-
DRAF
T
AgriDrone - A Ubicomp Prototype for Precision Farming
Med MatovuIT University of Copenhagen
[email protected]
Viktoras SvolkaIT University of Copenhagen
[email protected]
Paul HenckelIT University of Copenhagen
[email protected]
ABSTRACTPrecision agriculture aims at providing ICT tools that
allowfarmers to micromanage and make detailed plans for the workto
be done. Solutions have existed since the 1980s but havenot found
widespread adoption. With the rapid developmentof low-cost
open-source drones and advances in camera tech-nology the last
couple of years, we propose a system formonitoring crop development
by taking pictures and planningagricultural activities using drones
that can be controlled be-yond visual range using a standard web
interface. The systemwas evaluated in terms of its perceived
usability and a com-parative measure on the time it takes to cover
an area by footvs using drone captures. We found usability of our
system tobe high and calculated the coverage time of the drone to
be 3times faster than manual effort. This will allow the farmers
toschedule activities in a more informed way, saving time firstand
resources.
Author Keywordsdrones, aerial photography, precision
agriculture, precisionfarming, crop specific site survey, UAV,
decision supportsystem
INTRODUCTIONUsing drones or unmanned aerial vehicles (UAVs) in
agricul-ture initially started back in the early 1980s for crop
dust-ing[?]. Since then the UAVs have proliferated in
applicationsof aerial photography to imaging of crop fields to
assist withcrop production management [6, 1]. The agriculture
indus-try has grown and is seeking advanced ways to control
andmonitor crops, thus the term Precision
Agriculture/Farming(PA/PF) has emerged as a farming management
strategy. PAwas initially used as synonym for automatic steering
systemsin the early 1990s. Current PA applications collects and
pro-cess data from multiple sources for improving the
understand-ing and management of soil and landscape resources in
orderto handle crops in a more efficient way.
Even though solutions for PA exists today, widespread adop-tion
has not been catching on. There can be numerous reasonsfor this
including that PA is not equally suited for all types ofcrop [2],
but also that the systems available are too complexor more advanced
than needs to be for the utility of the farm-ers in their day to
day work.
We propose a low cost, low entry barrier system for mon-itoring
crop development and planning agricultural activitieswhich can be
used as part of a larger PA framework. This willbe achieved by
using drones to take pictures of farm parcels.
The problem of proficiently monitoring crop development inmodern
industrial agriculture increases with the size of thefarmland. For
this reason we established a partnership withRyegaard og Trudsholm
Godser1 which operate an area of1000ha with a yield of 850ha per
year. Their main crop iswheat and they are already by their own
needs, trying to de-velop ways to detect and respond to crop
threats in more effi-cient ways.
For planning and crop-development purposes we think dronescould
be useful within organic agriculture and permacultureas well.
RELATED WORKThe usage of UAV is not new, it have been in
production sincethe early 1900s. The technology was initially
driven by mili-tary applications in World War I and expanded by
World WarII. The military UAV applications are more advanced than
thecivilian applications. [4]
The civilians applications are likewise evolving in the
samedirections, due to their rapidly utilisation in various
applica-tions such as firefighting assistance, police observations
ofcivil disturbances and scenes of crimes, reconnaissance
fornatural disaster response, border security, traffic
surveillanceand precision agriculture [3, 4, 7].
Primicerio et al utilises unmanned UAV (VIPtero) equippedwith
camera for site-specific vineyard management. Nor-malised
Differential Vegetation Index (NDVI) values ac-quired by the
Tetracam ADC-lite camera mounted onVIPtero were compared to
ground-based NDVI valuesmeasured with the FieldSpec Pro
spectroradiometer to ver-ify the precision of the ADC system. The
vegetation indices
1http://www.ryegaard.dk/
1
-
DRAF
T
obtained from UAV images are in excellent agreement withthose
acquired with a ground-based high-resolution
spectro-radiometer.[5]
Merz et al addresses the design of an autonomous
unmannedhelicopter systems for remote sensing missions in
unknownenvironments. Focuses on the dependable autonomous
capa-bilities in operations related to Beyond Visual
Range(BVR)without a backup pilot by providing flight services.
Utilizesa method called Laser Imaging Detection and Ranging
(LI-DAR) for object detection which are applicable in real
worlddevelopment.[4]
Barrientos et al. utilises a team of UAVs to take pictures
inorder to create a full map, applying mosaicking proceduresfor
post processing and automatic task partitioning manage-ment which
is based on negotiation among the UAVs, consid-ering their state
and capabilities. Thus they combine a strat-egy which encompass
multi robot task planning, path plan-ning and UAV control for the
coverage of a crop field for datacollection. [1]
Table 1 compares the industry UAVs with the
research-basedprojects. It emphasise on product costs and survey
speed. Seethe table on page 3.
