enhancement of the navigation data quality · pdf filedata quality of an autonomous flying...

© 2012 – Bochum University of Applied Sciences

ENHANCEMENT OF THE NAVIGATIONDATA QUALITY OF AN AUTONOMOUSFLYING UNMANNED AERIAL VEHICLEFOR USE IN PHOTOGRAMMETRY

Heinz-Jürgen PrzybillaManfred Bäumker, Alexander Zurhorst

© 2012 – Bochum University of Applied Sciences

Content

§ Introduction

§ Technical aspects of the Mikrokopter system

§ Examination of positioning sensors– Actual status of positioning quality using low-cost GPS

sensors

– Enhancements of positioning quality

§ Photogrammetric applications

§ Conclusions

© 2012 – Bochum University of Applied Sciences

Introduction

§ Nowadays Unmanned Aerial Vehicles/Systems(UAV/UAS) are beyond the stage of testing.

§ Since 30 years they have been used for a widespectrum of applications.

© 2012 – Bochum University of Applied Sciences

History…1979/80

SchlüterPhotohelicopter

Pilot andNavigator

© 2012 – Bochum University of Applied Sciences

History…1979/80

First Tests

Hamburg ISP-Congress 1980

© 2012 – Bochum University of Applied Sciences

History…1979/80

Above: Archeologicaldocumentation, CologneLeft: North Sea tideland

© 2012 – Bochum University of Applied Sciences

Introduction

§ Reasons for this fact are the enormousdevelopments in consumer-electronics needed forthe operating of a copter, available in conjunctionwith low fees.

§ The „MikroKopter“ system is developed under theassistance of an internet community.

§ All electronic components used are standardproducts, fixed together to an efficient and low-prize UAV.

© 2012 – Bochum University of Applied Sciences

Introduction

§ The main focus for using the copter at HS BO arephotogrammetric applications.

§ These requirements derive the enforcement of highquality images and precise nagivation.

§ Actually a typical GPS-sensor (used in low-costsystems) achieves a differential GPS quality of• 1-2m in horizontal and

• 3-5m in vertical direction

© 2012 – Bochum University of Applied Sciences

Introduction

Aims of the investigation:

§ Enhancing GPS-navigation by implementing arealtime kinematic GPS solution.

§ Testing of alternative photogrammetric techniquesfor image orientation and processing.

© 2012 – Bochum University of Applied Sciences

The Mikrokopter Project

§ Completely documented under the Web Sitewww.mikrokopter.de .

§ A matter of do-it-yourself project.

§ Actually three MikroKopter systems are availablefor the investigations:– an Oktokopter of the HS BO and

– a Hexakopter and a Hexakopter XL of the project partnerAerometrics (www.aerometrics.de).

© 2012 – Bochum University of Applied Sciences

The Mikrokopter Project

Parameter Value

Number ofrotors:

4 – 12

Actual load: 250 g – 1000 g

Weight: 650 g – 1700 g

Flying time: 7 – 12 min

Distance: Visual range

Flying height: Max. 350 m(technically reliable)

Power supply: Lipo 11,1 V – 14,8 V

Sensors: Gyroscopes,accelerometers,compass, GPS,barometricaltimeter

© 2012 – Bochum University of Applied Sciences

Sensors

GPS-Module

Navi-Control

Flight-Control

Telemetry-ModuleBrushless-Controller

© 2012 – Bochum University of Applied Sciences

Camera (Ricoh GXR)

© 2012 – Bochum University of Applied Sciences

Aerial flight with operating altitudeof 100m

Camera:Ricoh GXR,f=18mm)

© 2012 – Bochum University of Applied Sciences

Aerial flights with operating altitudeof 50m

Camera:Ricoh GXR,f=18mm)

© 2012 – Bochum University of Applied Sciences

Oblique image of an electric tower

Camera:Ricoh GXR,f=18mm)

© 2012 – Bochum University of Applied Sciences

Examinationof Positioning Sensors§ u-blox GPS receiver: LEA-6S/6T

(horizontal positioning information)

§ 50 channels (GPS-L1/CA-Code, GALILEO OS).

§ SBAS-option (satellite based augmentation system) toacquire the correction signals of WAAS, EGNOS orMSAS.

§ At the Bochum location signals of the two EGNOS(European Geostationary Navigation Overlay System)satellites Inmarsat AOR-E and IOR-W are available andused to calculate a differential GPS position.

© 2012 – Bochum University of Applied Sciences

Examinationof Positioning Sensors

u-blox receivers as part of the GPS circuit board. Left:board surrounded by a protection shield to enhance GPS signal

© 2012 – Bochum University of Applied Sciences

GPS Capability Approval

To check the performance of the GPS-sensor diversetest were carried out:§ Long term measurements at a reference station

(static)§ Short term measurements at known ground control

points (static)§ Short term measurements during flight over a

known ground control point (dynamic)§ Aerial flights with operating altitude of 50m and

100m (dynamic)

© 2012 – Bochum University of Applied Sciences

Short term measurements duringflight over a known GCP (dynamic*)

* Copter-Tracking with tachymeter Trimble S6

© 2012 – Bochum University of Applied Sciences

Aerial flight with operating altitudeof 50m (dynamic)

© 2012 – Bochum University of Applied Sciences

§ The quality of positioning varies with a meandeviation from 3 – 4 m, not considering someoutliers of more than 10 m.

§ As a consequence of these effects the plannedoverlapping of adjacent images often exceeds atolerable limit.

