In order to manage and display large amount of urban as-built data, the information system is made by following parts.
1. 3D database: in SQL language with keyword search available. Considering data amount and display speed, scan region is divided the same as map coordinates. Individual scans in different periods are registered and divided into overlapping 3D blocks of point clouds and models. Whole regional model is used as a “3D map” to show relative location among buildings.
2. Web pages with measuring function: the construction and maintenance of urban models. In order to provide an interactive browsing webpage, large files of point clouds are in OCT format for display efficiency. The point models are in full scale, so the distance between any two points can be measured in a browser mode. In reality, the measuring function reduces potential field trip for dimensioning. The models which are browsed with real color and dimensions, combines as-built data and virtual reality interface.
3. Editing: In order to combine individual scans for regional display, the registration and hole-filling of 3D data have to be conducted. On-site scans of building facades are usually full of holes due to the interference of pedestrians, vehicles, and plants. The holes are filled using the scans made from complementary directions or using software by referring to the curvature around holes.
4. Data types: in 3D point clouds and polygon models. The former is the original data to be maintained and will be referred by consequent scans or studies. Another version of original point clouds also have to go through “noise reduction” and “sampling” process for manipulation efficiency. While the file size is usually very large, the number of points and polygons has to be reduced with appearance altered more in structural details and less in texture details. The creation of vector drawings is not included in this study.
5. Service: The scan data, with digital copyright, will be placed on internet and open to public with restriction to
Subjects
Scan of single historical building and neighborhood environment has been conducted in former studies (Fig. 2). In order to increase the feasibility of as-built data in a broader perspective, this project enlarges scan area and starts the
“digital landscape” scans of local main streets as the first step to build a scan information management system.
Fig. 2. Plafond and the entrance of Long-Shan temple (top) and the Taipei Confucius temple (bottom)
Buildings
In addition to record a component’s 3D configuration, scan data are also used to analyze embedded composition. A very special study was made to compare two plafonds designed and engineered by the same person based on as-built geometries (Fig. 2). Although the two plafonds were built under different spatial constraints, the comparison tries to find if the two plafonds shared the same design principles. Plafonds are made by complex wood bracket system as part of ceiling decoration. With the original design data missing and the joints are hidden after being assembled, traditional plafond study emphasizes analogical description of component layout. The as-built geometries are difficult to measure and to co-relate in 3D. Under this restriction, co-relating two plafonds engineered by the same person or different ones is even more challenging.
The reference-line model is similar a network (Fig. 3) which is made by the intersections of components as nodes and the components as links. The model is created after all the point clouds are registered. Three kinds of procedures were applied in the overlapping comparison: point clouds overlapping, one-dimensional scaling, and details comparison. The smaller plafond was enlarged horizontally (Fig. 4). Both plafonds share the same depth, but the Confucius’s has opening 1.205 wider than the Long-Shan’s due to the space available above and the manner of integration with wood main frame. The reference-line models of the both temples are close-related under the similar rules of composition and vertical segregation.
Buildings and components were usually scanned on the same period using different types of scanners (Fig. 5).
Building scans help define the relative location between components, in the mean time, to create as-built 3D computer model. Chinese temple for digital preservation is the first exemplification (Fig. 6). The model has to combine temple exteriors and interiors to show the double-eave mountain-styled roof structure on both sides. The
entire temple and its components have to be browsed on Internet up to the size of 200,000 points. The database also has to meet the specification of metadata. The preservation presents a reversed model construction process to integrate fragmental temple parts used to be retrieved individually.
The second case goes to old Japanese houses which are located in downtown area for over 50 years (Fig. 7). As the urban renewal has picked up pace rapidly in neighborhood, house owners intended to sell properties for apartment or commercial use. The environment impact caused by increased density forced original residents to reconsider the potential influence of these old houses. A new development scheme was proposed to maintain houses with small open spaces and lanes reorganized into an “urban oasis.” The environment survey, which includes the scan of remaining old houses and neighborhood apartments, becomes very important in defining the solid and void volume in this region. The scan records included a pine tree which is classified as endangered species.
Fig. 3. Wireframe model of Long-Shan temple (left) and Confucius temple (right)
Fig. 4. Plafond appearance and two overlapping slices of clouds
Fig. 5. Temple interior (left), detail scan (middle), and close-range scanner (right)
Fig. 6. Temple section and bell tower
Fig. 7. Old Japanese house in polygon representation
Historical district
A historical district in Northern Taiwan was scanned for its temple, residential houses, and public buildings (Fig. 8).