The parameters mentioned in the table description:
Frame type The type of UAV fixed or and rotary wings
Project Name If the project was an industry project or re-search
project
Retail Price How much the products cost
Control interface Is it controlled by a remote,pilot,
au-tonomously or continuous trajectories, and what systemsis
utilised to control them.
Imaging specification is utilised with photogrammetry
thatanalyses images providing information regarding therecognition
and identification of objects and their signifi-cance with respect
to the particular application.
Coverage per trip How long it takes to cover a specific
area.
Storage How the data is stored2
[TODO: include http://precisiondrone.com/ and Conserva-tionDrnes
+ include the parameters in the Main Contributionssubsection]
The parameters in Price and Survey speed are not com-pletely
up-to-date, accurate or comparable. There are fourfactors that
affect the survey speed: 1) Camera angle of view,2) Drone height,
3) Requested image overlap and 4) Dronecruise speed.
Most companies (except Trimble) did not provide informa-tion on
the value of the factors used for calculating the
survey2http://www.questuav.com/index.php,
http://uas.trimble.com/ux5,
http://www.cansel.ca/en/products/survey-instruments/data-collectors/trimble-ux5?vmcchk=1,http://www.cropcam.com/,
https://www.sensefly.com/drones/ebee.html, Pic-colo CC
http://www.cloudcaptech.com/piccolo command
center.shtm,http://www.geistware.com/rcmodeling/articles/beginner
1/#basic engine
speed. In Denmark authorisation must be gained to fly
above100m.
Regarding price it should be noted that we are not includ-ing
development costs for AgriDrone. One of the main ad-vantages of the
commercial UAVs compared with our OpenSource UAV is the camera
specification and the survey spec-ification. As the table shows,
the commercial UAVs are su-perior in those categories, but comes
with an expensive pricecompared to our UAV which is more cheaper
and adequate tocarry out the essential tasks as the commercial
products.
Another notable difference is that most of these other prod-ucts
are using a fixed-wing design. This implies that theyhave to be
pushed into the air on launch, either by hand or bycatapult, and do
a crash landing at the end of the mission.Conversely in our
prototype we have a rotary wing framewhich enables us to do
vertical take-off and landing (VTOL).VTOL enables the user to
initiate multiple aerial missionswithout low-level user involvement
in takeoff or landing.
Main ContributionsAs it has been mentioned in the interview
section , farmershave numerous tasks of routines to carry out in
their dailyactivities. The farms are really huge, thus it takes
time toplan for the tasks. However the solution presented enables
thefarmers to monitor and investigate the area in a sufficient
waycompared to current solution where the stakeholders have
toaccomplish the tasks manually by going out and checkingeach
place, which consumes a great deal of time.
Below we list some of the main features that our solution
con-tributes which none of the other projects offer.
GSM Our solution provides communication over the GSMdata
network. This allows for true remote control beyondvisual
range.
LPIS Our solution integrates with the Danish implementa-tion of
the EU Land Parcel Information Systems to presenta custom overlay
of the fields belonging to the farmer. InDenmark more than 300.000
fields are registered in LPIS.3In all of EU this figure is 135
million fields belonging to 8million farmers.4
Our systems contributes and facilities the farmers daily work.It
enables farmers to operate the UAV on the different devicesthat
have internet access thus providing flexibility.
