Land Information Science

Design Considerations for Remote Sensing Payloads onInexpensive Unmanned Autonomous Aerial Vehicles

Scot Smith, Zoltan Szantoi, John Perry, Fran1din Percival,and Brandon Evers

ABSTRACT: This l)al)wr describes the latest version of the University of Florida (UF)'s uninanned autoio-inoits vehicle (Rf,W). named the MAKO MAKO. The NIAKO MAKO can operate in fully autonomous w%av-ipoint navigation in1ode. including autoniomus takeoff and landing, allowing for repeatable, predictablegrotind coverage. Other features essential for itht mission profile include hand-launch ability for takeoff, aswIell as a waterproof Fuselage for aquatic landing. Low-altitude, high-r,esolution imaging is tacilitated by acruise speed of I5 ni/s. -Fle sensor payload on the MAKO is a digital single lens reflex (D)SLR) cameraolpcrawed at the shortesi exposure to minimize the etftect ofi motion bluiiing. The resulting images are capa-ble of resolving objects on the ground as small as six cm. The MAKO has been used for a number of appli-cations, including mapping ivading bird nests in the Florida Everglades and elsewhere, nionitloring theetlicacy of oetuliant spray progranis in Lake Okeechobee. identification of invasive exotic vegetation, andm1app)ing of bisoln.

KEYWORDS: Unnmanned automnous vehicle, remote sensing. digital single lens reflex

Introduction

Te ULniversity of Florida (LIF) lntPdisciplinary UjAV Research (;rotip hascompleted development of the fifth

generation (the MAKO) of small UAVs designedor- iiattiral resource and infrastructure applica-

tions (Watts et al. 2006; 2007; 2008). Over thepast 10 years, tJAVs have evolved to accommodatea rolist set of operational capabilities. Chiefamong these is the ability to operate in autono-mots waypoint navigation mode, allowing forrepeatable, predictal)le ground coverage. Otherfeatures essential for the mission profile inchldehand-lauinching for takeof fi-om nearly any plat-form, as well as a waterproof fuselage for aquaticlanding. ILow-altitude, high-resolution imaging isfacilitated by low-speed cruising (15 m/s), with a

Scot Smith, Zoltan Szantoi, and John Perry, School of ForestResources and Conservation, Geomatics Program, 301 ReedLab, P0. Box 110565, University of Florida, Gainesville, Florida32611-0565. E-mails: <sesmith@ufl.edu>, <zszantoi@ufl.edu>, and <justjohn@ufl.edu>. Franklin Percival, FloridaCooperative Fish and Wildlife Research, Building 810, Univer-sity of Florida, Gainesville, Florida 32611-0565. E-mail: <per-civalf@ufl.edu>. Brandon Evers, Department of Mechanicaland Aerospace Engineering, Micro Air Vehicles Lab, 114 NEB,University of Florida, Gainesville, Florida 32611-0565. E-mail:<bevers@ufl.edu>.

stall speed of 11 m/s and an efficient cruise speed(based on propeller selection) of up to 30 m/s.The all-electric drive system improves the reliabil-itY of the platform and minimizes airframe vibra-tion. The total weight of the airfr'ame. avionics,battery, and motor (excluding the remote sens-ing payload) is approximately 4.5 kg. A photo-graph of the MAKO is shown in Figure 1.

The weight specification for the remote sensingpayload is 1 kg (kg), with a maximuin payload of2.25 kg. The dimension of the sensor suite pay-load bay is approximately 15 cm x 15 cm x 25 cm,a total of 6000 cin'. In addition to physical con-straints, it is important to consider the powerrequired for the sensor payload that will reducethe available power for the avionics and motor.The standard payload specification of 1 kg draw-ing 1.5 A yields a specified flight duration of onehour.

Payload Controller

The MAKO's primary mission is the collection ofhigh-resolution, directly georeferenced, visible-spectrum imagery. The sensor suite and payloadcontroller developed for this platform is a casestudy in the design of remote sensing payloads forsmall UAVs. The following constraints and fea-tures were identified dturing the planning phaseand are used to anchor the design of the payload.

SurveYing and land In.fimration Science, U)1. 70. No. 3, 2010, /p). 131-1-37

Figure 1. MAKO launching from an airboat.

"* Physical constraints: The platform specifica-

tions given above limited the weight, size, andpower consumption of the payload.

