Exhibition International Symposium & Exhibition ... [PDF]

3 D GIS : the state of the art

Exhibition International Symposium & Exhibition on Geoinformation 2002 A historical overview of the First Levelling Network in Peninsular Malaysia Object Recognition in Aerial Imagery using Moments variant for Dataset Revision

JAWATANKUASA PEMETAAN DAN DATA SPATIAL NEGARA

BIL.2

2002

ISSN 1394 - 5505

Sidang Pengarang

Penaung Dato’ Hamid bin Ali, DIMP,KMN, PMC, PJC Ketua Pengarah Ukur dan Pemetaan Malaysia

Penasihat Dr. Abdul Kadir bin Taib, SDK, KMN

Ketua Editor Teng Chee Boo

8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19.

Editor Abdul Manan bin Abdullah Abdul Hadi bin Abdul Samad Amran bin Abu Rashid Dayang Norainie Bt. Awang Junidee

Ketua Rekabentuk/ Pencetak Muhammat Puzi bin Ahmat

Jabatan Perhutanan Sabah Jabatan Perhutanan Sarawak Jabatan Pertanian Sabah Jabatan Pertanian Sarawak Pusat Remote Sensing Negara Universiti Teknologi Malaysia Universiti Teknologi MARA (co-opted) Universiti Sains Malaysia (co-opted) Jabatan Laut Sarawak ( co-opted) Universiti Putra Malaysia (co-opted) Pusat Infrastruktur Data Geospatial Negara (Ma CGb1) (co-opted) Jabatan Perancang Bandar dan Desa (co-opted)

Kandungan

MAKLUMAT Buletin GIS diterbitkan dua (2) kali setahun oleh Jawatankuasa Pemetaan dan Data Spatial Negara. Sidang Pengarang amat mengalu-alukan sumbangan sama ada berbentuk artikel atau laporan bergambar mengenai perkembangan Sistem Maklumat Geografi di Agensi Kerajaan, Badan Berkanun dan Institut Pengajian Tinggi. Segala pertanyaan dan sumbangan boleh dikemukakan kepada:

Ketua Editor Buletin GIS Bahagian Pemetaan Jabatan Ukur dan Pemetaan Malaysia Tingkat 3, Bangunan Ukur Jalan Semarak 50578 Kuala Lumpur

JAWATANKUASA PEMETAAN DAN DATA SPATIAL NEGARA

1. 2. 3. 4. 5. 6. 7.

Jabatan Ukur dan Pemetaan Malaysia (JUPEM) Jabatan Tanah dan Ukur Sabah Jabatan Tanah dan Survei Sarawak Wakil Kementerian Pertahanan Jabatan Mineral dan Geosains Malaysia Jabatan Perhutanan Semenanjung Malaysia Jabatan Pertanian Semenanjung Malaysia

MukaSurat

Dari Meja Pengarang

1

_____________________________________

3D GIS: the state-of-the art

2

_____________________________________

Object Recognition in Aerial Imagery using Moments Invariant for Dataset Revision

7

__________________________________

International Symposium and Exhibition on Geoinformation

13

_____________________________________

A Historical Overview of the First Levelling Network in Peninsular Malaysia

15

___________________________________

Nota:

Kandungan yang tersiar boleh diterbitkan dengan izin Urusetia Jawatankuasa Pemetaan dan Data Spatial Negara

Dari Meja Pengarang Selangkah demi selangkah, era teknologi semakin pantas berkembang di dunia yang global ini. Tidak ketinggalan, Malaysia juga telah mengalami perubahan drastik menerusi era teknologi terkini. Oleh yang demikian, penggunaan teknologi maklumat seperti Sistem Maklumat Geografi (GIS) tidak dapat dielakkan malahan ianya terus berkembang dan digunapakai oleh pelbagai jabatan serta agensi kerjaan dan swasta. Seperti mana yang diketahui, Sistem Maklumat Geografi atau GIS merupakan suatu sistem pangkalan data berkomputer yang mana segala maklumat geografi sesuatu kawasan dapat disimpan, diolah dan dianalisis dengan menggunakan sistem komputer. Secara langsung GIS dapat membantu kerjakerja perancangan dan pentadbiran maklumat yang berkaitan dengan harta tanah, perhutanan, pertanian, alam sekitar, pengangkutan, penerokaan minyak, gas asli dan lain-lain. Oleh yang demikian, untuk mempertingkatkan ilmu serta memperkembangkan lagi minda, Buletin GIS diterbitkan sebagai bahan percambahan ilmu demi kepentingan bersama. Keluaran Buletin GIS kali ini memaparkan pelbagai artikel yang telah dihasilkan oleh pegawai-pegawai yang berpengalaman. Antara artikel yang dipaparkan kali ini ialah 3D: the state-of-the art, Object Recognition in Aerial Imagery using Moments Invariant for Dataset Revision, A Historical Overview of the First Levelling Network in Peninsular Malaysia dan juga laporan mengenai International Symposium and Exhibition on Geoinformation 2002. Dengan pemaparan artikel kali ini, diharapkan dapat memberi manfaat kepada pembaca untuk mengaplikasikan pengetahuan dalam bidang tugas yang berkaitan. Sesungguhnya pihak sidang pengarang amat berterima kasih di atas sumbangan-sumbangan seperti ini. Semoga penerbitan Buletin GIS akan lebih mantap lagi pada masa akan datang dengan sumbangan daripada agensi-agensi yang lain pula. Untuk kemudahan pihak tuan, semua sumbangan bolehlah dimajukan kepada Ketua Editor Buletin GIS atau email kepada [email protected].

Sekian, terima kasih.