3http://www.cowi.dk/menu/NyhederogMedier/Nyheder/Geografiskinformationogit/pages/digitaliseringaf300000marker.aspx
4http://ies.jrc.ec.europa.eu/our-activities/support-for-member-states/lpis-iacs.html
2
-
DRAF
T
ProjectName
Frame typeRetailPrice[EUR]
Control interface Imaging specification Coverage pertrip
Storage
QuestUAV200
Fixed WingSkyCircuitautopilot
18000 Custom made laptop and RC Hardware Modified LumixLX5
10MP
100Ha(1km2) in7-15 min SD card
QuestUAV300
Fixed WingSkyCircuitautopilot
20000 Custom made laptop and RC Hardware Tetracam MCA 6 100Ha
(1km2) in
7-15 min SD card
eBee Fixed Wing 8800 eMotion 2 software for desk-tops and
NoteBooks
Hardware 16 MP CameraSony NEX5R Photogram-metry Postflight Terra
3D-EB powered by Pix4D
150-1000Ha(1-10 km2) in 45min Resolution:3 cm/pixel
SD card
SwingletCAM
Fixed Wing 7800 eMotion 2 software for desk-tops and
NoteBooks
Hardware 16 MP CameraSony NEX5R Photogram-metry Postflight Terra
3D-EB powered by Pix4D
150-600Ha (1-6km2) in 30 minResolution: 3cm/pixel
SD card
CropCam Fixed Wing 5200 Custom made software forwindows 98
Hardware 12MP Pentax Op-tio A40
64Ha (0.64km2)in 20 min SD card
Trimble UX5 Fixed Wing 33100 Mission Planning Hardware 16.1 MP
mirror-less APS-C
100Ha (1 km2)in 31min Resolu-tion: 2.4cm/pixel
SD card
TrimbleX100
Fixed Wing less thanUX5 Mission Planning Hardware 10 MP
compact100Ha (1 km2)in 41min Resolu-tion: 3.3cm/pixel
SD card
AgEagle(early 2014)
Fixed WingPiccoloautopilot
5100 Piccolo Command Center Photogrammetry with Ag-Pixel
240Ha (2.4km2)in less than 60mins
MetaVRCFixed WingAPM2.5 au-topilot
MissionPlanner for Win-dows
Hardware Canon EOS MPhotogrammetry MetaVRTerrain Tools extension
toESRI ArcGIS
(up to 65 kmph)Resolution:3cm/pixel
SD card
AeryonScout
RotaryWings cus-tom autopilot
2200-3600 custom tablet Hardware Photo3S-NIRPhotogrammetry
Optionalsupplementary productpowered by Pix4D
(40 kmph) Reso-lution: 1cm/pixel
on-board andsent to basestation by ra-dio link
MicroDronesRotaryWings cus-tom autopilot
from 8000 GPS Waypoint navigationsystem
Photogrammetry poweredby Orbit GT
Resolution:1-7cm/pixel
VIPtero(modifiedMkrokopterHexa-II)
RotaryWings cus-tom autopilot
4000 koptertool Hardware Tetracam ADC-lite camera
0.5 ha. in 7.5min5.6 cm/pixelground resolutionat a flight
heightof 150m.
micro SDcard
CSIRO RotaryWings custom navigation computer Hardware RICOH
GX200digital camera with zoomleRICOH GR Digital III dig-ital
cameraRICOH GR Dig-ital III camera mod. forNIRFLIR Photon 640
ther-mal imaging camera GNCSystem (LIDAR)
SD card
AR100 Plat-form
RotaryWings custom backstepping+FSTcontroller
Hardware commercial zoomdigital camerato be tilted upto 100
deg
SD card
AgriDrone
RotaryWingsAPM2.5autopilot
600 web-application Hardware ArduCAM next25mm Photogramme-try
NDVI capable withDroneMapper or Pix4cloud75% image overlap
22ha in 12-20min(at 36 to 20kmph) Resolu-tion: 2cm/pixel
web-application(radio link)
Table 1: Comparison of commercial industry UAVs with research
based UAVs based on the costs and the survey speed.The last4
projects have been used in research projets on precision farming
specicly.
3
-
DRAF
T
METHODWork in the agriculture sector is very much a situated
experi-ence and the work required at one farm might not be the
sameas the work required at another farm at a given day. Becauseof
this nature of situated variability, we found it important toselect
some research methods that allows us to catch some ofthis situated
variability in a feasible way.
To gain an understanding of the domain, we initially made
aliterature survey and brainstorming session. Based on this wewere
able to formulate an interview guideline that we usedand made
telephone interviews with our stakeholders to getmore information
about what work scenarios play out in theirdomain, what factors
influence their work and how they en-visioned the use of drones
could be a part of their daily workdays.
Grouping together some of the things we found out during
ourinterviews, we quickly proceeded to make a low-fidelity
pro-totype a paper prototype that simulates the software run-ning
in a browser on a tablet device. Experience prototypingwas used to
evaluate our paper prototype together with ourstakeholder, and we
filmed the session to catch the subtle in-teraction taking place.
This feedback was then used to refineand implement our prototype
into an early product design.
Final evaluation was done as a usability study together with
aquantitative measure, namely a measure for the time it takesto do
an aerial survey compared against the time it wouldtake the farmer
to make the reconnaissance by foot and directobservation.
THE DESIGN OF THE PROTOTYPEOur design process was iterative and
comprised around 8weeks in total. In this period we gained a broad
understand-ing of Precision Agriculture and Arduino-based drones
andwere able to design and evaluate a prototype system.
