Table Of ContentCamera Networks
TheAcquisitionandAnalysisofVideosoverWideAreas
Synthesis Lectures on Computer
Vision
Editors
GérardMedioni,UniversityofSouthernCalifornia
SvenDickinson,UniversityofToronto
SynthesisLecturesonComputerVisioniseditedbyGérardMedionioftheUniversityof
SouthernCaliforniaandSvenDickinsonoftheUniversityofToronto.Theserieswillpublish50-
to150pagepublicationsontopicspertainingtocomputervisionandpatternrecognition.The
scopewilllargelyfollowthepurviewofpremiercomputerscienceconferences,suchasICCV,
CVPR,andECCV.Potentialtopicsinclude,butnotarelimitedto:
(cid:129) ApplicationsandCaseStudiesforComputerVision
(cid:129) Color,Illumination,andTexture
(cid:129) ComputationalPhotographyandVideo
(cid:129) EarlyandBiologically-inspiredVision
(cid:129) FaceandGestureAnalysis
(cid:129) IlluminationandReflectanceModeling
(cid:129) Image-BasedModeling
(cid:129) ImageandVideoRetrieval
(cid:129) MedicalImageAnalysis
(cid:129) MotionandTracking
(cid:129) ObjectDetection,Recognition,andCategorization
(cid:129) SegmentationandGrouping
(cid:129) Sensors
(cid:129) Shape-from-X
iii
(cid:129) StereoandStructurefromMotion
(cid:129) ShapeRepresentationandMatching
(cid:129) StatisticalMethodsandLearning
(cid:129) PerformanceEvaluation
(cid:129) VideoAnalysisandEventRecognition
CameraNetworks:TheAcquisitionandAnalysisofVideosoverWideAreas
AmitK.Roy-ChowdhuryandBiSong
2011
DeformableSurface3DReconstructionfromMonocularImages
MathieuSalzmannandPascalFua
2010
Boosting-BasedFaceDetectionandAdaptation
ChaZhangandZhengyouZhang
2010
Image-BasedModelingofPlantsandTrees
SingBingKangandLongQuan
2009
Copyright© 2012byMorgan&Claypool
Allrightsreserved.Nopartofthispublicationmaybereproduced,storedinaretrievalsystem,ortransmittedin
anyformorbyanymeans—electronic,mechanical,photocopy,recording,oranyotherexceptforbriefquotationsin
printedreviews,withoutthepriorpermissionofthepublisher.
CameraNetworks:TheAcquisitionandAnalysisofVideosoverWideAreas
AmitK.Roy-ChowdhuryandBiSong
www.morganclaypool.com
ISBN:9781608456741 paperback
ISBN:9781608456758 ebook
DOI10.2200/S00400ED1V01Y201201COV004
APublicationintheMorgan&ClaypoolPublishersseries
SYNTHESISLECTURESONCOMPUTERVISION
Lecture#4
SeriesEditors:GérardMedioni,UniversityofSouthernCalifornia
SvenDickinson,UniversityofToronto
SeriesISSN
SynthesisLecturesonComputerVision
Print2153-1056 Electronic2153-1064
Camera Networks
TheAcquisitionandAnalysisofVideosoverWideAreas
Amit K.Roy-Chowdhury and Bi Song
UniversityofCalifornia,Riverside
SYNTHESISLECTURESONCOMPUTERVISION#4
M
&C Morgan &cLaypool publishers
ABSTRACT
Asnetworksofvideocamerasareinstalledinmanyapplicationslikesecurityandsurveillance,envi-
ronmentalmonitoring,disasterresponse,andassistedlivingfacilities,amongothers,imageunder-
standingincameranetworksisbecominganimportantareaofresearchandtechnologydevelopment.
Therearemanychallengesthatneedtobeaddressedintheprocess.Someofthemarelistedbelow.
-Traditionalcomputervisionchallengesintrackingandrecognition,robustnesstopose,illumina-
tion,occlusion,clutter,recognitionofobjects,andactivities;
-Aggregatinglocalinformationforwideareasceneunderstanding,likeobtainingstable,long-term
tracksofobjects;
- Positioning of the cameras and dynamic control of pan-tilt-zoom (PTZ) cameras for optimal
sensing;
-Distributedprocessingandsceneanalysisalgorithms;
-Resourceconstraintsimposedbydifferentapplicationslikesecurityandsurveillance,environmental
monitoring,disasterresponse,assistedlivingfacilities,etc.
Inthisbook,wefocusonthebasicresearchproblemsincameranetworks,reviewthecurrent
state-of-the-artandpresentadetaileddescriptionofsomeoftherecentlydevelopedmethodologies.
The major underlying theme in all the work presented is to take a network-centric view whereby
theoveralldecisionsaremadeatthenetworklevel.Thisissometimesachievedbyaccumulatingall
thedataatacentralserver,whileatothertimesbyexchangingdecisionsmadebyindividualcameras
basedontheirlocallysenseddata.
Chapter1startswithanoverviewoftheproblemsincameranetworksandthemajorresearch
directions. Some of the currently available experimental testbeds are also discussed here. One of
thefundamentaltasksintheanalysisofdynamicscenesistotrackobjects.Sincecameranetworks
cover a large area, the systems need to be able to track over such wide areas where there could
bebothoverlappingandnon-overlappingfieldsofviewofthecameras,asaddressedinChapter2.