Ketua Editor Buletin GIS

1

3D GIS: the state-of-the art by Dr. Alias Abdul Rahman UTM

An increasing number of applications need advanced 3D tools for representing and analysing real world. At present, many software packages can handle a wide range of spatial problems. However, is the functional 3D GIS already a reality? This paper tries to find the answer by analysing both software available and efforts of researchers. An overview of several commercial systems and a 3D case study provide knowledge about 3D functionality offered in the market. The most significant achievements in the 3D research area concerning 3D structuring and 3D topology portray the current research status and delineate new research topics.

3D GIS in the market Nowadays, 2D GISs are widely used to handle many 2D GIS issues in a very efficient manner. However, the same kind of systems fail to operate if more advanced 3D tasks are demanded. There are already few systems available in the market that can be categorised as systems providing 3D solutions. We analysed four of them that constitute the largest share of the GIS market, i.e. ArcView 3D Analyst, (ESRI), Imagine VirtualGIS (ERDAS), GeoMedia (Integraph Inc.) and Geomatica (PCIGeomatics). All the systems provide excellent tools for 3D visualisation, animation and navigation through 3D textured models (Figure 1, 2). Most of them offer sufficient tools to manipulate 2.5D data such as surface generation, volume computation, draping images, terrain inter-visibility, etc. Access to data distributed on different servers is also greatly improved. All the systems revealed still little provision of 3D GIS functionality in terms of 3D structuring, 3D manipulation and 3D analysis. The full 3D geometry (the zcoordinate is basically an attribute), 3D topological relationships and 3D analysis are still areas to be entered by the general-purpose GIS vendor.

Introduction Currently, the most often quoted areas of human activities waiting for 3D solutions are 3D urban planning, environmental monitoring, telecommunications, public rescue operations, landscape planning, geological and mining activities, transportation monitoring, real-estate market, hydrographical activities, utility management and military applications. But the role of geo-information in all kinds of business processes is also getting quite transparent. Terms as “location-specific information” and “location-based services” become a part of the daily business language to denote the link between the virtual world of information (transactions, events, internet communication) and the real world of information (customers, inventory, shipping). Most business transactions rely on information systems as the geo-information is critical for many of them. Once the developments in the 3D GIS provide a compatible functionality and performance, the spatial information services will evolve into the third dimension. Here, we attempt to summarise the current status of 3D GIS development and suggest topics for further research and implementations.

Figure1: 3D ArcScene: 3D Visualisation

2

Spatial 8i has implementied 2D geometry types (point, line, polygon). Semantic characteristics of a feature are organised in GeoGraphics iSpatial, as one significant part of the information (semantic hierarchy, links to geometry types) is maintained at a database level. However, the notations (table names, columns, object definitions) have very specific application-oriented meaning. If the user decides to keep the database and change the CAD package, he/she will need to create the feature-geometry link from scratch. Real 3D geometric types are missing but the description of 3D data is possible. The Z value is maintained together with the X,Y values, i.e. it is not an attribute. Since the topological primitives are not implemented yet, real possibilities to analyse 3D data in GeoGraphis iSpatial and Oracle Spatial 8i are still missing. Tools in GeoGraphics iSpatial to create 2D topological layers or tools in Oracle Spatial 8i to perform spatial operations are provided but they operate with only X,Y coordinates. Some of the operations accept X,Y,Z coordinates but computations are still in 2D (Figure 3).

Figure 2: Imagine Virtual GIS: 3D Animation

Many other vendors (developing basically CAD, GIS and DBMS packages) realise the increased importance of geo-information and seek for appropriate solutions to maintain and analyse spatial data. A logical consequence of all the attempts is the agreement on the manner for representing, accessing and disseminating spatial information, i.e. the OpenGIS specifications. As a result, an increasing number of DBMS offer already storage, retrieval and analysis of spatial data. A growing number of CAD vendors develop tools to utilise the storage benefits of DBMS. Among all, we selected Oracle Spatial 8i and Microstation Geographics iSpatial to investigate the current status of the implementations. We completed a case study with 20021 buildings from the city of Vienna. The idea was to test the DBMS&CAD integration against query, visualisation and editing. Our experience is briefly summarised bellow. Both packages follow closely the OpenGIS concepts, i.e. a geographic feature has spatial characteristics that are represented by geometric and topologic primitives. Oracle

Figure 3: Oracle Spatial operator SDO_WITHIN_DISTANCE

3

Apparently the greatest benefits of the DBMSCAD integration are in the area of visualisation and editing of data. As frequently commented, the amount of data to be visualised in 3D increases tremendously and requires supplementary techniques (LOD, on-fly simplification, etc.) for fast rendering. Having 3D data stored in a database, the user has the possibility to extract only a limited set of data (e.g. one neighbourhood instead of one town) and thus critically reduce the time for loading. Locating, editing and examining a particular object become also quick, simple and convenient. Indeed, the elements that can be edited correspond to the geometry representation in Oracle Spatial 8i, e.g. the 3D topology of buildings cannot be preserved (Figure 4).

level of explicitly described spatial relationships varies. Each suggested 3D model exhibits efficiency and deficiency with respect to a particular applications and operations to be performed. Mechanisms for representing spatial relationships are yet another “hot area” of investigation. OpenGIS consortium has adopted two frameworks known as Egenhofer operators and Clementini operators based on the 9-intersection model. Although the topology is considered the most appropriate mechanism to describe spatial relationships, the study on applicability of other mathematical frameworks continues. Advances in the area of computer graphics have made visual media a major ingredient of the current interface in the communication and interaction with computers. Therefore the research related to the visualisation of real world 3D data is mostly “shifted” to the computer graphics society. Many viewers and browsers as stand-alone applications and plugins have been developed to quickly visualise and navigate through 3D models. New algorithms and implementations are reported daily. The design criteria, however, are fast rendering techniques based on internal structures rather than utilisation of database representations. TerraExplorer (SkyLine), the current leader for visualising large 3D textured data from real world with acceptable performance, also requires data restructuring (Figure 5).

