New possibilities of remote sensing, point clouds, laser scanning and UAV LS

Transcription

1 New possibilities of remote sensing, point clouds, laser scanning and UAV LS Juha Hyyppä Director, NLS/FGI, Director, CoE-LaSR Prof., NLS/FGI, Content What is NLS/FGI? General Principles in Laser Scanning (LS) 3D Remote Sensing Comparison Autonomous driving & Information Systems & Decision support systems Mini-UAV LS Personal and Mobile LS Plot level point clouds 2 1

2 FGI Research institute in National Land Survey of Finland (since 1. Jan 2015) FGI was the most productive research environment in Finland (output/input) About 110 Staff (roughly 90 scientist) Budget roughly 10 MEU (mostly outside financed) Highly academic institute (nearly 50% staff has PhD) Publish around 60 ISI Web of Science peer reviewed articles (200 peer reviewed articles) Three joint professorships with universities The only Finnish research institute that coordinates Center of Excellence by Academy of Finland Courtesy to JP Virtanen & J. Hyyppä, Aalto/FGI Contribution to NLS Topographic Database Crowdsourcing 2

3 Pohjoinen (m) 20/11/2015 Key Research Areas 1. Reference Systems National coordinate, height and gravity systems Nummela standard baseline and Metsähovi research station Deformation networks (Nuclear safety) Photogrammetric and laser scanning test fields 2. Changing Earth Land use change Land uplift and deformation Multitemporal remote sensing data analysis Global gravity change safety and environmental aspects Change of land topography Vegetation 3. Mobile Geomatics Mobile mapping GNSS technologies Indoor outdoor navigation smart mobile geospatial solutions Mobile GIS 4. Spatial Data Infrastructures Interoperable, standardized data services Harmonization and quality of geospatial data Network-based processing services Use and usability of GI-data and applications Visualization and cartographic presentation Mobile precise positioning with FinnRef 5 GNSS 4 PPP-RTK Itä (m) Users P 3 service project Intelligent transportation systems FinnRef Positioning accuracy targeting 0.5 m 3

4 Local deformations 3D Geographic Information for Hikers In the MenoMaps - project a mobile map service for the hikers has been developed. Department of Geoinformatics and Cartography 4

5 Forest map Courtesy to J. Oksanen et al. FGI Wintermap Courtesy to J. Oksanen et al. FGI 5

6 Courtesy to J. Oksanen et al. FGI Centre of Excellence in Laser Scanning Research: Together what is otherwise impossible Pulsed timeof-flight laser radar Juha Kostamovaara Univ. Oulu Hardware-driven approach Mobile and ubiquitous Laser Scanning Juha Hyyppä FGI Laser scanning for precision forestry Markus Holopainen Univ. Helsinki Laser scanning for built environment Hannu Hyyppä Aalto Univ. International benchmarking studies Courtesy to J. Hyyppä. FGI 6

7 Development of new sensors and test their feasibility for various applications Courtesy to A. Jaakkola, FGI Solid Potato Oy Your partner in R&D 14 7

8 General Principles in Laser Scanning Fundamentals of multiple discrete return technology 1st (and only) return from ground 1st return from tree top 2nd return from branches 3 rd return from ground Courtesy to Leica 8

9 ALS Output Output: x,y,z, intensity! Courtesy to Hannu Hyyppä, Petri Rönnholm,Aalto Platforms for Laser Scanning ALS Airborne Laser Scanning TLS Terrestrial Laser Scanning MLS Mobile Laser Scanning Courtesy to Hyyppä, Kukko, Kaartinen, Jaakkola FGI 9