Actual status of positioning quality

© 2012 – Bochum University of Applied Sciences

Actual status of positioning quality

Footprints of an aerial flightshowing the overlapping of adjacent images

© 2012 – Bochum University of Applied Sciences

§ Solution: integration of high performance GPSpositioning using real-time kinematic (RTK)technologies.

§ Requirements:- Availability of an appropriate GPS-receiver and of

- an applicabale software system

§ These requisitions could be met.

Enhancements of positioning quality

© 2012 – Bochum University of Applied Sciences

§ Equipment used for presented test:– One-frequency receiver of u-blox series (LEA-6T) in

combination with an own reference station

– RTK-calculations performed with RTKLIB, an open sourceprogram package for GNSS positioning

§ Under way are tests with a two-frequency receiver(however it is much more expensive, with a priceratio of 140€ : 2500€)

Enhancements of positioning quality

© 2012 – Bochum University of Applied Sciences

§ Distributed free of charge under GNU GPL v3-license.

§ It supports standard and precise positioningalgorithms with GPS, GLONASS, SBAS, GALILEO(which is enabled but not supported in currentversion) and QZSS (Japan).

§ Various positioning modes with GNSS for both real-time and post-processing (Single-point, DGPS/DGNSS, Kinematic, Static, Moving-baseline etc.).

RTKLIB Features

© 2012 – Bochum University of Applied Sciences

§ A wide range of standard formats and protocols forGNSS can be processed, like RINEXi, RTCMx, as wellas the proprietary messages of several GNSSreceivers.

§ The external communication interface supportsserial, TCP/IP, NTRIP and FTP/HTTP protocols.

§ All in all RTKLIB is a high-tech solution andpredestinated for a low-cost UAV-system.

RTKLIB Features

© 2012 – Bochum University of Applied Sciences

§ The u-blox raw-data are sent to the ground controlstation via telemetry.

§ Realtime calculations of the RTK-solutions arecarried out with RTKLIB.

§ (Raw-data of the UAV u-blox as well as those of thereference station are stored for additional post-processing.)

§ RTK-positions are transferred back to the UAVflight-control via telemetry.

Test scenario“Autonomous RTK-Flight”

© 2012 – Bochum University of Applied Sciences

Enhancements of positioning quality

Reference station with JAVAD Triumph-1 G3T (left).RTKLIB screenshot (right)

© 2012 – Bochum University of Applied Sciences

Enhancements of positioning quality

Accuracy of realtime calculationswith RTKLIB while positioning theu-blox sensor at the test field

red:single GPS-solution: 5 m – 20 morange:float-solution: 0.1 m – 1.0 mgreen:fixed-solution: 0.02 m – 0.1 m(first test: 90% of time)

© 2012 – Bochum University of Applied Sciences

§ With the exception of the initiation phase (lasting30 seconds) the RTK processing software calculateseither a fixed solution or a float solution.

§ Instead of an own references station it is possibleto use the data of a foreign reference station or areference service like the German SAPOS HEPS-Service.

§ Suitable tests will be carried out in the next weeks.

Enhancements of positioning quality

© 2012 – Bochum University of Applied Sciences

§ Aerial photogrammetry represents a typical applicationfor UAS (e.g. maps, orthoimages).

§ More and more the issue „pointcloud“ is of interest.§ An actual trend is set by the computer vision

community with (open source) products which process– feature extraction,– image orientation,– generation of pointclouds

fully automatically.

Photogrammetric Applications

© 2012 – Bochum University of Applied Sciences

§ Typical representatives of these solutions are:– Photosynth (Microsoft)

– 123D Catch (Autodesk)

– PhotoScan (Agisoft)

– Bundler/PMVS2 (Snavely)

§ Calculations are done via web-services or on a localhost.

Photogrammetric Applications

© 2012 – Bochum University of Applied Sciences

§ Bundler – a so called “structure-from-motion (SfM)system” operating on (large) unordered imagecollections.

§ PMVS2 – Patch-based Multi-view Stereo Software.

§ Bundler is responsible for the image orientationand produces sparse point clouds.

§ PMVS2 calculates denser point clouds (on the basisof Bundler results).

Bundler / PMVS2

© 2012 – Bochum University of Applied Sciences

Bundler / PMVS2 – Example 1

§ Orientation of 74 oblique images

§ Result: point cloud with a total of 1.488.381 points

© 2012 – Bochum University of Applied Sciences

Bundler / PMVS2 – Example 1

© 2012 – Bochum University of Applied Sciences

Bundler / PMVS2 – Example 1

© 2012 – Bochum University of Applied Sciences

Bundler / PMVS2 – Example 1

190.462points

Constructionof planes

© 2012 – Bochum University of Applied Sciences

Bundler / PMVS2 – Example 2

§ Orientation of 66 vertical aerial images

§ Result: point cloud with a total of 1.527.078 points

© 2012 – Bochum University of Applied Sciences

Conclusions

§ Navigation accuracy is a critical parameter for asuccessful operation of UAVs.

§ The actual DGPS accuracies have to be improved.

§ RTK-GPS is a usable technology but requires furthertechnical developments.

§ A partial direct georeferencing for UAV sensors(concerning their position in space) is realisable.

© 2012 – Bochum University of Applied Sciences

Conclusions

§ Aerial flight with UAVs – in context with classicalimage overlapping – suffers from insufficientnavigation quality.

§ As a result the overlapping ratio has to beincreased considerably.

§ Recently available software products for theorientation of „large unordered image collections“show encouraging results in generating spatialpoint clouds with „pleasant“ geometric quality.