Semi-structured InterviewsWith the domain in hand, we prepared a
short interview andthen got in contact with different people
involved in Agricul-ture in Denmark and located within 100km from
ITU. Thisgeographic restriction was to allow us to go visit the
farmersand try out our prototype in a situated context.
One of the first things we noticed was that everyone we
asked,knew and were enthusiastic about the potential of drones
foragriculture. This perhaps reflects the position of this
categoryof technology on the Gartner Hype Cycle5.
The interviews were designed to give us some key infor-mation
about current working procedures, determining mainchallenges, and
general conditions for work. While talkingover a telephone was not
enough to get a full picture of thescenarios of the stakeholders,
it provided us with a startingpoint for scenario analysis. Second
part of the interview wasdesigned to be a brainstorm on how they
perceived that dronescould be applied within their work.
The farmers seemed to be in consensus that the developmentof
drones within agriculture should start with large
farms5http://diydrones.com/profiles/blogs/where-drones-sit-on-the-gartner-hype-cycle
(> 500ha) since tenant farmers of small farms usually willbe
able to maintain an overview of the state of the farm
indi-vidually.
In Agriculture, not one day is the sameA common answer we got to
the question What is a normalday for you? was that they all replied
that not one day wasthe same. In Agriculture, the season, climate,
growth states,nutrient levels as well as human and facility
resources allplayed together and affected the required work and
ultimatelythe crop-yield. Examples of this include: 1) If wind is
strong,spraying with pesticides is prohibited, therefore spraying
usu-ally takes place 05.00 to 11.00 and 16.00 to 20.00; 2)
Ifquickspots6 develop on the field, they must be detected
andremoved quickly to protect the crop; 3) If crop is mature
andclimate conditions are dry, sunny and windy, then harvest
cantake place; 4) If parts of a field are covered in snow,
thenspraying with fertilisers is prohibited; 5) When a tractor
hasbeen loaded with pesticides, fungicides, herbicides or
fertilis-ers, it has to run until tank is empty, it is not advised
to keepremains in the tank for next run; 6) If grasslike weed have
de-veloped to two leaves, then they have to be sprayed to
avoidnutrient loss from the desired crop; 7) If [TODO: chec
withEric about facts] wheat have developed four leaves, then ithas
reached high maturity and can soon be harvested.
Another thing that came out of the interviews was that
inrelation to the planning and accounting/documentation workthat
takes place on the farm, being able to use maps with theproper
Markblok field parcel overlay was crucial to the us-ability of the
system. The Danish Markblok database is anational implementation of
the EU regulated Land Parcel In-formation System (LPIS)
initiative.7
We identified following application areas for drone-basedICT
systems within Agriculture based on our interviews:
Biomass estimation Weed detection Nutrient estimation Farmland
aerial monitoring, task management and commu-
nication
Facility monitoring and managementAside from these, application
areas such as detection of nitro-gen stress, water stress, pests
and crop diseases have also beenmentioned in literature [1] 8 and
researchers from KU are cur-rently working on a project on weed
detection and drones9
Paper PrototypingWe designed our system so that a web interface
is themain point of interaction. The drones base stations,
batteryrecharging requirements [TODO: reference
drone-charging6quickspots are small areas of weed on the field that
is expanding quickly7EU LPIS
http://marswiki.jrc.ec.europa.eu/wikicap/index.php/LPISIn Denmark,
the Danish AgriFish Agency under the Ministry of Food, Agriculture
andFisheries is the agency responsible of implementing the LPIS
nationally.
8http://www.weeklytimesnow.com.au/article/2013/10/25/586771
grain-and-hay.html9http://ing.dk/artikel/droner-spotter-ukrudt-i-danske-marker-161076
4
-
DRAF
T
group] and image transfer and post-processing should hap-pen
behind the curtains by the system requiring little userinvolvement.
To lower the entry barrier (increasing the rate oftechnological
adaptation), we decided to mimic the interfaceto something that the
user is already familiar with - GoogleMaps. Figure 1 shows our
initial mockup of a Google Mapinterface.
Figure 1: Google Map mockup used for early paperprototype
Then we printed and used scissors and pencil to create
apaper-based mockup of the UI and all the interactible ele-ments
down to 3rd level of interaction. See figure 2
Experience PrototypingWe used experience prototyping as an
evaluation method ofour paper prototype. This was carried out over
the course ofan hour in which the user was given the prototype and
told toimagine that this was a browser window on a
touch-enabledtablet computer. We assumed the role of silent
facilitatorsof interaction giving the user the space to
conceptualise thesystem in his own words. See figure 3
EvaluationThe user found the UI pretty intuitive and easy to
utilise. Wehad forgotten to draw the Markblok boundaries on the
mapand that was a point of critique because this is a necessity
tobe able to select which farmland to perform an action on.