The oldest house was built in 1832. The historical buildings are all allocated within a street block. This peaceful
town is full of tourists during holidays. The regional scans show old time commercial district and the addition of new private houses over decades.
Fig. 8. Regional scans of a historical district
Fig. 9. The auditorium and a typical local street view presented in 3D scanworld
Fig. 10. Scans and design alternatives
Communities
The Japanese houses along with others temples and historical buildings are allocated within 10 km diameter
downtown area. Just a few blocks south, an old campus auditorium has gone through three renovations (Fig. 9) since Japanese period of Taiwan. To have these buildings geographically connected, street scenes were scanned in between.
Tamsui, as the second example, was an international gateway on northern Taiwan in early days. Surround by mountains, Taiwan Strait is located to the west. As the last terminal of Rapid Transit System from Taipei, Tamsui has attracted many tourists for its famous local historical buildings.
The urban scene then joined design education through RP-aided design studios. A project was proposed near the mountain side of the satellite town which is full of tourists during weekends. The studio project has to incorporate the mountain view and river-side scene as design development criteria. RP-aided architectural design is a design pedagogy that verifies digital architecture using the physical models for environmental evaluation. Various issues were explored, such as the inter-relationships between images and forms, virtual layout and substantiation, and design process in general. RP was used to verify the possible forms (Fig. 10) in a scaled footprint that shed light on the design development.
Regional cultural and commercial characteristics help defining scan issues as follows.
1. the old street appearance: The scans are mainly made to the static contents of the environment. Tamsui can provide visitors and tourists more sightseeing information by showing street facades with 3D scans of shops with signs. Instead of illustrating 2D images in tourist guide, the scans are presented as the most updated 3D as-built models so a visitor can identify a shop with the least deviation from what a person can see in reality.
2. the historical identity of different periods: Old buildings show the influence made by early immigrants from Europe, Japan, and mainland China styles. Building styles and decorations can be recorded by 3D scans.
3. the urban development clues: Tamsui is developed based on its distinguished river and mountain views. Old streets run parallel to the river. The geographic scan of land form will show the development trails of past decades.
Scan also shows street as well as the movement of vehicles and pedestrians. The sidewalk part of scan, which is a very common urban scene, illustrates the entangling pedestrians and the actual usage of motorcycle parking space nearby.
On-going urban scans
Connecting previous spots through the scans of urban circulation system is the ongoing works. The scan schedule is planned based on two principles: to expand former scan scope of individual buildings to regions and to record the urban development process. Previous scan-related works present six years efforts which will be registered with new scans to enlarge the covered area. New scans are also carefully planned to follow the trails of undergoing urban construction as part of chronological records for past-present comparison. For now three main traffic axes running north-south and east-west are selected for a scan period of 1-1.5 years each. A satellite town is selected after that for the record of country scenes with more local culture-oriented description.
The scan would include 3D circulation system of streets, sidewalks, Taipei Rapid Transit System (TRTS), and skywalks. Buildings and streets will also be scanned before, under, and after construction as a chronological urban development record available for the working schedule monitoring of as-built construction process.
Discussions
As the result, former related studies have extended the record of 3D geometries into following information related
1. the definition of digital contents: Though the scans along traffic axis, the most important urban appearances, behaviors, plants can be retrieved at the same time and be included in the same point cloud models.
2. the integration of land registration information: Combining former vector drawings of survey and scan data provides up-to-date urban information in 2D and 3D more detailed than current 2D registration information does.
3. the illustration of chronological events: The urban activities have been chorological influence of regions over years. Overlapping periodical scans of the same places can illustrate the changes explicitly.
4. the integration of fragmental urban elements: Multiple urban elements which are allocated beyond human scale can be integrated in 3D for a clear inspection of inter-relationship.
5. the characteristics abstraction: Scans of facades and skylines help define the urban characteristics and icons.
6. the inclusion of “noise”: Traditional scan interference made by pedestrians and vehicles are considered as a record of human-environment interaction. The interferences, which are usually called “noise”, will be maintained purposely.
Scanners
This study uses Cyclone HDS 2500, HDS 3000, and Optix 400L scanners for long- and short-range scans (Fig. 11).