Distributedprocessingisanotherchallengeincameranetworksandrecentmethodshaveshownhow
todotracking,poseestimationandcalibrationinadistributedenvironment.Consensusalgorithms
thatenablethesetasksaredescribedinChapter3.Chapter4summarizesafewapproachesonobject
andactivityrecognitioninbothdistributedandcentralizedcameranetworkenvironments.Allthese
methods have focused primarily on the analysis side given that images are being obtained by the
cameras.Efficientutilizationofsuchnetworksoftencallsforactivesensing,wherebytheacquisition
and analysis phases are closely linked.We discuss this issue in detail in Chapter 5 and show how
collaborative and opportunistic sensing in a camera network can be achieved. Finally, Chapter 6
concludesthebookbyhighlightingthemajordirectionsforfutureresearch.
KEYWORDS
wideareatracking,distributedvideoanalysis,Kalmanconsensus,distributedtracking,
recognition,activesensing,opportunisticsensing
vii
Amit:To my parents for all they have done
Bi:To my parents
ix
Contents
Preface .................................................................xiii
1 AnIntroductiontoCameraNetworks........................................1
1.1 Researchdirections ..................................................... 1
1.1.1 CameraNetworkTopology ........................................ 1
1.1.2 WideAreaTracking .............................................. 2
1.1.3 DistributedProcessing ............................................ 3
1.1.4 CameraNetworkControl(ACTIVEvision).......................... 4
1.1.5 MobileCameraNetworks ......................................... 5
1.1.6 SimulationinCameraNetworks .................................... 7
1.1.7 ExperimentalTestbeds ............................................ 7
1.1.8 ApplicationDomains ............................................. 8
1.2 Organizationofthebook ................................................ 9
2 Wide-AreaTracking ..................................................... 11
2.1 ReviewofMulti-TargetTrackingApproaches.............................. 11
2.1.1 KalmanFilter-BasedTracker...................................... 12
2.1.2 ParticleFilter-BasedTracker ...................................... 12
2.1.3 Multi-HypothesisTracking(MHT)................................ 13
2.1.4 JointProbabilisticDataAssociationFilters(JPDAF) ................. 14
2.2 TrackinginaCameraNetwork-ProblemFormulation...................... 14
2.3 AReviewonCameraNetworkTracking .................................. 16
2.4 On-LineLearningUsingAffinityModels................................. 17
2.5 TrackletAssociationUsingStochasticSearch .............................. 19
2.6 PersonReidentification ................................................. 24
2.7 LearningaCameraNetworkTopology ................................... 27
2.8 ConsistentLabelingwithOverlappingFieldsofView ...................... 29
2.9 Conclusions .......................................................... 31
3 DistributedProcessinginCameraNetworks................................ 33
3.1 ConsensusAlgorithmsforDistributedEstimation.......................... 33
x
3.2 DecentralizedandDistributedTracking................................... 35
3.2.1 DecentralizedTracking........................................... 35
3.2.2 DistributedTracking............................................. 36
3.3 ConsensusAlgorithmsforDistributedTracking............................ 36
3.3.1 MathematicalFramework ........................................ 36
3.3.2 ExtendedKalman-ConsensusFilterforaSingleTarget ............... 37
3.3.3 JPDA-EKCFforTrackingMultipleTargets ......................... 40
3.3.4 HandoffinConsensusTrackingAlgorithms......................... 44
3.3.5 ExampleofDistributedTrackingusingEKCF ....................... 44
3.3.6 SparseNetworksandNaiveNodes-TheGeneralizedKalman
ConsensusFilter .................................................48
3.4 CameraNetworkCalibration............................................ 53
3.4.1 DistributedDataAssociation...................................... 53
3.4.2 DistributedCalibration........................................... 54
3.4.3 DistributedPoseEstimation ...................................... 56
3.5 Conclusions .......................................................... 57
4 ObjectandActivityRecognition .......................................... 59
4.1 ObjectRecognition .................................................... 59
4.1.1 ObjectRecognitionUnderResourceConstraints..................... 60
4.2 Time-DelayedCorrelationAnalysis ...................................... 61
4.2.1 SceneDecompositionandActivityRepresentation ................... 62
4.2.2 CrossCanonicalCorrelationAnalysis .............................. 63
4.2.3 Applications .................................................... 63
4.3 ActivityAnalysisUsingTopicModels .................................... 64
4.3.1 ProbabilisticModel .............................................. 65
4.3.2 LabelingTrajectoriesintoActivities ................................ 66
4.4 DistributedActivityRecognition......................................... 66
4.4.1 ConsensusforActivityRecognition ................................ 67
4.5 Conclusions .......................................................... 71
5 ActiveSensing .......................................................... 73
5.1 ProblemFormulation................................................... 73
5.1.1 ActiveSensingofDynamicalProcesses ............................. 74
5.2 ReviewofExistingApproaches .......................................... 77
5.3 CollaborativeSensinginDistributedCameraNetworks ..................... 78
5.3.1 SystemModeling................................................ 79
Description:As networks of video cameras are installed in many applications like security and surveillance, environmental monitoring, disaster response, and assisted living facilities, among others, image understanding in camera networks is becoming an important area of research and technology development. Ther