10 Characteristics of LS Airborne Laser Scanning Point density pts/m 2 Homogenous point density Elevation accuracy 5-30 cm Planimetric accuracy cm Few hundred m to several km Transfer costs Cost-effective for areas larger than 50km 2 Multitemporal ALS applications studied relatively little (Murakami et al. 1999, Yu et al. 2004) Reference book Shan and Toth (2009), Vosselman and Maas (2009) Airborne mini-uav scanning New research area (idea in Zhao et al., 2006, first demo for mapping Jaakkola et al. 2010) Point density 10-several hundred pts/m 2 Elevation accuracy 5-30 cm Altitude tens of m Usable for research and multitemporal studies Presently commercial applications Reduces CO 2 consumption of research Courtesy to J. Hyyppä, FGI Characteristics of LS Mobile laser scanning Point density in the range of 100 to several thousands pts/m 2 More homogeneous (TLS) point density Point accuracy of few centimeters (egg) when collected with good GNSS coverage Applicable range of few tens of m Viewing angle different from ALS Higher variation in the range data (e.g. round traffic poles are not round in the data) Vast data sets when data collection is done continuously Need for automatic and interactive algorithms Terrestrial Laser Scanning Point density in the range of pts/m 2 at the 10 m distance Distance accuracy of few mm to 1-2 cm Operational scanning range from 1 to several hundred m Feasible for small areas less than few tens of m distance Processing time challenging: image processing technigues applied Small variation in data, e.g. Distance variation low -> surface normal etc can be calculated Reference book: Vosselman and Maas (2009) Courtesy to J. Hyyppä, FGI 10

11 Courtesy to M. Lehtomäki, FGI MLS Point Cloud Processing 1/2 Unorganised 3D Point Cloud Pre- Processing Removal of Ground and Facades Segmentation Point Segments Attribute Extraction Classification of Segments Object Location Prediction Predicted Object locations 21 Courtesy to M. Lehtomäki, FGI MLS Point Cloud Processing 2/2 Point Cloud Ground Removed Segmented/clustered Classified Segments Tree Lamp post Traffic sign Car Pedestrian Undefined 22 11

12 First publication in using LS in map updating Matikainen, L., Hyyppä, J., and H. Hyyppä, Automatic detection of buildings from laser scanner data for map updating. Workshop '3-D reconstruction from airborne laserscanner and InSAR data', Dresden, Germany 8-10 October The International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol. XXXIV, Part3/WG13, Courtesy to L. Matikainen, FGI OK New Not found Classified as tree 3D Remote Sensing Comparison 12

13 DTM, DSM, ndsm Courtesy to St-Onge DSM DEM, DTM CHM, ndsm 25x25 m 13

14 2x2 m Interpreting ALS Point Clouds Courtesy to TU Wien DSM Crown Volume ndsm Tree Stems DTM Near-Surface Vegetation 14

15 Utilizing individual tree information in laser assisted forest inventory Point clouds future of forest resource assessment 3D Remote Sensing comparison Airborne and Spaceborne systems, 90 plots Helicopter LS & Tomoradar (tbd) ALS (2 point densities) ALS Full Waveform (Finnmap) NLS open data 2012 Airborne images ALS data SAR Tandem X interferometry TerraSAR X Radargrammetry Worldview Stereo 15

16 RMSE (%) 20/11/2015 Relative RMSE for forest attribute estimates using area-based approach and random forest based on various RS 3D data and 91 sample plots of 32 m x 32 m size. 35,00 30,00 25,00 20,00 15,00 Mean height (m) Mean DBH (cm) Basal area (m2/ha) Volume (m3/ha) Biomass (Mg/ha) 10,00 5,00 0,00 ALS-900m ALS-2500m ALS-NLS AP WV2 TSX TDX Courtesy to X. Yu, FGI 31 Autonomous Driving Information Systems Decision Support Systems 16

17 Courtesy to A. Jaakkola, FGI 33 Courtesy to A. Jaakkola, FGI 34 17

18 Full process chain merging virtual and physical worlds Zhu, L., Hyyppä, J., Kukko, A., Kaartinen H., and R. Chen, Photorealistic building construction from mobile laser scanning data. Remote Sensing 3(7), Indoor MLS Courtesy to R. Kaijaluoto, FGI 18

19 Courtesy to A. Jaakkola, FGI 37 Forest SLAM based driver support and forest inventory system Partner: Aalto, Finnish Forest Research Institute Companies: Ponsse, Parker-Vansco, UPM- Kymmene, Stora-Enso, Metsäliitto, Metsäteho Forest SLAM Driver selects the trees to be harvested in the HMI display Semiautomatic control of movements of the crane and harvester head in normal easy conditions Driver has more time for perception and decision making The collected map of trees improves the analysis and planning of forest operations when uploaded to GIS Page 38 Courtesy to A. Visala, Aalto 19

20 Mini-UAV Laser Scanning FGI Sensei first mini- UAV LS (2009) NovAtel SPAN-CPT Ibeo Lux/Sick LMS151 AVT Pike F-421C Specim V10H Flir Photon 320 2nd generation Courtesy to A. Jaakkola, FGI 20