Figure 4: Query, edit and post of a building

3D GIS in the research The research in 3D GIS is intensive and covers all aspects of the collecting, storing and analysing real world. 3D analysis and the related issues (topological models, frameworks for representing spatial relationships, 3D visualisation) are mostly in the focus of investigations. The 3D topological model is the key issue, since it is related to the representation of a large group of spatial relationships, e.g. inclusion, adjacency, equality, direction, intersection, connectivity. Although the large number of 3D models reported in the literature, the research is concentrated around few basic ideas, as the

Figure 5: Terra explorer: Real-time 3D

navigation trough large textured 3D models

4

Observations on the demand for 3D City models show user preferences for photo-true texturing. Currently, a limited number of packages offer means for mapping images onto geometry (e.g. facades). Trading phototrue texture requires a variety of topics to be considered, e.g. standardisation of parameters for image-geometry references and image organisation at a database level.

Another significant area of 3D GIS research is devoted to Web applications. The Web has already shown a great potential in improving accessibility to 2D spatial information (raster or vector maps) hosted in different computer systems over the Internet. The first attempt to disseminate and explore 3D data, i.e. VRML, appeared to be rather “heavy” for encoding real geo-data due to the lack of a successful compression concept. Despite the drawbacks, the language became a tool for research visualisation. Researchers could concentrate on data structuring and analysis and leave the rendering issues to VR browsers offered freely on Internet. GeoVRML (VRML extended with geo-nodes) and Geographic Modelling language (GML) are another promising opportunities for representing 3D data on the Web. Based on XML concepts, GML provides larger freedom, flexibility and operability than VRML.

The significant step, however, is made. Developers and researchers from different areas have united their efforts toward developing a 3D GIS. The understanding of GIS is changing. Instead of a monolith, desktop, individual system, GIS is becoming an integration of strong database management (ensuring data consistency and user control) and powerful editing and visualisation environment (inheriting advanced computer graphics achievements).

Discussion The study on the commercial systems showed clearly that the functional 3D GIS is not an easy task. The 3D progress in traditional GIS packages is mainly in the area of data presentation and surface analysis. DBMS&CAD&GIS developers have rapidly adopted OpenGIS specifications but the focus is mostly on the geometry. 2D topological representations and operations as well as the real 3D geometry type are still to be implemented.

Acknowledgements The authors express their gratitude to Bentley Systems Europe, The Netherlands, the Institute of Computer Graphics, TUGraz, Austria and the mapping company ADP, Graz, Austria for their contributions to the Vienna case study.

Further reading The third dimension with respect to topological issues is still in the hands of the researchers. Although quite significant number of works devoted to 3D structuring, the consensus on a 3D topological model is not achieved yet. Still a lot of 3D GIS functionality is left for addressing. 3D buffering, 3D shortest route, 3D inter-visibilities are some of the most appealing for investigation. Integration of object-oriented approaches with the 3D GIS raises new research topics toward standard object descriptors and operations at a database level.

Bentley, 2002, accessed April http:// www2.bentley.com/products/ ESRI, 2002, accessed April http:// www.esri.com/software/arcgis/index.html Oosterom, P. v, J. Stoter, W. Quak and S. Zlatanova, 2002, The balance between geometry and topology, In: Proceedings of Spatial Data Handling, 8-12 July, Ottawa, Canada OpenGIS specifications, 2002, OpenGIS Consortium, Inc., accessed April http:// www.opengis.org/ SkyLine, accessed April http:// w w w. s k y l i n e s o ft . c o m / c o r p o r a t e / corporate_home.asp

3D GIS requires appropriate means not only for visualising and exploring 3D textured models but also for building and storing them.

5

Biographies of the authors Dr. Alias Abdul Rahman is an Associate Professor and the Director, Center for Geographic Information and Analysis (CGIA) and also attached to the Department of Geoinformatics at Universiti Teknologi Malaysia. His interests are mainly on 3D GIS, DTM, and spatial data structuring and modelling. Dr. S. Zlatanova is a Researcher at TUDelft, The Netherlands. Her interests are on 3D GIS: object reconstruction, data structures, spatial relationships and visualisation. Dr. Morakot Pilouk is currently a Software Development Manager at ESRI, Redlands, California, United States.

Affiliation Assoc. Professor Dr. Alias Abdul Rahman, Department of Geoinformatics, Universiti Teknologi Malaysia, 81310 UTM Skudai, Johor, Malaysia, [email protected] Dr. Eng. Siyka Zlatanova, Delft University of Technology, Thijsseweg 11, 2629JA, Delft, The Netherlands, [email protected] Dr. Morakot Pilouk, Environmental System Research Institute (ESRI), 380 New York Street, Redlands, California, USA, [email protected]

6

Object Recognition in Aerial Imagery using Moments Invariant for Dataset Revision Teng Chee Hua

Jabatan Ukur & Pemetaan Malaysia Seksyen Geodesi, Bangunan Ukur, Jalan Semarak, Kuala Lumpur

Abstract Significant information about physical objects in a digital topographic dataset exists within the objects themselves. Shape descriptors, in particular, can be determined as characteristics of classes of object such as building, road, grassland and water body. Such descriptors also include moments invariant, indices of shape which are constant regardless of the scale, rotation or translation of objects. This paper examines the role of such moments in characterising objects in a large scale topographic dataset. Further information about objects can be obtained from their typical reflectance values on colour aerial photography. Additionally, digital elevation model may be generated during the creation of orthoimage to assist objects grouping into ground level and above-ground objects. The use of moments invariant indices in moderating the revision process for the spatial dataset is examined. It is concluded that there is some scope for incorporating such ‘amplified intelligence’ in the revision flow-line.

significant use of aerial photographic imagery, either to help ascertain where ground-based revision techniques need to be applied, or to provide a source for remote measurement and interpretation of new detail. The imminent initiation of systems for retrieving high spatial resolution data from space platforms will increase the use of imagery in the revision flow-line. Unfortunately, although measurement of accurate coordinate information from both aerial and space imagery can be easily accomplished, the meaning within the imagery may be less apparent. The existence of change needs to be detected and its nature needs to be determined. Sophisticated change detection and image interpretation It is suggested that information extracted from the spatial objects within a digital topographic spatial dataset can be used to assist in the process of image interpretation. Thus a ‘knowledge bank’ detailing the characteristics of typical features in the topographic landscape can be created and unknown features detected in the imagery can be matched against this. The information used can consist of measurements obtained directly from the coordinates of objects, for example buildings and roads, along with the spectral signatures in the imagery of objects whose existence is known.

