Capture ofVehicleSurroundingsUsing a Pairof Panoramic Annular Lens Cameras
Guoguang Yang · 2007
Inthispaperwe propose anewsystemtodetect obstacles roundavehicle forintelligent driver assistance system. Thesystemconsists ofapairofpanoramic annular lens(PAL) cameras, whichareinstalled on thetopofthevehicle. The panoramic annular lenscanproject theviewoftheentire 360 degrees aroundtheoptical axisontoan annular plane. Two cameraswhicharefixed properly cancapture allviewofthe surroundings ofthevehicle. Theyarecalibrated byusingthe spherical perspective projection modelto unwrappedthe conventional perspective imagewithoutdistortion. A front panoramic imageanda backpanoramic imageofthevehicle formaVehicle DynamicPanoramic Map(VDPM).A driver can watchthesceneinanydirection ofthevehicle through the VDPM, whichwillmakethedrivesafer. Threecoordinates systems areestablished inordertoobtain theobstacle position information fromthevehicle. We createa virtual binocular vision situation using thevideo streamfromeachonepanoramic camera.The binocular analysis technique witharea-based correlation isusedtorecognize theobstacle andcomputethe distance fromthevehicle. Theexperiment result showsthebasic feasibility ofthis system. Itisbelieved thatapanoramic annular lenscamerawillplayan important roleinthefutureof intelligent driver assistance systems. I.INTRODUCTION Almostonetenthofmillion persons diefromtraffic accidents every yearinChina, whichtakes thefirst place in theworld. Saferautomobile driveandsmoother traffic condition arestrongly demanded inthedevelopment ofthe cities. Overthepast three tofour decades, safety belts, airbags andsomeimproved design andstructure ofthevehicle have indeed saved countless lives. Butall these safe systems belong tothepassive safety. Unlike passive system, active safety can predict apossible accident orcollision andalarm adriver to takeaction toprevent thepossible accidents. Theactive safeties usually usesensors whichareultrasonic sensors, microwave detectors, thermal infrared imagers andvideo cameras. Amongallthesensors, video sensors attract more andmoreattention duetotheir lowcost, fast response, easy installation andmaintenance, andwideareamonitored. Considerable research (1)hasbeenperformed fordeveloping intelligent driver support systems using video sensors. The best intelligent drive assistance system based onvideo sensor needmonitor allsurroundings ofthevehicle. Somanyvideo sensors areneeded andthesystem will become complex. Luckily, wenowhaveapanoramic camera whichcangive apanoramic viewofthesurround ofthevehicle. Inthis paper, wedevelop apanoramic video system fortheintelligent driver assistance system using apair ofpanoramic annular lens (PAL) cameras. Thepanoramic annular lenswithsmartdesign structure issosmall that itcaneasily beinstalled onthecar andalmost notbefound. Twopanoramic annular lens cameras cancapture thefront panoramic scene, backpanoramic scene andpartial side scene withimagedistortion. We proposed a novel algorithm called spherical perspective projection model tocorrect this distortion. Thecorrection algorithm isdifferent conventional perspective algorithm whichisthebasic model ofcomputer vision. Besides webring theconcept ofVehicle Dynamic Panoramic Map(VDPM)whichiscomposed oftwo panoramic images. An obstacle detection algorithm isalso usedtodetect objects position information round thevehicle. A driver notonlycanwatch thevehicle surroundings inreal- timebutalso inquire about theobstacle position information through VDPM.Thesewillmakethedrive safer andmore intelligent. Theoutline ofthepaper isasfollows. Section 2describes theprinciple ofthepanoramic camera. Section 3brings and interprets theconcept VDPM.Section 4presents anobstacle detection algorithm andshowsthereal experiment results. We conclude andproposethefuture workwhichneeds improvements insection 5.