Making php see, confoo 2011
0 likes544 views
The document discusses introducing OpenCV, an open source computer vision library originally developed by Intel. It was designed with over 500 algorithms for real-time computer vision. The talk introduces some basic usage of OpenCV in PHP, including loading and saving images, image processing techniques like smoothing and morphology, edge detection algorithms, template matching to find objects in images, and using histograms for color matching. Examples of applications mentioned include object recognition, gesture tracking, and structure from motion.
![Thecode
<?php
use OpenCVImage as Image;
use OpenCVHistogram as Histogram;
echo "Loading samplen";
$i = Image::load("elephpant_sample.jpg",
Image::LOAD_IMAGE_COLOR);
$hsv = $i->convertColor(Image::RGB2HSV);
$planes = $hsv->split();
$hist = new Histogram(1, 32,
Histogram::TYPE_ARRAY);
$hist ->calc($planes[0]);
echo "Loading main imagen";
$i2 = Image::load("dragonbe_elephpants.jpg",
Image::LOAD_IMAGE_COLOR);
$hsv2 = $i2->convertColor(Image::RGB2HSV);
$planes2 = $hsv2->split();
$result = $planes2[0]->backProject($hist);
$result ->save("bp_output.jpg");](https://image.slidesharecdn.com/making-php-see-110311145834-phpapp01/85/Making-php-see-confoo-2011-67-320.jpg)
![Haarclassifier
<?php
use OpenCVImage as Image;
use OpenCVHistogram as Histogram;
$i = Image::load("sailing.jpg",
Image::LOAD_IMAGE_COLOR);
$result = $i->haarDetectObjects("
haarcascade_frontalface_default.xml");
foreach ($result as $r) {
$i->rectangle($r['x'], $r['y'],
$r['width'], $r['height']);
}
$i ->save("sailing_detected.jpg");](https://image.slidesharecdn.com/making-php-see-110311145834-phpapp01/85/Making-php-see-confoo-2011-88-320.jpg)
More Related Content
Similar to Making php see, confoo 2011 (20)
More from Bachkoutou Toutou (12)
Making php see, confoo 2011
- 1. MakingPHPSee Confoo2011 MichaelMaclean mgdm@php.net http://mgdm.net
- 4. IntroducingOpenCV “TheOpenComputerVisionLibraryhas>500 algorithms,documentationandsamplecodefor realtimecomputervision.”
- 5. IntroducingOpenCV OriginallydevelopedbyIntel,releasedasopen sourceundertheBSDlicence
- 6. IntroducingOpenCV http://opencv.org
- 8. Why? Iwantedtolearnaboutit,butusealanguageI wasmorecomfortablewith
- 11. Applications Gesturetracking(perhapsnotthatusefulin PHP ...)
- 12. Applications Structurefrommotion Creating3Dmodelsofobjectsfrom2Dinputs
- 13. Applications Structurefrommotion Again,maybenotinPHP,butoneday...
- 15. ABriefDisclaimer It'snotmyaimforthistalktogiveashort courseoncomputervision
- 18. ABriefDisclaimer Iwillnotexplainthetheorybehindmanythings (asImightnotunderstanditmyself)
- 19. ABriefDisclaimer I'mnotreallyamathsgeek,soifthereareany inaccuracies,pointthemout
- 20. Gettingit It'swillbeonGithubveryshortly http://github.com/mgdm
- 21. Gettingit You'llneedtheOpenCVlibrary,whichisinmost Linuxdistributionsthesedays YouneedPHP5.3
- 22. Gettingit I'venotdoneaWindowsbuildyet–it'sonthe TODO Patchesarewelcome ☺
- 23. Installing It'sthenormalPHPextensionbuildsystem phpize ./configure make make install ● Addextension=opencv.sotophp.ini
- 24. Basicusage EverythinginthelibraryisundertheOpenCV namespace Ifyou'renotfamiliarwithnamespaces,check http://php.net/namespaces
- 26. LoadingandSavingImages $image = OpenCVImage::load("test.jpg"); $image ->save("test2.jpg"); YounowhaveanImageobject Thesameimagehasbeensavedastest2.jpg Ifthathasworked,thelibraryissetupcorrectly. Congratulations☺
- 29. BasicsofOpenCV Thisincludesthingsliketransposition,adding, multiplying,addingscalars,etc
- 30. BasicsofOpenCV Thereisaverylargelistofthesebasicoperators NotallofwhichIunderstand...