Keywords: object recognition; aerial imagery; GIS dataset; image processing; moments invariant 1. Introduction The aim of this paper is to determine the extent to which knowledge that exists within a topographic spatial dataset can supplement and assist in the revision of digital dataset using aerial photographic techniques. Specifically, metrics have been obtained from digital data and have been used to inform the process of object recognition on, and data extraction from, aerial photography. The updating of national topographic digital datasets is currently, or inevitably will be in the future, the major task of all mapping and spatial data supply organisations.

2. The digital map data An important stage in the creation of polygonal dataset is the identification and incorporation of ‘dark links’ - non-extant lines used to closeoff areas and break up long and large objects into manageable units. For the purpose of object recognition polygon data need to entirely cover the landscape. Object-oriented techniques will be most suitable for such purpose as can be found in Ordnance Survey OS96 next generation

Traditional methods of revision involve

7

with an object in an image are written in the form

National Topographic Database (NTD2) data structure. The information that can be extracted from the above digital dataset pertaining to objects includes a range of metrics that can be used to characterise them. Mean values may be calculated for area, perimeter, compactness index (Blair and Biss, 1967) as follow:

mp q =

where f(x,y) is the density function describing the object. The integrals are taken over the entire image. The central moments are given by expressions of the form:

D Compactness Index =

¾¾x p * yq * f(x,y) dxdy

2 F Ý r2 d A A

where D = area of shape A, dA = small element of area A and r = the radial distance from the centroid to the elements of area A. The shape A is a discrete image (raster).

where

and

are the coordinates of

2.1. Calculation of moments A variety of further metrics has been applied to the digital image analysis of objects in general. Pratt (1991) described the possible use of topological attributes, distance, perimeter, area measurements, shape orientation descriptors, fourier measurements and spatial moments in shape analysis. McKeown (1984) employed criteria such as average intensity, area, compactness, linearity and template shape matching to search for certain types of regions on aerial imagery. Domenikiotis et al. (1995) applied moments invariant (particularly those they regarded as defining ‘spread’ and ‘slenderness’) to line segments detected on satellite imagery and used them in their knowledge-based system to decide whether the extracted line was valid to include as part of a derived road network dataset. In this paper, the spatial moments is used to search for objects on aerial photographic imagery and attempted to match the results against the moments calculated for each class from the enhanced NTD2 dataset.

~ = m pq

p ~q (x - ~ x) * (y - y) * f(x,y) dxdy ¾ ¾

the centroid of the object where

J Mu(p,q) = Ý

K

Ý x p *y q *f(x,y)

j =1 k =1

k

j

where J and K are the numbers of rows and columns, respectively, of the array of pixels defining the object. The various orders of moments can be equated to properties of the objects described (Schalkoff, 1989). Table 2 explains the various order of moments and interprets their meaning with respect to the nature of the shape of the object.

Moments invariant are mathematical shape descriptors which are invariant to translation, scale and rotation, yielding identical indices for the shape of geographical objects regardless of their position, size and orientation on the map. The moments invariant is able to take into account data in vector form (NTD2) and raster form (aerial imagery). For raster data, the twodimensional moments of p+q order associated

8

Central moment

Order (p+q)

Interpretation

Second-order (variance)

“horizontal centralness”

Second-order (variance)

“vertical centralness” “diagonality”; indication of quadrant (with reference to centroid) where the object has more mass

Second-ord (variance) Thirdorder (“skewness”)

“horizontal divergene”indicates the relativ extent of the left of the region compared to the right

Third-order (“skewness”)

“vertical divergence”; indicates the relative extent of the bottom of the region compared to the top

Third-order (“skewness”)

“horizontal imbalance”; location of the centroid with respect to half horizontal extent

Third-order (“skewness”)

“vertical imbalance”; location of centroid with respect to half vertical extent

Table 2: Interpretation of 2nd and 3rd order central moments

The zero-order moment is equal to the total number of all the object pixels which define its area.

three independent invariants are :

The first-order central moments

These second- and third-order moments are absolutely invariant to both proper and improper rotations. A set of seven second- and third-order moments invariant to translation, scale and rotation can be found in Hu (1962) and Pratt

,I

and

are always equal to zero which is the moment at the centroid with respect to itself. The central moments are location invariant and can be normalised to become scale invariant:

M

cos G sinG x - sin G cosG y

. The first

pq

p q = 55x *y *f(x,y)

where f(x,y) in this case is the discrete image function. One-dimensional vector moments can be calculated for the outline of an object, or indeed a line segment itself. Vector moments can take account of parts of a shape allowing for objects and models to be compared using all or only portions of their boundaries. Both open and closed line strings can be assessed, along with occluded and/or flawed objects whose global characteristics are indeterminate (Wen and Lozzi, 1993). The calculation of vector moments can be simpler and faster than for two-dimensional raster objects.

and also for the improper orthogonal transformation or rotation and reflection

x’ cos G sinG x y’ = sin G - cosG y He shows that for the second-order moments,

.

