Sunteți pe pagina 1din 12

AboutSRIN

Introduction
ProblemStatement
Content Methodology
Overview TheViolaJonesAlgorithm
Results&ErrorAnalysis
ConcludingRemarks
References
Samsung group is a South Korean MnC with headquarters in Seoul.
It was founded by Lee Byungchul in 1938 as a trading company.
Since 1990, Samsung has increasingly globalized its activities in the
electronics industry.
AboutSRIN Samsung R&D Institute India, Noida (SRI Noida) is one of the 29 global
Research and Development centers of Samsung Electronics and was
established in 2007.
The Company is focused on embedded and mobile software testing and
development.
Computervision modelingandreplicatinghumanvision
usingcomputersoftwareandhardware.
RecognitionProblem determiningwhetherornotthe
imagedatacontains somespecificobject,feature,or
activity
Introduction
Varietiesofrecognitionproblem
Identification identificationofaspecific vehicle
ClassificationGoogleGoggles
Detection detectionofpossible abnormalcellsortissuesin
medicalimages
Givenaninputimage,determine ifthereareobjectsof
Problem agivenclassintheimageandwheretheyarelocated.
Statement Theclassofobjectstobedetected isthefaceofwild
animals.
TRAININGSTAGE
OpenCVutilitiesused:

1.create_samples(usedto
generatedatasetand
FeatureVector
extractfeatures)
Methodology Sample
Creation Feature
Classifier
Training
Extraction 2.opencv_traincascade
Training Cascade (usedforclassifier
Dataset Classifier training)
TESTINGSTAGE OpenCVUtilitiesUsed:

detectMultiScale() returns
boundingboxeswhere
objecthasbeendetected

BoundingBoxes compareHist() returnsa


Test Intensity
Methodology Dataset Classifier Histogram
Matching
valuebetween0and1(one
meaningexactlysame
histograms)
InputImage Classifier Output
Metricusedtodefine
similarityisBhattacharyya
distance:
ThisObjectdetectionframeworkwasproposedin2001
byPaulViolaandMichaelJones.
Cascade Classifier
Training Itwasprimarilymotivated to providecompetitive
"TheViolaJones objectdetection ratesinrealtime.
Algorithm"
Trainingmightbeslow,butdetectionisfast.
Afewpropertiescommontohumanfaces:
Theeyeregionisdarkerthantheuppercheeks.
Thenosebridgeregionisbrighterthantheeyes.

TheViola
Jones Algorithm
Rectangular
Features

Source: (Viola, Paul and Michael J. Jones, "Rapid Object Detection using a Boosted Cascade of Simple
Features", Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2001.
Volume: 1, pp.511518)
Rectanglefeatures:
Value = (pixels in black area) (pixels in white area)
Threetypes:two,three,fourrectangles,Viola&Jonesused
tworectanglefeatures
Inthefigure,AandBaretworectanglefeatures,Cisthree
rectangleandDfourrectanglefeature.
TheViola
Jones Algorithm
Rectangular
Features

Source: (Viola, Paul and Michael J. Jones, "Rapid Object Detection using a Boosted Cascade of Simple
Features", Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern
Recognition, 2001. Volume: 1, pp.511518)
Astandardimagesubwindowofsize24X24maycontainalargenumber
ofrectanglefeatures.
ViolaandJonessuggestedthataverysmallnumberofthesefeatures can
TheViola becombinedtoformaneffectiveclassifier.
Jones Algorithm Theoverallformofthedetectionprocessis thatofadegeneratedecision
ClassifierLearning tree.

andDetection Apositiveresultfromthefirstclassifiertriggerstheevaluationofa
secondclassifier,andsoon.
Anegativeoutcomeatanypointleadsto theimmediaterejectionofthe
subwindow.
TheViola
Jones Algorithm
Schematic of
DetectionCascade

Source: (Viola, Paul and Michael J. Jones, "Rapid Object Detection using a Boosted Cascade of Simple
Features", Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern
Recognition, 2001. Volume: 1, pp.511518)
[1]Viola,PaulandMichaelJ.Jones,"RapidObjectDetectionusinga
BoostedCascadeofSimpleFeatures",Proceedingsofthe2001IEEE
ComputerSocietyConferenceonComputerVisionandPatternRecognition,
2001.Volume:1,pp.511518.

References [2]Opensource ComputerVision(OpenCV)HaarCascadeDocumentation


(src: http://docs.opencv.org/2.4/doc/tutorials/objdetect/cascade_classifier/c
ascade_classifier.html)

[3]EclipseIDEDocumentation
(src:http://www.eclipse.org/documentation/)

S-ar putea să vă placă și