"* Autonomy: The payload and avionics of the

plane should be mutually independent. Thatis, it should be possible to fly the MAKO with-out the payload and operate the payload withoutthe plane. This is particularly important duringthe design and testing phases, and also elimi-nates the need to engineer the payload to thelevel of reliability required for avionics.

"* Consumer off-the-shelf hardware (COTS):Modularity, extensibility, ease of replacement,and minimization of development cost and sale

price all deter the use of custom components."* Remote (wireless) and direct (cabled) inter-

faces: It is an operational necessity to haveremote access for command and communica-tion with the payload, both for sensor control

and status updates. The large data sets gener-ated by the payload, however, preclude the soleuse of wireless interfaces, so hardwired inter-

faces for data transfer and payload trouble-shooting are necessary.

"* Unified data bus: Limiting inter device commu-nication to standards such as USB, Ethernet,and RS-232 simplifies the cabling and choice of

connectors, an important consideration in the

tight confines of the airframe."* Unified data storage: Centralizing the data stor-

age on one device eliminates the complexity

and physical overhead of having redundantdata storage devices.

"* Unified sensor control: Command and commu-nication with the payload in-flight must be aggre-

gated for transmission over a single wireless datalink. This implies the use of a device that has

direct control over all sensors and subsystems.The majority of these design considerations fol-

low practically from the notion of a small, inex-

pensive UAV, such as the physical constraints. Not

as obvious, howeveI, is the use of COTS hardware.However, the development and operation of pre-

vious generations of small remote sensing UAVsinformed many of the other constraints. For

example, the proclivity to use COTS hardwarecame from the realities of implementing cus-

tom hardware, which required specialized knowl-

edge and extensive redesign for even relativelyminor changes in the sensor configuration, such

as changing the camera model (Bowman 2008).

Another pertinent example was the autonomy ofthe payload. In previous generations, the navi-

gation system used by the avionics was exploitedto provide direct georeferencing to the payload.

However, the limitations of this became appar-ent once it was found that the navigation accu-

racy was insufficient for the payload purpose,

yet, it was impractical (and even infeasible) to

upgrade the navigation components of the auto-pilot (Wilkinson 2007).

Several of the design considerations were met

simultaneouslywith the decision to use a MicrosoftWindows XP-compatible computer. Pairing theVIA Technologies, Inc.® Pico-ITX form factor

motherboard, the EPIA'T1 P700-10,L, with anIntel' X18-M solid state drive provided U.S.B,GigE, and RS-232 interfaces; a single data storage

device; a centralized sensor controller; and the

ability to rapidly develop extensible sensor con-trol software. The use of the Windows XPTM oper-ating system had the disadvantage of not providing

low-level real-time access to the hardware laver,but greatly simplified the ability to connect COTS

sensors that often have proprietary SDKs avail-able only for Windows XPIM.

The wireless interface is provided by connect-ing a wireless modem to the payload computer. In

a compromise of the fully autonomous designconsideration, this connection was provided by"piggy-backing" the payload communications on

the wireless data link used by the autopilot.(Procerus Technologies-', Kestrel AutopilotTM).However, this was a standard service provided by

the autopilot and the eliminated the need for

redundant wireless data links on the plane.The navigation sensors are of paramount impor-

tance to the mapping accuracy of the direct geo-referencing solution. The navigation sensorsprovide the EOPs and hence the absolute orien-

tation of the photogrammetric solution. The twocanonical components of' the navigation systemare the Inertial Mapping System (IMU) and GPS(Global Positioning System) sensors, which arefurther integrated using a state estimation filter

I 3'2 $u �vem'ing a 11(1 l.a iid In/ormaUon !,rie?ureSurvrv ing and Land h!ionnalion Sciemce17?2

((Chatfield 1997). Additional sensors such as mag-netonieters and air pressure sensors are oftenused to augment the navigation performance.

Size, weight, and cost restrictions severely limitthe selection of the navigation sensors. In gen-eral, the selection for small UAV platforms is lim-ited to the two lowest-accuracy classes of thesedevices: a MEMS-based IMU and a code-solutionsingle-frequency GPS. As a result, the primary fac-tors in selecting the INS/GPS were the relativeaccuracy, robust design, and ease of integration.Frhe Xsens, Inc.'N MTi-G'IM was chosen for imple-mentation. Key features of the MTi-G"'M include abuilt-in Kalman filter for a real-time orientationparameter solution, a raw data output mode for

postprocessing refinement, a high update rate,and a simple serial data communication protocol[Xsens M'Ti-G]. Inportantly, an interface is alsoprovided for device synchronization based on theGIPS IPPS signal.