- 31. ImageProcessing OpenCV(predictably)hasmany,manyfunctions thatdovariousthingstoimages Theserangefromthefairlymundanetothevery powerful
- 32. Smoothing Notveryexciting,butagooddemonstration–a Gaussianblur
- 33. Smoothing $dst = $image->smooth( OpenCVImage::GAUSSIAN, 31, 0, 0, 0); $dstwillnowcontainaslightlyfuzzyversion oftheimage
- 34. The input
- 35. The result
- 36. Smoothing,continued Whywouldyouwanttodothat? Perhapstoremovenoisefromanimagetakeninlow light
- 37. Smoothing,continued ● Variousmethodsareavailable,accessiblevia differentparameterstosmooth() ● Medianblur ● Gaussian ● Bilateralfiltering
- 39. ImageMorphology Theseareaclassofoperatorswhichconvolvean imagewithakernel Thekernelislikeamask,whichisscanned acrosseverypointintheimage Ithasananchorpoint,whichrepresentsthepixel beingcurrentlytransformed
- 40. Kernels 0 1 0 1 2 1 0 1 0 Forsomeoperations,eachnumber representstheamountof“influence”that pixelhasontheoutput
- 42. Dilation Thismaximumvaluebecomesthevalueofthe pixelatthatpositioninthenewimage Theupshotofthisis,brightareasgetlarger
- 43. Erosion Thisisfundamentallytheoppositeofdilation Insteadofthemaximumvalueunderthekernel, welookfortheminimum Brightareasgetsmaller
- 44. What'sthepoint? Brightareasmightbepartofthesameobject, butsplitintoseveralpartsintheimage (Shadows,obstructions)
- 45. What'sthepoint? Theseoperationswillcausetheseareastojoin up,makingthemeasiertoidentify
- 46. Edgedetection OpenCVfeaturesseveralalgorithmsforedge detection YoumightwellhaveseenthesebeforeinGIMPor Photoshop Youcanusethemtogetoutlinesofobjects
- 47. Sobel TheSobeloperatorisawaytogetaderivative ofanimage (Mathsgeeksmaypointoutthisistechnically incorrect–justrunwithitfornow)
- 48. Sobel Youendupwithanimagewhichhasbright areaswherethereislotsofcontrast
- 49. Theresult
- 50. Canny Amorecomplexalgorithm,whichtakes derivativesinxandy Somemoreprocessingresultsinafarclearer line
- 51. Theresult
- 52. Andnowforsomething completelydifferent moreinteresting
- 53. Templatematching YoucanuseOpenCVtotryandlocateanimage withinanotherimage You'dusuallydothiswithasmallimageofan object,andalargersearchimage
- 55. Myotherimage ThankstoMichaelangelovanDamfortheCreativeCommonspicture http://www.flickr.com/photos/dragonbe/3411273755/
- 56. Theresult
- 58. Theresult ThereasonitmatchesmostoftheElePHPantsis becausethesmallimageis,well,small,andhas beenresavedasaJPEG
- 59. Theresult Thus,thepixelsitissearchingwithareno longeridenticaltotheoriginals
- 60. Histograms Histogramsareawayoffittingvaluesintoslots, orbins,asOpenCVcallsthem Theyarefundamentallythesameasanarray, withacountofvalues,andcanbegraphedasa barchart
- 61. I'mnotgoingtodoanything quitethatdull
- 62. Histograms OpenCVuseshistogramstorecordasignature offeaturesinanimage
- 63. Histograms Forexample,youcanusethemtorecordthe coloursinasectionofanimage (Stayawakeattheback)
- 64. Backprojection IfIconvertthisimagetoHSVformat,Icangeta versionoftheimagewhichencodesthehueas onechannel
- 65. Backprojection ThenIcanconvertthisintoahistogram,toget ahashofsortsoftheproportionofcoloursin theimage Then,OpenCVwilltakethishistogramandlook foritinanotherimage
- 66. Anaside... Thisiswhatanimagelookslikewhenit'sconvertedtoHSVandsaved asifitwasstillRGB