.

six combinations are all absolute invariant to translation, scale and rotations with reflection. The seventh is similar but changes sign with reflection and is thus able to detect mirrored objects. The simplest case of two-dimensional raster moments are with binary data where f(x,y) = 1 when the pixel is inside the object, f(x,y) = 0 outside. In such an image the general form for moments can be simplified to

This scale invariance is only valid for scale changes which are similar both in the x and y direction.Rotational invariance can also be achieved by combinations of the normalised central moments. Hu (1962) derived the rotations invariant by subjecting the x and y coordinates to proper orthogonal transformation or rotation:

the two independent invariants are

and

(1991).Allarecombinations of

where

x’ y’ =

21I21

and

For the third-order moments, the

9

Fig. 4: Orthophotograph 3. The image data The imagery used to assess the planned system is typical of the aerial photography used in current map revision practice. Coloured photography at a scale of 1:5000 covering the same area as the digital map data was obtained. These images were scanned at a resolution of 300 dpi. Stereo coverages were accessed, allowing for the automated creation of a digital surface model (DSM) in a digital photogrammetric system. The DSM therefore includes surfaces of elevated objects like buildings, trees and hedges. Surface fitting techniques were used on the DSM to generate the digital terrain model (DTM), which only describes the topographic surface. The DTM could also be generated from existing digital contour data. Baillard et. al. (1996) suggested the possible use of a classical “top hat” filtering method on the DSM to generate the topographic surface. Orthoimagery was generated from the scanned imagery(Fig.4) ensuringregistration with the digital map data, so that information from that source can be easily compared and incorporated with the imagery.

The need to obtain information regarding the existence and nature of differences between the map data and the image data led to an investigation of image processing and interpretation. A number of image handling techniques were examined, all making use of the Red-Green-Blue (RGB) elements of the colour image. An initial image segmentation was undertaken to denote regions of similar reflectance. Haralick and Shapiro (1985) suggest segmentation should be based on overall grey tone, significant differences between adjoining regions, simple and spatially accurate boundaries between regions and few holes and discrepancies in region interiors. There is no universal segmentation method and an empirical approach was applied to the complex aerial imagery used in this study. Use was then made of the three colour components of the aerial photograph and the RGB components were separated in the

10

5. Future investigations

processing software. The correlation among these components, however, would lead to unsuitable segmentation results and transformation analysis was attempted. Instead of the more standard Principal Components Analysis, Ohta et al. (1980) devised what they considered a more rigorous method for RGB images which creates new elements based on the RGB values as follows:

A major aim of this project is to automate the detection and determination of the nature of change in an urban landscape. The outline system above demonstrated that GIS dataset can assist the creation of metrics which characterise spatial. Automated methods should be able to match an unknown object on the imagery against a knowledge-based database and thus instinctively determine and incorporate the object into the GIS dataset. Such automated matching could be undertaken using artificial intelligence techniques. It is anticipated that such a system must be able to handle the fuzziness of models and objects extracted from the imagery. Artificial intelligence (AI) techniques could also be applied during the earlier image segmentation and classification stages: the potential of neural network methods, in particular, has already been recognised widely by remote sensing scientists, for example, Kanellopoulos and Wilkinson (1997). A future paper addresses these issues and outlines a technique developed to incorporate AI techniques into this flowline.

I1 = (R+G+B)/3 I2 = (R-B)/2 I3 = (2G-R-B)/4

These were calculated and normalised to remove negative values. A statistical filter was applied to smooth minor irregularities in the image. A height above-ground (HAG) surface is obtained but subtracting the DSM from the DTM. The HAG surfaces are used to assist image segmentation. The digital map data was used to assist object’s reflectance extraction and subsequent image segmentation. Segmentation can also be achieved through the detection of edges of the regions, locating positions of abrupt changes in intensity values. Standard applications of filtering techniques such as the Prewitt, Sobel and Navatia-Babu edge detectors were attempted. The complexity of the image, however, along with the numerous shadows and irregular vegetation features produced a high level of noise in the resultant segmented image. The necessary, subsequent linking of the edges to form continuous and uniform edge lines is difficult to automate, especially when the image is noisy.

6. Conclusion The research undertaken in this project has shown that it is possible to create a series of standardised metrics for objects held in a digital topographic dataset. These metrics have encompassed simple arithmetic measurements along with more complex moment calculation. The objective of automatically updating the digital dataset has relied on these metrics to guide the interpretation of new features detected on conventional aerial photography. Extraction of the new features relies on a series of image processing tasks and change detection procedures, resulting in definitive spatially located objects whose properties are derived from comparison with the standard metrics. This technique can be usefully applied to the updating of digital topographic datasets and it relies on established digital photogrammetric and image processing systems, it shows promise for application in many national and regional mapping agencies throughout the world.

4. Change detection

Change detection is undertaken in order to extract only those objects which were new on the aerial photography. Conventional methods of change detection rely on subtraction of the old aerial image from the new (after adjustment to ensure common scale, reference and contrast). A similar technique was undertaken in this study, by subtracting the ‘polygonised’ map data from the segmented aerial image. Objects were extracted by region growing from a seed point within the new object.

11

References Baillard, C., O. Dissard, O. Jamet, H. Maitre (1996). Extraction and characterization of above-ground areas in a peri-urban context. Proceedings of the IAPR TC-7 Workshop “Remote Sensing and Mapping”, Austria. Blair, D.J., and T.H. Biss (1967). The measurement of shape in geography: An appraisal of methods and techniques. Bulletin of Quantitative Data for Geographers. Domenikiotis, C., G.D. Lodwick, and G.L. Wright (1995). Intelligent interpretation of SPOT data for extraction of a forest road network. Cartography, 24 (2), 47-57. Haralick, R.M., and L.G. Shapiro (1985). Image segmentation techniques. Computer Vision, Graphics and Image Processing, 29 (1), 100-132. Hu, M.K. (1962). Visual pattern recognition by moment invariants. IRE Transactions on Information Theory, 8 (2), 179-187. Kanellopoulos I., and G.G. Wilkinson (1997). Strategies and best practice for neural network image classification. International Journal of Remote Sensing, 18 (4), 711725. McKeown, D.M. (1984). Map-guided feature extraction from aerial imagery. Computer Science Department, Carnegie Mellon University, Pittsburgh, USA. Ohta, Y., T. Kanade, and T. Saki (1980). Color information for region segmentation. Computer Graphics and Image Processing, 13 (3), 222-241. Pratt, W.K. (1991). Digital Image Processing (2nd Ed). John Wiley & Sons, New York, USA. Schalkoff, R.G. (1989). Digital Image Processing & Computer Vision. John Wiley & Sons, New York, USA Wen, W., and A. Lozzi (1993). Recognition and inspection of manufactured parts using line moments of their boundaries. Pattern Recognition, 26 (10), 1461-1471.