A final hardware component common to manysensor suite implementations is a device that pro-

vides sensor synchronization facilities. This is par-

ticularlv true where the desired remote sensingproducts are dependent on the fusion of the datafrom multiple sensors, as is the case for direct geo-refemrencing. Often, this facility is integrated intothe navigation sensors due to their dependencyon accurate timekeeping, but the appearance ofthis feature is less frequent in the class of naviga-tion sensors compatible with a small UAV. Becausea reasonable ofIlthe-shelf solution has not beenidentified, this was the sole hardware componenton the sensor suite that was custom-designed. Adetailed description of the device and its methodof operation is available (Perry and Childs 2009).

Optical Sensors

Given the flexibility of the payload controller, theincreased options for the sensor suite were lim-ited only by the practicalities of size and weight.There are three practical options for optical sen-sors on the MAKO:1. High definition aerial video camera2. Compact rangefinder digital camera3. Digital single lens reflex camera

High-definition video (HDV) refers to anyvideo system of a higher resolution than that ofa standard-definition (SD) video and most com-monly involves display resolutions of 1280 x 720

pixels. An example of a relatively lightweightHl)V camera is shown in Figure 2. With a weight

Figure 2. High definition aerial video camera (PanasonicHDC-SD9).

Figure 3. Image from an HOV mounted on a UAV. Source:Draganfly Intions (www.draganfly.com).

of 275 g (gr), the camera can be flown in theNIAKO platform (for results, see Figure 3).

One of the most important performance param-eters of an optical sensor is resolution as definedby the smallest object on the ground that can bedetected. One means of indirectly measuring thisis the array of pixels tor a system. Larger arraysyield higher resolution. The best HDV camerashave a resolution of 1920 x 1080 pixels.'

At a flying height of 100 m, a 3 mim focal lengthHDV svstem has the ability to detect targets on

the ground that are approximately 3 m x 3 m ifthe target to background contrast is high. Forsome applications, this resolution might be ade-quate. For example, if the purpose of the missionwere to assess the extent of a wildfire, a resolutiongreater than 5 ii would be sufficient to be useful.However, if the mission were to calculate theamount of potential fuel for the fire in adjacent

areas, this resolution would probably be toocoarse.

' For comparison, a typical 12 mega pixel, mp, compact rangefinder camera has a pixel array of approximately 4000 x 3000 pixels.

I W. 70, Not. 3 133

Compact or rangefinder cameras are camerasdesigned primarily for simple operation. Alnexample of a typical compact digital camera isshown in Figure 4. Most rangefinder cameras usefocus-free lenses or autofocus for Focusing andautomatic systems for setting the exposure.Compact cameras are distinguished from single-lens reflex cameras (DSLRs) in several respects,but the two most significant differences for UAVapplications are (1) size and weight and (2) sen-sot size. Compact or rangefinder cameras aresmaller and lighter than any DSLR on the mar-ket. The MAKO aircraft can fit two rangefindersin the payload bay and only one SIR. This wouldbe important ift for example, you wanted to simul-taneously shoot a natural colormission and an infrared colormission.

A Canon Powershot 650 compactcamera was flown on the MAKOtor several missions, including oneover Cedar Key Florida. An exam-ple of the imagery obtained withthis camera is shown in Figure 5.In this example, the objective ofthe mission was to detect pelicannests located in the trees. It waspossible to detect the nests butimpossible to see whether or noteggs were present.

The second major differencebetween a compact rangefinderand a DSLR for UAV applica-tions is the size of the sensor.The sensor on most compactrangefinders is a firaction of thaton most DSLRs. For a given num-ber of pixels, the larger sensorallows for larger pixels or photosites which then provide widerdynamic range and lower noiseat high ISO (International Orga-nization for Standards) sensitivity Figure 5. Imagelevels. As a consequence, DSLIRsproduce better quality images ingeneral, but especially under high contrast or lowlight situations.

(Compact. rangefinder cameras, in general, areoptically inferior to even standard grade lensesdesigned for a DSLR camera. One reason for thisare the design compromises needed for variablefocal length lenses found on all compact camerasas opposed to fixed focal length lenses.