- 67. Thecode <?php use OpenCVImage as Image; use OpenCVHistogram as Histogram; echo "Loading samplen"; $i = Image::load("elephpant_sample.jpg", Image::LOAD_IMAGE_COLOR); $hsv = $i->convertColor(Image::RGB2HSV); $planes = $hsv->split(); $hist = new Histogram(1, 32, Histogram::TYPE_ARRAY); $hist ->calc($planes[0]); echo "Loading main imagen"; $i2 = Image::load("dragonbe_elephpants.jpg", Image::LOAD_IMAGE_COLOR); $hsv2 = $i2->convertColor(Image::RGB2HSV); $planes2 = $hsv2->split(); $result = $planes2[0]->backProject($hist); $result ->save("bp_output.jpg");
- 68. Theresult
- 69. Theresult Asyoucansee,theresultisabinaryimage showingtheplaceswherethehistogram matchedthecoloursintheimage
- 70. Theresult Youcouldprobablygetabettermatchwitha bettersample
- 71. Theresult Youcanthengoonanddootherthingswiththe morphologyoperators,tofindtheextentsof eachmatch
- 72. Theresult Oryoucanusethatimageasamaskforother operations
- 73. Otherusesforhistograms Incombinationwiththeedgedetectionstuff fromearlier,youcanmakeahistogramofthe edgegradients
- 74. Otherusesforhistograms Thismeansthatyoucanthentryandmatch particularshapesusingahistogram Thisisusedbysomeforgesturerecognition
- 75. Machinelearning Inadditiontotheimageprocessingfunctions, OpenCVincludessomemachinelearning capability
- 77. Machinelearning Thealgorithmsaregeneral,andnotreally specifictocomputervision
- 78. Machinelearning Generally,thealgorithmsaretrainedusinga largedataset,andthentestedagainstanother
- 79. Machinelearning Therearetwomainwaysofimplementingthis ● Supervisedlearning,inwhichtheinputdatais “labelled” ● Unsupervisedlearning,wherethedatahasnolabels
- 80. Supervisedlearning Inthismethod,thealgorithmknowsthatthe featureitisbeingtrainedoniseitherpresentor notpresent
- 81. Supervisedlearning Itcanthenlearnthatcertaincharacteristicsare distinctiveamongthosewherethefeatureis present
- 82. Supervisedlearning Forexample,youcanhaveadatasetof10,000 images,inwhich8,000containfacesand2,000 donot
- 83. Unsupervisedlearning Thesealgorithmsarereferredtoasclustering algorithms Theyareusedtomatchupsimilarimages, withoutknowledgeofwhatiswithin
- 84. Training OpenCVcomeswithtoolsfortrainingthe variousalgorithmsitsupplies I'mnotgoingtoexplainthemhere
- 85. Haarclassifier Thisisafairlyspecializedimplementationofa learningalgorithm
- 86. Haarclassifier Itisusedtodetect“mostlyrigid”objects Likefaces!
- 88. Haarclassifier <?php use OpenCVImage as Image; use OpenCVHistogram as Histogram; $i = Image::load("sailing.jpg", Image::LOAD_IMAGE_COLOR); $result = $i->haarDetectObjects(" haarcascade_frontalface_default.xml"); foreach ($result as $r) { $i->rectangle($r['x'], $r['y'], $r['width'], $r['height']); } $i ->save("sailing_detected.jpg");
- 89. Theinputimage
- 90. Theresult
- 91. Otherfeatures Thereareotherinterestingfeaturesofthe librarythatI'venotimplementedyet,butthey willbetheresoon
- 92. Inpainting Repairingdamagetoimages–forexample ● strippingwatermarks ● fillingintextureaftertheimagehasbeenrotated Todothisproperly,youhavetocreateamask, andI'venotworkedoutareasonableAPIfor that
- 93. Contourdetection OpenCVcanuseitsedgedetectionalgorithmsto findthecontoursofanimage Thesearestoredinasequence,whichcanbe iteratedover
- 94. Imagecapture OpenCVcangrabimagesfromacamera Iwillnowattempttodemothis...
- 95. Thanksforlistening Pleaseratemytalk! http://joind.in/2830 Getintouch mgdm@php.net http://mgdm.net @mgdm