Based on the feedback we also decided to deprecate our con-cept
of Task Management from the GUI. We had originallyconceived this as
a task management component for plan-ning human resources, but
during the experience prototypingit quickly became apparent that
this was not a use case. Theargument is that rather than trying to
augment the captureddata with a layer of task management logic, it
would be betterto remain transparent to the captured data and
provide APIsthat can be utilised in tractor GPS computers,
boomspray con-trollers or other peripheral computer-controlled farm
equip-ment.
(a) Main page GUI
(b) Interaction-dependent GUIelements
Figure 2: Paper prototype
5
-
DRAF
T(a) User interacts with View Data page
(b) User interacts with 3rd layer functionality on Main page
Figure 3: Experience Prototyping: Evaluation of thepaper-based
prototype
The Final DesignThe system consists of a web-based application
that providesall the information that is needed to carry out the
drone oper-ations on the field via its web-interface. The web-based
ap-plication provides a communication link between drones anduser
via GSM/GPRS connection to the web interface.
Figure 1 illustrates the main idea behind the implementa-tion.
We have 6 main parts in the application. The dronewhich is equipped
with an APM2.5 Arduino board, a uBlockGPS module, RC telemetry
link, GSM/GPRS module that im-plements the necessary Mavlink
communication protocol10and a Sony NEX5R modified NIR for NDVI11.
The secondlayer is the Web application which controls the
communi-cation between the integrated web-based application and
theflying drone. Using GSM module, the drone sends the sta-tus
information via the GPRS connection to the web appli-cation, then
the web application gathers and process the in-formation it gets
and do other steps according to the tasks(e.g. sends HTTP post with
the data to the interactive web-based map). The third layer we have
is the web-based in-teractive interface from which the stakeholders
can monitoreverything.The fourth layer is database, where we store
allnecessary transaction information. The fifth layer is the
inte-gration with openlayers.com API to provide a base map forour
interface, together with an overlay map of all danish farmparcels
(provided by the Markblok database of Danish Agri-
10https://pixhawk.ethz.ch/mavlink/11Open Source Single Camera
NDVI http://flightriot.com/vegetation-mapping-ndvi/
Fish Agency12). The sixth layer integrates a cloud or ded-icated
server based app for photogrammetry converting ourdrone captures
into orthomosaics that can be used for analy-sis.
Figure 4: System architecture
Flying the DroneThe drone model: 3DR DIY Hexa Copter. It was
pre-build with main features: camera, gps, telemetry,
autocontrolboard, arduino APM board. To this concept we added a
GSMmodule with the SIM card, which extended the possibilitiesfor
the drone. On the arduino based GSM module we im-plemented and
integrated a service, which can communicatewith our main web
application on top of Google (We foundtwo similar but currently
defunct projects13). The connectionwas made through GPRS and HTTP
json based commands.The flow of the drones operations are
illustrated by this se-quence:
1. The drone receives the GPS coordinates from the
web-application and a trigger to start mission
2. Starts the motors/engine
3. Fly to specified GPS coordinates
4. When reaching the coordinates, drone starts taking picturesat
the user-specified setting
5. Drone goes home to base station
6. Captured images are sent to web application over WiFi linkfor
post processing
The same workflow applies to the Scheduling mode as wellas the
instant-command mode.
Web ApplicationThe web-application in this project takes the
most valuableposition. It controls the communication between web
inter-face and the flying drones. All requests and responses
are12https://kortdata.fvm.dk/geoserver/web/?wicket:bookmarkablePage=:org.geoserver.
web.demo.MapPreviewPage13https://code.google.com/r/andreas-apm2-wip/
and https://code.google.com/p/
ardupilot-cellular-extension/
6
-
DRAF
T
based on JSON. The request is sent through the HTTP pro-tocol.
The application is hosted using Google app engine,it gives us 99.9%
system up-time, and ensures minimum re-sponse time. We provide
small API in the application, whichallow us to connect our
application to another third party ap-plications like periphery
farm equipment. It stable and recog-nise (response) only due to
predefine commands: get status,send gps coordinates, get pictures,
get biomass index at coor-dinate, check battery and etc.