DSLRs have only one lens, and a mirror divertsthe image from the lens into the viewfinder. That

Figure 4. Canon 650 Powershot.

s from the Canon Powershot 650 of Seahorse Island, Florida.

mirror then retracts when the picture is taken sothat the image can be recorded on the sensor. Anadvantage of a DSLR over a compact rangefinderis sensor size. The larger the sensor, the betterthe image quality will be. The largest sensorsare referred to as "full-frame," and are the samesize as 35 mm film (the image fbrmat is 24 mm x36 rnm). These sensors are Used in DSI,Rs tooheavy for most UAVs. Most DSLRs use a smallersensor commonly referred to as APS-C sized,that is, approximately 22 mm x 15 mim, which is

'34 S,rve�, iig i/fl (I I iintl Ifl/0/ fl/fit/Oh .S//e////'13•4 Sunrt'vingorl1and Lw tltý)k-iialioll Science('

al)bout 40 percent of the area of a ftull-frame

sensor. Other sensor sizes found in DSLRs

include the Four Thirds System sensor used by

Olympus that is 26 percent of full frame.

Another advantage of DSLRs over compact (ig-

ital cameras is that DSLR cameras allow the user

to change lenses. The ability to exchange lenses

to select the best lens for the mission is very

important. The MAKO has flown lenses as small

and light as the standard grade Olympus Zuiko

25 inm lens when resolution was not paramount

to tdIe mission and lenses as large and heavy as the

higher grade Olympus Zuiko 50 mm lens when

resolution was important.Many lens manufacturers produce three grades.

Olympus, For example, produces standard, high

a.I

Figure 6. Images from (a) 25 mm vs. (b) 35 mm vs. (c) 50for equivalent focal length.

Figure 7. Olympus E-420 DSLR camera with OlympusZuiko 25 mm pancake lens.

grade, and super high grade lines of lenses. Thedifference between the grades is qualitv of mate-rials, including glass, and build quality. Typically,higher grades of lens are larger and heavie.linage quality difference can be pronounced (seeFigure 6).

The Olympus Zuiko 25 mm standard grade lensproduced a relatively low resolution image com-pared with the 35 mm standard grade lens. The50 mm high grade lens produced a significantlyhigher resolution image than the other lenses.

The Olympus 420 DSLR

The primary imaging sensor on the MAKO is anOlympus E-420 DSLR camera (see Figure 7). This

digital camera was selected pri-marily for its compact size androbust feature set, including asoftware interface that allows

complete control of the canierasettings, exposure triggering, and

Eb. captured image transfer to the

host.An Olyinpus Zuiko 25 mm stan-

(lard grade pancake lens wasselected because of its lightweight and compact size. A weightand size comparison between the

mm Olympus lenses 420 and other popular DSLRs isshown in Table 1. An importantconsideration in moving to a

DSLR over the Canon Powershot 65() was supe-rior optical performance. The overall success of

the platform as a remote sensing tool relies asmuch on the quality and resolution of the finalimagery as on the spatial accuracy of thegeoreferencing.

During a typical mission, the Olvmpius 420 is

operated at the shortest possible exposure speedto minimize the effect of motion blurring. As aresult, aperture priority mode is usually set to anexposure duration of 1/2000th of a second orhigher and an ISO sensitivity of 100. The Olympusautomatic exposure-metering program, set toshutter priority, normally sets the aperture tof/2.8 with these parameters under normal day-light conditions. No discernable motion blurringoccurs at these settings, and increasing the

Type _

Dimensions (mm)

Olympus E-420 T129.5 x 91 x 53

Nikon D60

1 126 x 94 x 64

Canon EOS 450D

1 129 x 98 x 62

W eight (g) (no battery)380

Table 1. Olympus E-420 size and weight comparison with similar cameras.

471 475

135

0 OLYMPUYMPL

---- ----- I

Udl. 70, No. 3

I

exposure duration to 1/4000thof a second does not result in amarked improvement in sharp-ness and has less contrast in low-light conditions.

Image Capture Rate

The TruePic III Image Processorenables the camera to shoot at upto 3.5 frames per second insequenced shooting and fullmegapixel mode, until the bufferspace of the memory card reachescapacity. Our capture rate is lim-ited to about 2.2 seconds, the timeit takes to trigger an image cap-ture, focus and expose the sensor,process the image, and transfer itto the payload computer.

Examples of the images takenwith the Olympus 420 are shown Figure 8. Examin Figure 8. Figure 8a shows bee and (b) theIndian Canal on Lake Okeecho-bee and Figure 8b was taken inthe Florida Everglades.