HDS 3000 can reach 300 m with 360x270 degree of scan. Optix 400L is feasible for scan depth of 40-100 cm under 30 micron deviation. Both HDS 3000 and Optix 400L can retrieve color point clouds and polygons. Data can either be wrapped into polygons afterward (HDS 2500 and 3000) or output meshes directly (Optix 400L).
Fig. 11. Cyrax HDS 2500, 3000, and OptiX400
Limitations
3D scan technology still bears certain limitations:
1. The interference of obscures: Obstruction to laser beams makes the measurements behind objects impossible. A smaller hand-held scanner would be helpful for a smaller space like a crevice. This problem can be solved by registering scan data made from other directions.
2. The cost: The initial cost of photogrammetry software is relatively low. The software with educational discount is usually under USD1,000, much less than the USD 100,000 long-range laser scanner used in this study.
3. The weight and size of a long-range 3D scanner: The 3D scanner (2500) of this study weighs about 27.8 kg and measures the dimension of 15.8" x 13.25" x 16.9" (scanner) and 12.25" x 11" x 9.375" (power supply). The weight and size are not designed for an easy lift to any place without elevators.
Conclusion
3D scanners were used for the digital preservation of building components, buildings, village, and urban street scenes.
Scan data, in terms of point clouds, were retrieved to represent the as-built geometric information. The digital models
construction process. While this scope of study is extended from a temple to street blocks, individual building are included as part of the urban scenes to provide a more precise description of as-built environments for building design and planning control.
The RP output and the VR models contribute to one of the most accurate data sources for environmental evaluation.
The RP approach, which also offers a good opportunity for practice with media tools, liberated the dependency on particular design tools, fostering a better interaction with the environment context.
acknowledgements
Part of this study was sponsored by National Science Council under the project number NSC 96-2221-E-011 -156.
REFERENCES
1. Dick, A.R., P.H.S. Torr and R.Cipolla, 2004: “Modelling and Interpretation of Architecture from Several Images”, Intl J. of Computer Vision, Vol.60, nº 2, 111-134.
2. Fernández-Martin J.J., SanJosé J.I., Martínez J., and Finat J.: Multirresolution Surveying of complex façades: a Comparative analysis between digital photogrammetry and 3d laser scanning, CIPA Symposium, Torino, 2005.
3. Finat J., Ordering criteria and information fusion in 3D laser surveying of small urban spaces, University of Valladolid, 47011 Valladolid, Spain, 2005
4. Früh, C. and Zakhor, A, 2004. “An automated method for large-scale, ground-based city model acquisition”. Intl J. of Computer Vision, 60(1), 5-20.
5. Hartley R. and Zisserman A. Multiple view geometry in Computer Vision. University of Cambridge, UK, 2000.
6. Ikeuchi, K., M.Sakauchi, H.Kawasaki, I.Sato: “Constructing Virtual Cities by using Panoramic Images”, Intl J.
of Computer Vision, Vol.658 nº 3, 237-247, 2004.
7. Martínez, J.; J.Finat, L.M.Fuentes, M.Gonzalo, A.Viloria: A coarse-to-fine curved approach to 3d surveying of ornamental aspects and sculptures in façades, CIPA Symposium, Torino, 2005
8. SanJosé J.I., Finat J., Fernández-Martin J.J., Martínez J., M.Fuentes L., Gonzalo M.: Urban lasermetry.
Problems and results for Surveying urban historical centres: Some pilot cases of spanish Plaza Mayor, CIPA Symposium, Torino, 2005.
9. Slabaugh,G.G., W.B-Culbertson, T.Malzbender, M.R.Stevens and R.W.Schafer: “Methods for volumetric reconstructions of Visual Scenes”, Intl.J. of Computer Vision, Vol.57, nº 3, 179-200, 2004.
10. Teller, S., M.Antone, Z.Bodnar, M.Bosse, S.Coorg. M.Jetwa and N.Master: “Calibrated, Registered Images of an Extended Urban Area”, Intl. J. of Computer Vision, Vol.53, nº 1, 93-107, 2003.
11. Triggs, B., P.F.McLauchlan, R.I.Hartley and A.W.Fitzgibbon: Bundle Adjustement: A modern synthesis, in B.Triggs, A.Zissermand and R.Szeliski, eds: “Vision Algorithms: Theory and Practice (Corfu, 1999)”, LNCS 1883, Springer-Verlag, 1999, 298-372.
