JOINT PROCESSING OF UAV IMAGERY AND TERRESTRIAL MOBILE MAPPING SYSTEM DATA FOR VERY HIGH RESOLUTION CITY MODELING A. Gruen1, X. Huang1,2, R. Qin1, T. Du1, W. Fang1, J. Boavida3, A.Oliveira3 1 Singapore-ETH Center, Future Cities Laboratory, CREATE, Singapore - agruen@geod.baug.ethz.ch; 2 LIESMARS, Wuhan University, China - hwangxf@gmail.com 3 Artescan -3D Scanning, Portugal - jboavida@artescan.net 1. Introduction 2. Project and data fusion 3. Conclusions
DARCH 1. Introduction SEC-FCL project UAV over NUS campus Singapore ETH Centre for Global Environmental Sustainability Future Cities Laboratory (Simulation Platform) 2
SEC-FCL project UAV over NUS campus Singapore ETH Centre for Global Environmental Sustainability Future Cities Laboratory (Simulation Platform) Project purposes + Pilot project to refine DARCH data processing algorithms & sw + Test-bed for demonstrations of technology and products + Results will be applied by a variety of different users for analysis, animation and simulations (autonomous vehicle driving, hydrology, crowd movement, etc.) 3
Challenges/difficulties in UAV urban data acquisition Permission application and flight restrictions Radio interferences Limited take-off/landing spaces Short flying times NUS Campus: Steep terrain, high buildings, tropical vegetation
Challenges/difficulties in MMS urban data acquisition GNSS obstructions by buildings and trees deteriorates accuracy Limitation to streets, doesn t allow all buildings façades acquisition Traffic produces façade and street level obstructions and data noise
2. Project and data fusion Multi-sensor data (1) Vertical aerial UAV images at 5 cm footprint (2) Oblique UAV images (in planning) (3) Raw point clouds from MMS at 5cm point spacing (4) Terrestrial images from off-the-shelf cameras (5) Ground Control Points (GCPs) (6) Existing data (maps) Output 3D hybrid site model, achieved by integration of these input data 6
Mission Parameters Mapping area: 2.2 Flying height: 150 m above the ground Number of images per flight: 25 on average Number of flights: 43 Strip overlap: 80% Across-strip overlap: 60% Number of images for processing: 857 GSD: 5 cm (ground resolution) Expected accuracy: 5-10 cm horizontal and vertical
Actual Blockstructure 43 sub-blocks 8
Main Process Workflow The working procedure can be generalized into several steps, including: (a) UAV images aerial triangulation DARCH (b) Integration of UAV- derived control point data to geo-reference and adjust the MMS point cloud data (c) Modeling of the roof landscape from UAV images (d) Measurement of the DTM from UAV images (e) 3D modeling of façades from MMS data and (if needed) from terrestrial images (f) Modeling of DTM from MMS data (g) Fusing façade and roof models and the DTMs to generate a complete geometry model (h) Optional: Texture mapping from aerial and terrestrial images
Main Process Workflow - Data processing steps DARCH
Object measurement & modeling strategy Buildings/roof-landscape: Cyber City Modeler: semi-automatic procedure Facades: 3ds Max, manual modeling from point cloud DTM- manual measurement: Profiles + break-lines, combining with mobile LiDAR data for area under plant canopy and contours from older maps Vegetation: Parametric measurements: one point on tree top, tree diameter. Use of plant pre-defined models. Light poles (>900): Similar approach Texture: Self-developed software for roof texturing.
DARCH NUS Campus, point cloud from image matcher
DARCH NUS Campus: Overview and resolution
NUS Campus, 3D models from CC-Modeler DARCH
NUS buildings roof texture map DARCH
NUS Campus: DTM integration DARCH
DARCH NUS Campus: Texture mapping
NUS campus Mobile Mapping RIEGL WMX-250 with two RIEGL VQ-250 laser scanners 600 Hz 16 km road, 34.4 GB raw point cloud data, 700 points/m2 on road, 3 hours 5.25 GB of video sequences
DARCH Mobile Mapping: NUS campus point cloud data
DARCH Control point selection 169 points measured manually in UAV stereos precision < 1pi
Accuracy evaluation X-diff Y-diff Z-diff Res Z-diff 1-0.02-0.01 9.55 0.00 2 0.04 0.04 9.57 0.02 3-0.1 0.11 9.41-0.14 4 0.17 0.11 9.76 0.21 5 0.01-0.03 9.52-0.03 6-0.01 0.03 9.36-0.19 7-0.02-0.02 9.51-0.04 8-0.04-0.16 9.43 0.12 9-0.22-0.1 9.28-0.27 10 0.12 0.1 9.55 0.00 11-0.09 0.07 9.62 0.07 12 0.06 0.06 9.67 0.12 13-0.05 0.14 9.61 0.06 14 0 0.03 9.43-0.11 15 0.32 0.17 10.16 0.61 16-0.06 0.1 9.35-0.20 MEAN 0.00687 0.04 9.548 0.0 STD 0.122 0.087 0.20 0.20
DARCH NUS (CREATE) building reconstruction raw data UAV images MMS laser scanning
NUS building reconstruction UAV images and MMS laser scanning (3ds Max) roof from UAV images Wrapped point cloud Complete Model
NUS (CREATE) building reconstruction UAV images and MMS laser scanning DARCH
Conclusions This pilot project delivers valuable experiences for future applications and research topics: (1)The accuracy of multiple data fusion is carefully evaluated; (2) the experiment shows that using the UAV data to georeference MMS data affected by GNSS obstructions is feasible and successful and can save a lot of surveying field work; (3) complete building models are created by integration of UAV data, terrestrial images and MMS data. 25
Conclusions Directions of DARCH future work: (1)how to improve the level of automation in georeferencing of MMS data? (2) how to improve the accuracy of data registration by utilizing features from both data sources? (3) how to do the micro-adjustment to achieve a perfect match between building roofs with the façade model? (4) how to improve and speed up the façade modeling procedure? 26
Thank you! 27