12

International Symposium and Exhibition on Geoinformation 2002 22-24th October 2002 Kuala Lumpur

The Symposium and Exhibition

About the Event

The conference was held on the 22-24th October 2002 at the Nikko Hotel, Kuala Lumpur. The 2002 International Symposium and Exhibition on Geoinformation is the fifth in the series of annual symposium, jointly organized by the Institution of Surveyors Malaysia (ISM) and four local universities (UTM, UiTM, UPM and USM). This year, UTM is hosting the Symposium at the international level. In this Symposium, we are able to attract a wide spread of delegates, from various background of geoinformation related practitioners and users. We have received significant contribution from speakers, a total of 90 technical papers altogether, in a range of topics covering Spatial Information Technologies, Precise Surveying, Remote Sensing, GPS, Geodetic Engineering, Hydrographic Surveying, Cadastre and Land Information Management. The Steering Committee for the conference was chaired Dr. Abd. Kadir Taib (Chairman, Land Survey Division of ISM), and Assoc. Prof. Dr. Shahrum Ses was a chairman for the Organizing Committee.

The symposium has been officiated by the Deputy Minister of Land and Cooperative Development. The Director General of DSMM and the Vice Chancellor of UTM were also present in the opening ceremony. The conference brought together various geoinformation communities to discuss the new development and technologies in geoinformation. We have a list of renowned members of the International Advisory Panel for vetting the selected papers. The conference agenda is divided into 9 main technical sessions. Five keynote sessions were scheduled and presented by Prof. C. Rizos (Australia), Dr. J. Drummond (Scotland), Prof. Cracknell (England), Dato’ Hamid Ali (DSMM) and Prof. Mohamad Ibrahim (UPM). Poster sessions were also planned for those who would like to explain their findings interactively. At the end of the conference, a Plenary Session has been held with the outstanding panels chaired by Prof. Ayob Sharif (UTM). The Session specifically being focused on the R&D activities in Geoinformation with particular attention given to pattern related to the field. A total of 20 companies and agencies related to Geoinformation have been participated in the exhibition. The main sponsors for the Symposium are the PEJUTA, Top Optics and Sepakat GIS. A total of 250 participants have been attending the symposium including overseas delegates.

Objectives The conference goal is to create awareness and to promote the use of Geoinformation in the area of science & technology with particular focus on the development of the New Economy. The conference is the common platform to learn and exchange ideas on existing Geoinformation methods of utilizing the latest technology available for economic development. The conference demonstrates how Geoinformation can facilitate and assist in information exchange and dissemination in international science & technology development.

13

International Symposium and Exhibition on Goeinformation 2002

1. Delegate

from local and International attending the symposium

2.Deputy Minister of Ministry of Land and Cooperative Development at the symposium

3. Director General (The Chairman of the Lembaga Jurukur Tanah Malaysia) presenting a souvenir to the Vice Canselor of UTM

4. Opening ceremony for the International Symposium and Exhibition on Geoinformation 2002

A HISTORICAL OVERVIEW OF THE FIRST LEVELLING NETWORK IN PENINSULAR MALAYSIA Azhari bin Mohamed Geodesy Section Department of Survey and Mapping Malaysia Kuala Lumpur [email protected]

Abstract

first established at the turn of the last century. This network is referred to as the First Order Levelling Network 1967 (FOLN67). This was followed by a second vertical control network called the Precise Levelling Network (PLN). Table 2.1 presents the chronicle of the main events related to the development of the vertical control networks in Peninsular Malaysia.

Like most countries around the world, a nationwide levelling network in Peninsular Malaysia was first established at the turn of the last century. This network is referred to as the First Order Levelling Network 1967 (FOLN67). This article presents a historical overview of this network and also relates to the need for a new one in view of the advances that have occurred over the past decades. 1.0

Background

3.0

Malaysia still base its geodetic, cadastral, engineering and mapping activities with the production of accurate maps and survey information on networks that were observed during the 1900s. Such systems have been the 2-dimensional triangulation networks and a separate 1-D height networks. The realisation and maintenance of the national reference systems for horizontal and vertical control in Peninsular Malaysia falls under the jurisdiction of the Department of Survey and Mapping Malaysia. Historical documents and records related to the early formation of the first vertical control network in Peninsular Malaysia are lacking and scanty. Furthermore, the available reports contain only limited information about the network. 2.0

The First Levelling Network

The first precise levelling programme in Peninsular Malaysia was initiated in 1912 with the first line levelled between Port Kelang and Kuala Lumpur. However, one of the primary difficulties encountered during the early days of precise levelling was one of access. As seen from the 1944 map of Peninsular Malaysia obtained from the Survey of India as illustrated in Figure 3.1, a number of high grade roads traversed the country but the network was not close and left many large areas without ready access. Roads built were mostly concentrated in the western states. The road system consisted of a main road running the length of the country from Johor to the Thailand border. The other main road ran eastwards from Kuala Lumpur to Kuantan and then northwards to Kota Bharu. On the other hand, the rail system consisted of two main lines: first, Johor to Kuala Lumpur to Butterworth and to the Thailand Border and secondly, Gemas through Pahang and to Kota Bharu. Thus, it was only logical that the levelling lines of the FOLN67 were mainly governed by the road and rail networks.