Interactive web-based map applicationAfter the interviews and
the evaluation of the paper prototypewe decided to change the
following:
Status page The status page is the main page and allows userto
instantiate new missions quickly. Either as a scheduledevent to be
executed later or an instant event to be executednow.
Schedule page Our implementation provide the functional-ity
where you are able to schedule drone to fly to specificplace at
certain date and time.
Concept of Task We changed this from our original post-flight
human-centered notion of task planning to a pre-flight
drone-centered notion of task planning.
LPIS/Markblok Integrating with the national database offarmland
boundaries allowed us to provide a way of userinteraction where the
could just click to select a given fieldinstead of drawing custom
polygons.
Multiple fields and actions We enabled user to specify mul-tiple
fields and actions for single missions.
The system is compatible with multiple devices. So the sys-tem
can be accessed from smartphone as well as from com-puter or
tablet. Nowadays mobile phones are one of the mostwidely used
devices in the world, this is the main reason whywe decided to make
a browser-based system. Figure 5 showsthe final design of the web
interface.
Challenges encountered in developmentBelow we list some of the
challenges we encountered in theprocess of developing our
system.
Calibration We had to perform a live calibration on ourcompass.
It was difficult to find the compass calibrationposition because no
positions were shown the same appliesto calibrate
accelerometer.
Compassmot This part came us a surprise. It was an oner-ous task
to carry out. It required us to disconnect the pro-pellers ,flip
them over and rotate them one position aroundthe frame. And
simultaneously we had to push the copterdown into the ground when
the throttle is raised i.e. thepropellers are rotating.
Battery challenges (charging problems) At the first begin-ning
the connection cables doesnt fit our needs, that is whywe need to
manually replace the connection cables on thebattery itself. In
addition we had some problems with bat-tery charger.
Propellers During first assembly we didnt know that pro-pellers
of the drone are different and they need to be assem-bled in
special way. So at the beginning the drone couldnot take off
because of the propellers pulling and pushingin the wrong way,
while we were thinking that there someproblems with
calibration.
GSM/GPRS module To fully accomplish communicationbetween APM2.5
board drone and web-application wesuppose to connect and integrate
GPRS shield on APM2.5board. We did Arduino UNO communication
channel withGPRS shield and web-application, but it was challenging
tomove everything on APM2.5 board. Firstly, GPRS shieldrequires 5V
to work with.It was difficult to find the 5V pinalso on APM board.
Moreover we had problems connect-ing to the internet using some GSM
providers SIM cards.Also we could not figure out how to transfer
Arduino UNObased implementation code to APM2.5 board, the
compi-lation were failing due to missing different libraries.
EVALUATIONThe challenges that we encountered during this project
meantthat we could not in effect implement the fully workign
sys-tem as per our design. In order to do the evaluation anywaywe
used a version of the Wizard of Oz method in which weasked the user
to imagine that interacting with the web inter-face would control
the drone.
One of our goals was to have an intuitive interface so that
thebarrier of entry could remain low. Based on the evaluationfrom
our test person, this goal appears to have been achieved.Another of
our goals was to integrate a communication linkbetween the GSM
module and our web server. As mentionedbefore we could actually get
this to work in testing with theArduino UNO board, but porting it
to the APM2.5 wasnt suc-cessful. The third goal that we wanted to
show was that util-ising a drone can be a big timesaver for the
farmer. Our testperson informed us that, if it should be very
thorough, then in-specting a 20ha field would take him 3 hours. By
calculationsbased on an altitude of 100m and image overlap of 75%,
thenit would take the drone 12min at top speed (36kmph) and 20min
at a cruising speed of 20kmph. After capturing the datawe estimate
another 30 min in post-processing of the imageson a server to
produce the orthomosaic maps at 2cm/pixelresolution which the user
then can inspect. This means thedrone-based solution is
approximately 3times faster than thecurrent method. The bottleneck
is likely to be the rate atwhich we can transfer the images off the
drone to the cloudfor processing and this processing time increases
when thenumber of images increases.
Now, because we have only evaluated the system with one
testperson we do not have enough data to make a very strong
andcompelling argument. At Ryegaard there are 12 fulltime all-year
employes. Our test person was one of these employes.Their
organisational structure is very flat meaning that theyare all
involved in field inspection tasks when needed. Thatthey are all
exposed to the same type of field inspection tasksis a good
indicator that they will all find our system equallyuser friendly
as our main test person did.