21 World s first mini-uav Laser Scanner SENSEI Jaakkola, A., Hyyppä, J., Kukko, A., Yu, X., Kaartinen, M., Lehtomäki, M., and Y. Lin, 2010, A low-cost multi-sensoral mobile mapping system and its feasibility for tree measurements. ISPRS journal of Photogrammetry and Remote Sensing 65 (6), rd gen UAV LS Lighter Higher PRF Better range accuracy Better planimetric accuracy 21

22 Applications and Impact Ecological for measurements (no transfer cost/pollution of large planes) Effective for Change-detection and multitemporal surveys (data also obtained when needed) Simulates ALS, but also for reference data collection Costs are going down New UAV applications Powerline measurements Operational pilot in September 2014 (first in the world) UAV LS 44 22

23 Tree Extraction with SENSEI Method n Horizontal mean Horiz. StDev Height mean Height StDev 100 Manual cm 14 cm -15 cm 30 cm Automatic cm cm 2 cm 34 cm Jaakkola, A., Hyyppä, J., Kukko, A., Yu, X., Kaartinen, M., Lehtomäki, M., and Y. Lin, 2010, A low-cost multi-sensoral mobile mapping system and its feasibility for tree measurements. ISPRS journal of Photogrammetry and Remote Sensing 65 (6), Courtesy to A. Jaakkola, FGI 46 23

24 Courtesy to A. Jaakkola, FGI 47 24

25 Courtesy to A. Jaakkola, FGI World s first mini-uav-ls operational corridor mapping, Sept 2014 Courtesy to A. Jaakkola, FGI 50 25

26 Mobile and Personal Laser Scanning: System and applications Dr. Antero Kukko Centre of Excellence in Laser Scanning Research Finnish Geospatial Research Institute FGI Courtesy to H. Kaartinen, FGI EuroSDR MLS Accuracy evaluation ROAMER June 2009 Intensity Riegl March 2010 Elevation Mitsubishi Sep 2010 RGB Sensei May 2011 Kaartinen, H., Hyyppä, J., Kukko, A., Jaakkola, A. and H. Hyyppä, Benchmarking the Performance of Mobile Laser Scanning Systems Using a Permanent Test Field. Sensors 12 (9), ; 26

27 Elevation error [m] 20/11/2015 MLS Accuracy Evaluation Elevation error vs. distance Courtesy to H. Kaartinen, FGI MLS A MLS A recalibrated MLS B MLS B recomputed MLS C MLS D MLS E Distance from trajectory [m] PLS - Integrated Technology Courtesy to A. Kukko, FGI For collecting 3D data on the move (i.e. mobile mapping) GNSS-IMU Positioning Ultra High Speed Laser Scanner Cross-track Profiling To gain: Mobility and accessibility Efficiency Object complexity 54 27

28 PLS under thick canopy Courtesy to A. Kukko, FGI In GNSS shadowed areas accuracy is reduced due to the loss of signal: RMSE XY: 0.73 m STD XY: 0.82 m RMSE h: 1.10 m 8 targets, n=53, traj. length 3 km Trajectory correction: mean XY 15.9 cm and h 8.1 cm 55 PLS in Forestry Courtesy to A. Kukko & X. Liang, FGI For collecting complete structural data in complex environments Automation in tree inventories at plot and individual tree level Liang et al.,

29 Industrial engineering Courtesy to A. Kukko, FGI As-built inventories for plant layout design Existing structures Pipelines Connections Masses and materials Complex scenes Demand for level of detail Demand for accuracy 57 Frame-based Spectrometry Dr. Antero Kukko Centre of Excellence in Laser Scanning Research Finnish Geospatial Research Institute FGI 29

30 Reflectance 20/11/2015 Courtesy to E. Honkavaara, FGI FGI hyperspectral, photogrammetric UAV measurement system: Summer 2014 setup 59 Output products: hyperspectral image mosaics, DSMs, point clouds, spectra, BRF, σ XYZ, σ ρ Courtesy to E. Honkavaara, FGI Wavelength (nm) 30

31 New Sensor Technology Future of Remote Sensing and GIS Utilizing individual tree information in laser assisted forest inventory 31