Historic Synopsis of Vertical Control Networks in Peninsular Malaysia

The establishment of a height network provides a system of vertical control whereby a height value of a point may be defined. Like most countries around the world, a nation-wide levelling network in Peninsular Malaysia was

15

Year

Event

1885

: The first Survey Department was instituitionalised in Johore

1909

: The Federated Malay States Survey Department (FMSSD) was formed

1912

: The mean sea level at Port Kelang was determined through a one-year tidal observation, referred to as the Land Survey Datum 1912 (LSD12) The first precise levelling programme commenced with first line levelled between Port Kelang and Kuala Lumpur

1941

: Levelling operations halted on outbreak of Second World War

1949

: Levelling operations resumed

1963

: The FMSSD was renamed as the Department of survey and Mapping Malaysia (DSMM)

1967

: Precise levelling programme completed with the network known as the First Order Levelling Network 1967

1976

: Survey Regulations 1976 published with Appendix 1B Instructions for Precise Levelling

1979

: Field inspection revealed about 50% of bench marks of the network were missing on damaged A study was conducted to re-define the vertical datum for peninsular Malaysia

1980

: Survey Circular No. 1/1980 Densification on Levelling Network in Peninsular Malaysia was issued

1981

: The determination of mean sea lenel at Port Kelang and later at eleven other tide gauge stations around Peninsular Malaysia began

1984

: Director of Topography Circular No. 1/1984 Survey Instructions for precise Levelling was issued

1985

: The second precise levelling programme commenced, known as Precise Levelling Network PLN)

1988

: The introduction of Motorised Levelling technique

1989

: Director of Topography Circular no. 1/1989 Survey Insructions for the Densification of Levelling Network in Peninsular Malaysia was issued

1993

: The setting-up of a SOKKI Laser Interferometer System LM 7900 for staff calibration at DSMM Introduction of new design for bench mark monuments

1994

: Official launching of the National Geodetic Vertical Datum (NGVD) as the new vertical datum for Peninsular Malaysia based on 13 years of tidal observation at Port Kelang

1999

: Field operation of the PLN completed

2000

: Processing and Adjustment of PLN commenced Table 2.1 The development of vertical control networks in Peninsular Malaysia

16

3.1

slowly, with an additional of several hundred kilometers more. By 1949, the precise levelling network had grown to over 2352 km in distance. The extent of the network as in 1944 is shown in Figure 3.1. The network was finally completed in 1967 and it is known as the First Order Levelling Network 1967 or FOLN67. The configuration of the FOLN67 as illustrated in Figure 3.2 is mainly governed by its road and rail patterns. The reasons for this were obvious: availability, security, easy accessibility and communications as well as less possibility of development and disturbance to the BMs. The final network contains a total of 2872 km of levelling distances with about 87 lines. It also comprises of 11 primary loops with a total of about 2532 bench marks.

Land Survey Datum 1912 (LSD12)

Probably, the earliest attempt to establish a vertical datum in Malaysia was carried out by the British Admiralty in 1908 at Port Swettenham, which is now named Port Klang. However, the MSL was derived from the 5-month of tidal observations and was strictly local in nature. It was intended to fulfill the ‘surroundings’ requirements during the period, and never used as the national vertical datum. The reference BM is no longer in existence today due to the damage caused by heavy development within the vicinity of Port Kelang. It was not until 1912 that the British Admiralty established what to become the first national levelling datum for Peninsular Malaysia. It was based on a one-year tidal observations carried out by the HMS Waterwitch between noon of September 1, 1911 and May 31, 1912, at Port Kelang. This datum continued to serve vertical control users in Peninsular Malaysia for more than 80 years. 3.2

Configuration of the Network

The first precise levelling line in the Peninsular Malaysia was established in 1912. It started at the reference bench mark BM1 at Port Swettenham (now known as Port Kelang) and ended in Kuala Lumpur and followed a railway line. A distance of about 40 km was covered. The levelling survey was considered significant as it represented the connection of the tidal datum to all subsequent BMs across the nation. Two other levelling lines branched out during the same year. BMs were planted along railway lines to Tanjung Malim in the north and to the interstate border of Selangor and Negeri Sembilan in the south, covering a distance of about 40 km and 70 km respectively. The lines later expanded to other states with over 1080 km of levelling accomplished in a sporadic manner. In 1941, the levelling operations were interrupted on the outbreak of the Second World War in Peninsular Malaysia. A period of inactivity followed for almost eight years and when the war was over, the operation resumed. The precise levelling activities continued steadily and at times very

17

Figure 3.1. Map of the extent of the first levelling network in Peninsular Malaysia in 1944

Figure 3.2.

Map of the final configuration of the FOLN67 for Peninsular Malaysia

18

3.3

There are several indications that distortions and errors existed in the FOLN67. DSMM has carried out numerous studies that revealed the weaknesses and anomalies in the network. As was mentioned above, the LSD12 was based on a one-year tidal observation at Port Kelang. Discrepancies were noted in the levelling lines between four other tide gauges and Port Kelang. The results revealed a positive slope of MSL of between 0.3 to 1 feet at Port Kelang. Although the reason for the large differences remains unexplained, the presence of systematic errors and the effects of sea surface topography (SST) were suspected.

Bench Marks

There were two types of bench marks that made up the FOLN67: standard bench marks (SBM) and intermediate bench marks (IBM). These are illustrated in Figures 3.3 and 3.4 respectively. SBM is set flush with the ground. The reference point is a deep rod with a numbered hexagon screw cap and driven vertically into the ground. It is sited in a buried steel-reinforced concrete chamber. The point is protected with 0.5 inches thick removable steel cover that measures 12 inches by 12 inches. An SBM is established at junction points and also in major towns. On the other hand, an IBM consists of a brass plate set in concrete monument above ground. It is placed every 1 km. 3.4

Similar trends have also been observed when the FOLN67 was connected to the levelling network of Thailand in 1966 and 1967. The connection at Sg. Golok and Padang Besar revealed differences of between 0.29 and 0.37 metres below LSD12. Again there was no attempt been made to investigate the cause of the differences, although unreliable tidal records from tide gauge stations in both countries were also suspect. However this could also be an indicative of the presence of distortions in the levelling network.