Experimental Work and PerformanceAnalysis

Applications the MAKO has been used for thusfar fall into two categories: (1) detection of physi-ological changes in vegetation not perceptible bythe human eye and (2) wildlife inventories. Forthe first category, the MAKO has been used todetect the reaction of vegetation in LakeOkeechobee Florida to herbicides. Imagery wastaken before and after application of the herbi-cide. It was found that the imagery was able to dis-play a change in the spectral reflectance of thevegetation before it became visually obvious. Thisis very useful because agencies that regularlyapply herbicides need a method to measure theefficacy of their treatments. The UAV is a muchmore efficient means of doing so than fieldwork.

In the second category, the UAV has been usedto perform inventories of wading birds in Florida.The DSLR system allows the interpreter to counteggs in nests of some birds. It can also detect rep-tiles such as alligators and large snakes. Suchapplications are of obvious benefit given that thenumber one cause of fatalities of field Fish andWildlife Service employees are plane crashes. Ifthe UAV can serve as a surrogate to mannedflights, many lives will be saved.

nplesEver

of images from the Olympus 420 from (a) Lake Okeecho-glades.

Conclusions

Small inexpensive UAVs such as the MAKO havea large number of potential applications andpromise to open a new era in remote sensing. Wefound that the advantages of a DSLR comparedto a compact rangefinder camera or an HDVmake it a better choice. DSLRs have better imagequality, more lenses to choose from, and usuallymore versatility in terms of exposure control. Asthe design of UAVs improves with respect to pay-loads, DSLRs can be mounted with higher-quality(and heavier) lenses that will further expand thepotential applications.

REFERENCES

Bowman, S. 2008. Design and validation of an autono-moUs rapid mapping system using a small UAV, Mas-ter's Thesis, University of Florida, Gainesville,Florida.

Chatfield, A. 1997. Fundamentals of high accuracyinertial navigation. Progress in ikstronautics andAeronautics Series, Volume VI 74. Reston, Virginia:AIAA. p. 800.

Perry, J., and j. Childs. 2009. Timing on the fly: Syn-chronization for direct georetferencing on smallUAVs. InrsideGNSS 4 (6): 34-40.

Watts, A., F. Percival, L. Pearlstine, P. Irju, B. Dewitt,S. Snmith, A. Mohamed, and S. Bowsman. 2006. Devel-opment of an autonomous unmanned aerial

I �'6 .Surveving andl I (old In/or?nation Science136 Survey,ing and La?id Il6friiatiari Scienwe

vchicle (UAV) system for Wildlife and EcologicalResearch. Poster presentation, 3rd National Confer-ence on Coastal and Estuarine Habitat Restoration,New Orleans, Louisiana, 9-13 December.

Watts, A, F. Percival, L. Pearlstine, P. Ifju, B. Dewitt,S. Smith, A. Mohamed, and S. Bowman. 2007.Unniantned airctaft systems for natural resource

monitoring and ecological research. Plenary Pre-sentation, Association for Unmanned Vehicle Sys-tens International, Washington, DC, 6-9 August.

Watts, A., W. Bowman, A. Abd-Elrahman. A. Mohamed,.j. Petry, Y. Kaddoura, and K. Lee. 2008. Unmanned

aircraft systems (UASs) for ecological research andnatural-resource monitoring. Journal of EcologicalRestoration 26(1): 13-4.

Wilkinson, B. 2007. The design of georeferencing tech-niques for an unmanned autonomous aerial vehiclefor use with wildlife inventory surveys: A case studv ofthe National Bison Range, Montana. Master's Thesis,University of Florida, Gainesxille, Florida, USA.

IA)?. 7U, No, 3 13,137UoL 70, No,

COPYRIGHT INFORMATION

Author:

Title:

Source:

ISSN:

Publisher:

Smith, Scot; Szantoi, Zoltan; Perry, John

Design Considerations for Remote Sensing Payloads on InexpensiveUnmanned Autonomous Aerial Vehicles

Surv Land Inf Sci 70 no3 S 2010 p. 131-7

1538-1242

National Society of Professional Surveyors

6 Montgomery Village Avenue, Suite 403, Gaithersburg, MD 20879

The magazine publisher is the copyright holder of this article and it is reproducedwith permission. Further reproduction of this article in violation of the copyright isprohibited. To contact the publisher: www.acsm.net

This article may be used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub- licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.