7
-
DRAF
T(a) Main Page GUI of AgriDrone (b) Schedule Page GUI of
AgriDrone
Figure 5: Web interface of AgriDrone system
As part of a more general discussion, it is also worth noting,as
our test person commented. That the real difficulty facedis in
knowing how new measures of biomass or weed densityshould be
translated into actual measures of the amount ofchemicals to use on
the fields. With precision farming andthe ability to look at the
field with greater variability, then weare also able to spray
chemicals in matching variability, butinitially this does not
necessarily entail that less chemials willbe used - only trial and
error will lead to an understanding ofhow chemical usage can be
reduced in the long run. This isimportant because, in light of the
interviews, the value of thedrone system is proportional to the
value that the farmer cansave by adopting the drone system. Saving
human resourcesis one thing, but being able to use the information
captured ina way as to save costs on chemicals is probably what is
goingto be the crux of the argument.
In Denmark the rules and regulations for flying with dronesare
governed by Trafikstyrelsen in BL9-414. Specifically itmentions
that operation must be more than 150m away fromresidential areas
and main roads, 5km away from aerodromesand military areas, and at
a maximum height of 100m withfull pilot control override at all
times during flight. Exemp-tions can be granted for research and
commercial applica-tions. We think that as imaging technology
becomes moreadvanced, the desire to go above 100m in altitude is
going toincrease because people would want to capture bigger
areasfaster.
FUTURE WORKDuring the research and the user-participation, many
ideascame to us. We will list them here.
Suggestions for improvements based on Farmer:
1. Place multiple waypoints on map
2. Integration with Trimble GPS control system on tractors
Copter14http://www.trafikstyrelsen.dk/DA/Civil-luftfart/Luftfartserhverv/
Unmanned-Aircraft-Systems-UAS.aspx
The 3DR Hexa-C Copter that we are using has a maximumspeed of
36kmph. The performance of a drone has a lot to dowith the
aerodynamic properties caused by the physical de-sign of the drone.
3DR has since we started this project, dep-recated the HexaCopter
model and replaced it with the Y6model which is a tricopter frame
with rotors on both sides.They have also provided a guide to do
this modification one-self 15. Use the 3DR Iris QuadCopter16 or
some of the mini-UAV quadcopters available[TODO: find link to the
micro uavsystems].
PhotogrammetryDronemapper.org17 allows us to upload our photos
and turnthem into orthomosaics which will be a much nicer way
topresent to the stakeholders. Pix4uav Cloud
In the open-source world, maybe there will be possiblity touse
the frameworks GRASS 18 or ImageJ19 to integrate thephotogrammetry
features within our own web framework.
To improve precision of measurements, it might be necessaryto
introduce ground control points on the fields or maybe justat the
base station.
Data ManagementAs the geospatial image data grows we need better
waysof control and optimizations. The tools provided by
thePostGIS20 project allow easy hookup with a
PostgreSQLdatabase.
Integration with Danish Agro ICT
15Guide to turning the HexaCopter into a Y6
http://3drobotics.com/wp-content/uploads/2013/06/Y6-Conversion-Kit-Assembly-Instructions-Rev4.pdf
16Iris QuadCopter:
http://store.3drobotics.com/products/iris17Guide to Dronemapper
service - http://dronemapper.com/guidelines;
Desktop solutions: BAE Mosaic Manager
http://www.geospatialexploitationproducts.com/content/products/product-modules/mosaic-manager,
http://grass.osgeo.org/,
http://www.sensefly.com/operations/maps-and-3d.html,
http://conservationdrones.org/ http://pix4d.com/products/
18GRASS GIS http://grass.osgeo.org19ImageJ
http://rsbweb.nih.gov/ij/20PostGIS http://postgis.net
8
-
DRAF
T
PlanteIT21 offers a mobile application that registers whenusers
enter and leave a farm and prompts user to do time-tracking of
activities. Extending this solution with the infor-mation gathered
via AgriDrone will allow us to push infor-mation to the user in a
location-aware context.
Camera SystemAside from the TetraCam option mentioned in [5]. In
theConservationDrones project they use cameras that do NDVIin one
camera. These cameras are modified by the companyLDP LLC22 and they
offer conversions of many consumercameras.
Integration with Agricultural ProductsIn our usecase the farmer
had a Fendt tractor with TrimbleGPS system. It would be very
beneficial to provide a methodof sending the data (tables of GPS
coordinates and corre-sponding biomass index and other measured
parameters) tothe Trimble system that the farmer already brings
with himinto the field.