Comments on First Order Levelling Network 1967

The BMs of the First Order Levelling Network 1967 constitute the official height system for mapping, engineering and other applications during the past century. The levelling operation started during a period where technological computing capabilities that existed in early 1900’s were limited. Even so, the network had served the objectives set out quite well, by providing an adequate base for the national topographic programme and for general engineering purposes in Peninsular Malaysia.

The above studies showed that the network had its inadequacies. The datum was derived from tidal observations for one year and thus did not take into account the 18.6 years nodal moon cycle. The information on the level instruments, levelling data, field and processing procedures were either not well documented or missing. Observations were built up piecemeal and more on the basis of expediency. As such, it was suspected that the levelling lines were surveyed one after another and adjusted to previous values as completed. This piecemeal adjustment process certainly would produce a network of vertical controls having distortions. Thus, the network was heterogeneous and no single adjustment was ever attempted. Furthermore, the network is based on observed heights such that the values were not corrected with respect to the gravity field, hence no orthometric corrections were applied.

The execution of the establishment of FOLN67 had been a piecemeal process. The way that the FOLN67 was established tended to reflect a response to immediate needs and demands during its time. A majority of the levelling lines were concentrated in developed regions while in some areas there are scanty observations. Thus the network was generally incomplete and did not cover adequately in areas to the east and south of Peninsular Malaysia. The lack of background information on the levelling data had complicated in-depth investigation into the network. Information on the equipment used, field procedures, calibration methods, method of adjustment and corrections being applied to the levelling data were also not available.

19

Horizontal type

Figure 3.4. Intermediate Bench Mark

Figure 3.3. Standard Bench Mark

Indeed, poor scale control and large systematic errors are imminent in past levelling networks. They may also exhibit spatial and temporal correlations as well as unmodelled systematic errors that could be greater than the random errors. The FOLN67 network was built sporadically over 50 years had become obsolete and new levellings are in dire need to meet modern demands. Thus, a new vertical control network that can adequately meet the needs in any given part of the country should be established. 4.0

Vertical type

studies. It was also during this period when it became apparent that significant distortions that existed within the First Order Levelling Network 1967 had caused it to no longer able to satisfy many of the demands made out of it. FOLN67 originated in response to the needs of an expanding number of users. It was established using the most appropriate technology available during that time and thus was adequate as such. However the technological advances during the later half of the last century have brought increased requirements for a more accurate vertical control points.

The Need for a New Levelling Network

The readjustment of the FOLN67 network was impractical since the levelling data were not available to the department. The best and perhaps the only possible solution was to relevel the whole network and add new routes. There had been unsuccessful attempts in the past by other countries to combine new levelling with older work. Baker (1974) reported that new levelling tends to absorb corrections from older network, thus requiring an upsetting of the adjustment of the latter.

It was clear that the First Order Levelling Network 1967 was done on an adhoc basis and finally completed after a period of over 50 years, thus giving rise to its overall shortcomings. However, it should be appreciated that the establishment of FOLN67 belongs to its own time domain and therefore is as creditable an effort as any other following it. The late 1970’s period was one of a rapid economic development in Peninsular Malaysia. This, coupled with the increase in land values, had not only prompted the demand for the vertical controls but also created a greater need for height information of higher accuracy. It was during this era that there was a dire need for an accurate geodetic vertical control network to satisfy the needs of surveying, mapping and other scientific

Tectonically, Peninsular Malaysia sits on a relatively stable plate. However, through time, certain places in the country occasionally experience subsidence caused especially by the removal of underground water. Discrepancies thus exist between the affected areas and other stable ones as the heights of BMs would not be related to a common datum. A new adjustment would not rectify the

20

shortcomings of the network and hence due to the above factors, a relevelling exercise was the best option to undertake. The establishment of a new vertical control network also means that destroyed bench marks be replaced and re-determination of height values of the remaining BMs using a more precise levelling technique and a rigorous network adjustment to ensure better accuracy and homogeneity.

began to initiate definitive steps in the early 1970’s to prepare for a new vertical control for Peninsular Malaysia. This was to be known as the Precise Levelling Network (PLN). REFERENCE Mohamed, A.B. (2002) An Investigation of Vertical Control Network in Peninsular Malaysia using GPS and Terrestrial Data. PhD Thesis in Preparation, Universiti Teknologi Malaysia, October, 288pp.

Thus, the combined influence of the factors that contributed to the need of a new vertical control network for Peninsular Malaysia can be summarised as follows: Ý Ý Ý Ý Ý Ý Ý Ý Ý Ý

5.0

ad-hoc nature of the levelling operation, survey methods, instruments and standards of accuracy were not documented, long period of survey to complete the network, low coverage of the bench marks, LSD12 is based on MSL derived from just 1-year tidal observational data, no consistency with levelling networks of neighbouring countries, Thailand and Singapore, no single adjustment of the network ever attempted, no application of orthometric correction, most BMs are either missing or damaged, and increasing requirements for accuracy and consistency by modern users.

Conclusion

The evolution of the first levelling network in Peninsular Malaysia known as the First Order Levelling Network 1967 or FOLN67, the adopted vertical datum and comments on its shortcomings that led to the need to conduct a second levelling are given. Due to the numerous weaknesses inherent in the FOLN67 network, DSMM began to initiate definitive steps in the early 1970’s to prepare for a new vertical control for Peninsular Malaysia. However, it is noted that the establishment of FOLN67 belongs to its own time domain and therefore was as creditable an effort as any other following it. Due to the numerous weaknesses inherent in the network, DSMM

21

Disedia dan dicetak oleh JUPEM No. 119-03