segmented the fitting curve by limiting the threshold value. Finally, we checked the accuracy of curve fitting by means of cross-sampling point fitting. The result shows that this method is efficient, reliable and the precision meets the production requirement.Key wordspoint check segmental curve fitting, fixed-step approximation, cross-sampling (Page:98)Application of 3D Free Station Based on Helmert in Determination of GRP for High Speed Railway Abstractby CHEN Deli

speed railway, we used the variance component estimation theory to calculate the

Aiming at the problem of free station in the GRP measurement of high

3D coordinates of free space station with high accuracy. And then, we derived

3D indirect adjustment model of total station instrument for free station based

on Helmert and strict formula of linearized error equation, applied them into

the GRP determination ofdetail. Through calculation and analysis, we find that the 3D indirect adjustment

high speed railway, and described its working flow in

model for free station based on Helmert is correct. It is feasible to apply it into the

GRP determination of Key wordshigh speed railway.

determination (Page: free station, Helmert, indirect adjustment, high speed railway, GRP101,

)

Research on the Refinement of Provincial Quasi-geoid Abstractby RAO Caiqiang

the quasi-geoid in Hubei Province, and used 220 independent elevation anomaly

Based on elevation anomaly control points and EGM2008, we refined

points to validate the refinement results. The result shows that the mean square

error is underground engineering, etc) and natural structures (such as Karst cave) were ±0.051 m. If the influence of human activities (such as reservoir filling,

ignored, the mean square error can be optimized less than 0.04 m. The accuracy of EGM2008 gravity field model is 3Moreover, EGM2008 can significantly promote the accuracy in homogeneities ～4 times of the current one in Hubei Province. caused by the terrain factors. EGM2008 has advantages on high definition, high resolution and accuracy homogeneities, so it is recommended that the level plane of Hubei Province, even whole China, need to be refined using EGM2008.Key wordsresidual elevation anomaly, gridding, independent checkpoint EGM2008, elevation anomaly, elevation anomaly control point, (Page:104)Design of Quality Inspection System for Inefficient Land Survey Results byAbstracttechnology to analysis and design the quality inspection system for inefficient In this paper, we integrated used GIS technology and computer LU Zhibingland survey results from three aspects, such as system architecture design, database design and system function design. Through this system, we realized the automatic inspection functions of quality problems such as the completeness of the result, the correctness of the spatial reference, the standardization of the result, the consistency of the map, the consistency of the logical relationship, and assessed the quality inspection result. The system which had high operating efficiency and simple operation was widely used in practical projects, and controlled the quality of inefficient land survey results effectively.

Key words (Page: inefficient land survey, quality inspection, quality assessment

108)Research on the Algorithms of Parameter Estimation for Linear Regression on TAbstractotal Least Squares by YANG Genxinestimation of linear regression on total least squares at first, and further discussed the In this paper, we introduced the common algorithms of parameter algorithms which were proposed in literatures. Through theoretical deduction and demonstration, we found that the algorithms of literatures did not achieve expected purpose of total least squares and there was certain bias in theory. According to this, we derived two total least squares iteration algorithms for linear regression parameter, and compared with other algorithms through the example, in order to verify the correctness and rationality of algorithm presented in this paper.

Key wordsdecomposition, mixed total least squares total least squares, linear regression, iteration algorithm, singular value (Page:111)

Research on Deformation Prediction of Subway Foundation Pit Based on

Trend Analysis Model Abstractthe moving average prediction model and the exponential smoothing prediction Based on the trend analysis of time series analysis method, we built by GOU Shengrongmodel. Based on actual engineering monitoring data, we carried out the deformation prediction of foundation pit. And then, we analyzed the safety and stability of foundation pit combined with the point mutation theory in order to verify the accuracy and reliability of the prediction model. The results of example test show that the two prediction models have slightly different results due to their different parameter selection, but they all have high prediction accuracy. When

appropriate parameters are selected, a better prediction effect can be obtained. The

highest prediction accuracy is obtained by the moving average prediction model when smoothing prediction model when n=1, while the highest prediction accuracy is obtained by the exponential actual deformation of the foundation pit. Meanwhile, the fitting degree of point a=0.7. Both of them can better reflect the mutation theory reaches 0.988 9, it verifies the accuracy and reliability of the prediction model. This study can be used as a reference for similar projects.Key wordsdeformation prediction, stability analysis deep foundation pit, trend analysis, point mutation theory, (Page:115)Analysis of the Spatial Cluster Development Characteristics of Pearl River Delta Urban Agglomeration from 2000 to 2017 Abstractby WANG Haiyunto 2017, we used the geographical concentration, location quotient, urban flow Based on the images and statistics of the Pearl River Delta from 2000 intensity, economic membership grade and urban expansion to analyze the agglomeration levels of population, economy, industrial, and the spatio-temporal

evolution process of spatial relationship in Pearl River Delta urban agglomeration

from 2000 to 2017. The conclusions show that network connection in the study area is not high, it still has a clear “Core-Periphery” ①the urban agglomeration spatial

characteristic, and people agglomeration lags behind in the economic developed areas. huge differences in the intensity of urban flow between cities. Guangzhou ②The inter-city economic relations are very loose in this region, there are plays a leading role and it has a strong correlation effect. However, Shenzhen, which mainly depends on industrial outward service, has lost its advantage in the region, its comprehensive function yet need to be enhanced. of urbanization are closely related to economic agglomeration, and they ③The characteristics have high consistency in space. of the lands are obvious, but the economic links are not enough. The links ④The geographical agglomeration characteristics between Zhuhai, Guangzhou and Shenzhen should get more focused.Key wordsintensity (Page: urbanization, spatial aggregation, economic relation, urban flow 119)

Automatic Quality Inspection Method of Digital DOMAbstractdata scheduling management, we made full use of existed DOM and DLG to In this paper, based on key technologies of image matching and big by LIANG Feiimplement an automatic quality inspection method of DOM. The experimental result shows that this method has high automation and reliable accuracy, and improves the DOM inspection efficiency. The method has been applied in production practice. Key words (Page: DOM, DLG, quality inspection, automatic inspection, data linking125)Improved Edge Detection Algorithm Based on Canny Operator Abstractthe traditional edge detection algorithm in solving the problem of suppressing

In the field of image edge detection, Canny algorithm is better than

by ZHU Qiulinnoise and preserving the edge, but it still lacks in preserving the edge details and reducing the false edge. In this paper, we proposed an improved edge detection algorithm based on Canny operator. This algorithm used the bilateral filter with preservative edge denoising instead of Gaussian filter to denoise, and improved the non-maximum rejection and threshold selection. The experimental result shows that the algorithm can detect more edge details than the traditional algorithm.Key words edge detection, Canny operator, bilateral filter (Page:128)