Crowd Sourcing and User Submitted ReadingsIn an article from IBM
Research 23, they mention another usescenario which we have not yet
explored. The scenario theypropose is to allow users to, with their
mobile device, takepictures in the farm of that is uploaded to the
website and canbe analysed by distributed experts. It bears some
resemblanceto Trimbles ConnectedFarm platform24.
Manual OverrideEven though various failsafe parameters are
already in place,to increase safety, disaster management and user
control fur-ther, there should be an option of manual override to
force thedrone to land. One way to do this could be to have a
AbortMission and a Emergency Halt button available as part ofthe
webinterface. But that would only work with internet ac-cess. A
better way would perhaps be to use the GSM moduledirectly. Where
placing a call to the drone would invoke theEmergency Halt
behaviour and land the drone instantly onground.
Authentication and scaleabilityWe didnt implement any kind of
authentication in our systemwhich means that anyone on the internet
could in effect trig-ger a drone to start flying a mission. To be a
useful systemfor farmers each tenant farmer needs to be secured
exclusivedrone operation rights and data-access to the fields
belongingto him. To allow for collective drone-sharing, it might
maesaense also to allow farmers to create groups in which theyshare
operational access to a shared set of drones with basestation
between their properties.
Integration with wireless sensor networks (WSN)Our interviews
showed that another important parameter forfarmers is to know about
the microclimatic variations
across21http://it.dlbr.dk/DLBRPlanteIT/22http://www.maxmax.com/digital
still cameras.htm23IBM Research - The future of PA:
http://www.research.ibm.com/articles/precision
agriculture.shtml24http://www.connectedfarm.com/
the fields - soil moisture levels, soil nutrient levels and
croplocal temperature conditions. These kinds of data would notbe
feasible to acquire using a drone flying at 100m altitude.Instead
we propose a solution that involves a WSN of Wasp-motes using
Libeliums Agricultural sensors25 as a startingpoint.
CONCLUSIONModern industrial agriculture is a very highly coupled
systemwith dependencies on many uncontrollable components. Thisis
in general why introducing a system that can facilitate
earlydetection early response is a good idea.
ACKNOWLEDGEMENTSWe would like to thank all the people who
participated in theinterviews and helped ground this project:
stdansk Land-brugsradgivning, Nordsjllands Landboforening,
Ryegaardand Trudsholm godser and our supervisors and technical
per-sonnel at ITU.
REFERENCES1. Barrientos, A., Colorado, J., Cerro, J. d.,
Martinez, A.,
Rossi, C., Sanz, D., and Valente, J. Aerial remote sensingin
agriculture: A practical approach to area coverage andpath planning
for fleets of mini aerial robots. Journal ofField Robotics 28, 5
(2011), 667689.
2. Bramley, R. Lessons from nearly 20 years of
precisionagriculture research, development, and adoption as aguide
to its appropriate application. Crop and PastureScience 60, 3
(2009), 197217.
3. Eisenbeiss, H. The autonomous mini helicopter: apowerful
platform for mobile mapping. Int. Arch.Photogramm. Remote Sens.
Spat. Inf. Sci 37 (2008),977983.
4. Merz, T., and Chapman, S. Autonomous unmannedhelicopter
system for remote sensing missions inunknown environments. In
Conference on UnmannedAerial Vehicle in Geomatics (UAV-g) (2011),
1416.
5. Primicerio, J., Di Gennaro, S. F., Fiorillo, E., Genesio,
L.,Lugato, E., Matese, A., and Vaccari, F. P. A flexibleunmanned
aerial vehicle for precision agriculture.Precision Agriculture 13,
4 (2012), 517523.
6. Xiang, H., and Tian, L. Development of autonomousunmanned
helicopter based agricultural remote sensingsystem. In ASABE Annual
Meeting. Portland, Oregon(2006).
7. Zecha, C., Link, J., and Claupein, W. Mobile sensorplatforms:
categorisation and research applications inprecision farming.
25http://www.libelium.com/development/waspmote/documentation/agriculture-board-technical-guide/
9
IntroductionRelated WorkMain Contributions
MethodThe Design of the prototypeSemi-structured InterviewsIn
Agriculture, not one day is the same
Paper PrototypingExperience PrototypingEvaluationThe Final
DesignFlying the DroneWeb ApplicationInteractive web-based map
applicationChallenges encountered in development
EvaluationFuture WorkCopterPhotogrammetryData
ManagementIntegration with Danish Agro ICTCamera SystemIntegration
with Agricultural ProductsCrowd Sourcing and User Submitted
ReadingsManual OverrideAuthentication and scaleabilityIntegration
with wireless sensor networks (WSN)
ConclusionAcknowledgementsREFERENCES