出席國際學術會議心得報告
會議名稱 SPIE's 2007 International Symposium on Remote Sensing
發表論文題目 The construction of as-built digital urbanscape - from building components to a city
一、參加會議經過
本次Remote Sensing 2007研討會在義大利Florence市舉行，由the International Society for Optical Engineering (SPIE)主辦，地點在Palazzo degli Affari Conference Centre，時間為9/17-20，
共為期四天。本人以國科會補助專題研究-「數位都市景觀現況紀錄-從建築構件至城市」(NSC 96-2221-E-011-156) 內 容 「 The construction of as-built digital urbanscape - from building components to a city」一文發表，而本次參加經費亦由專題研究中經費補助。
本研討會事實上是結合 Remote Sensing 及 Security+Defense 二會，二者均參加要加不少錢，
但現場似乎無嚴格把關，所以時間允許也可聽到不同領域之演講。而少見的是註冊費竟然要 加 20%稅，令人無法相信。以下是相關議題：
6743: Remote Sensing of the Ocean, Sea Ice, and Large Water Regions 2007
6750: Lidar Technologies, Techniques, and Measurements for Atmospheric Remote Sensing 6745: Remote Sensing of Clouds and the Atmosphere
6744A: Sensors, Systems, and Next-generation Satellites
6747: Optics in Atmospheric Propagation and Adaptive Systems
6749: Remote Sensing for Environmental Monitoring, GIS Applications, and Geology 6746: SAR Image Analysis, Modeling, and Techniques
6748: Image and Signal Processing for Remote Sensing
6742: Remote Sensing for Agriculture, Ecosystems, and Hydrology
6744B: Global Earth Observing System of Systems, Implementation Strategies and Applications 本研討會計有歐、亞、非、美洲國家參加，論文發表者計 65 篇文章，另有 57 篇海報。研討 會分成下列子項進行：
Session 1: Remote Sensing and Archaeology I
Session 3: Environmental Monitoring I Session 4: Environmental Monitoring II Session 5: Environmental Monitoring III Session 6: Processing Methodologies I Session 7: Urban Remote Sensing I Posters—Tuesday
Session 8: Urban Remote Sensing II Session 9: Urban Remote Sensing III Session 10: Processing Methodologies II Session 11: Processing Methodologies III Session 12: Sensors and Platforms I Session 13: Sensors and Platforms II
Session 14: Geology and Hazard Monitoring Session 15: Volcanic Monitoring
每日議程均由 keynote/invited speech 開場，同一時間在不同地區分成數場舉辦，因此常因興 趣、專長必須交錯選擇聽取。本人在第三日 Session 8: Urban Remote Sensing II 中發表。以下 就有趣、先進之發表介紹：
在 session 1: remote sensing and archaeology I 中，Techniques for improving the detection of archaeological features from satellite imagery 一文介紹使用 5-8m 解析度的 Corona / Ikonos imagery 及用 linear 和 enhancement 方式於 edge detection，contrast type 有 direct 及 proxy measurement 二種，最高 image resolution 可小於 1m。目的在於瞭解 local condition 及 contrast different usages over time。同 session 中 Beck 有二篇文章，竟然是預先錄製現場播放，第二篇 介紹在印度 Sanchi 基地古蹟保存工作，目的有二：prospection、time change。另介紹現有衛 星影像資料庫多種，各有不同解析度，連帶影響判讀難易、精確度，如 low cost images 之 Labdsat(15-30m)、Aster、Spot(4-15m)、Corona (3m)、SRTM 均是。高解析度影像有 Quickbird 及 Corona (0.6-2m)，所以也做了一資源在當地適用性比較。
在 unmanned/unattended sensors and sensor networks IV 中
Carapezza 是 invited speaker，該文介紹未來科技，這是我聽過最有趣課題之一，有點像是科 幻電影情節，或是 Discovery 頻道介紹的未來科技。內容是運用動物本能發展人用科技，以戰 爭為對象，可見本研討會領域非常寬廣。實例如下：
1. 步兵荷重 100lb vs. 40lb：希望可由 100lb 減至 40lb。這已經操越材料科技可解決問題範 疇，而必須更進一步仰賴網路、偵測科技協同運用方可從本質上達成。
1. 步兵荷重 100lb vs. 40lb：希望可由 100lb 減至 40lb。這已經操越材料科技可解決問題範 疇，而必須更進一步仰賴網路、偵測科技協同運用方可從本質上達成。