I-Biometrics yizilinganiso zomzimba kanye nezibalo ezihlobene nezici zomuntu. Ukuqinisekiswa kwe-biometric (noma ukuqinisekiswa okungokoqobo) kusetshenziswa kwisayensi yekhompyutha njengendlela yokuhlonza kanye nokulawula ukufinyelela. Iphinde isetshenziswe ukuhlonza abantu emaqenjini aqashwe.

Izihlonzi ze-biometric ziyizici ezihlukile nezilinganiswayo ezisetshenziselwa ukumaka nokuchaza abantu ngabanye. Izihlonzi ze-biometric zivame ukuhlukaniswa njengezimpawu zomzimba ezihlobene nesimo somzimba. Izibonelo zifaka phakathi, kodwa azigcini nje ngeminwe, imithambo yesundu, ukuqashelwa kobuso, i-DNA, ukunyatheliswa kwesundu, i-geometry yesandla, ukuqashelwa kwe-iris, i-retina, kanye nephunga/iphunga.

Ubuchwepheshe bokuhlonza i-biometric buhilela isayensi yamakhompyutha, i-optics kanye ne-acoustics kanye nezinye isayensi yemvelo, isayensi yezinto eziphilayo, izimiso ze-biosensors kanye nezibalo ze-bio, ubuchwepheshe bokuphepha, kanye nobuchwepheshe bobuhlakani bokwenziwa kanye nezinye isayensi eziningi eziyisisekelo kanye nobuchwepheshe bokusebenzisa obusha. Kuyizixazululo zobuchwepheshe eziphelele zemikhakha eminingi.

Eminyakeni yamuva nje, ngokuthuthukiswa kobuhlakani bokwenziwa, ubuchwepheshe bokuhlonza i-biometric buye bakhula kakhulu. Njengamanje, ubuchwepheshe bokuhlonza ubuso buyindlela emele kakhulu i-biometrics.

Ukuqashelwa kobuso

Inqubo yokubona ubuso ihlanganisa ukuqoqwa kobuso, ukutholwa kobuso, ukukhishwa kwezici zobuso kanye nokubonwa kokufaniswa kobuso. Inqubo yokubona ubuso isebenzisa ubuchwepheshe obuhlukahlukene njenge-algorithm ye-AdaBoos, inethiwekhi ye-convolutional neural kanye nomshini we-vector wokusekela ekufundeni komshini.

Inqubo yokubona ubuso

Njengamanje, ubunzima bokubona ubuso bendabuko okuhlanganisa ukuzungeza ubuso, ukuvaleka, ukufana, njll. kuthuthukiswe kakhulu, okuthuthukisa kakhulu ukunemba kokubonwa kobuso. Ubuso be-2D, ubuso be-3D, ubuso obunama-spectral amaningi Imodi ngayinye inezimo ezihlukile zokuzivumelanisa nokutholwa, izinga lokuphepha kwedatha kanye nokuzwela kobumfihlo, njll., kanye nokwengezwa kokufunda okujulile kwedatha enkulu kwenza i-algorithm yokubona ubuso be-3D ingeze amaphutha okubikezela kwe-2D, Ingakhomba ngokushesha ubunikazi bomuntu, okulethe impumelelo ethile ekusetshenzisweni kokubonwa kobuso obunezinhlangothi ezimbili.

Ngesikhathi esifanayo, ubuchwepheshe bokuthola i-biometric okwamanje busetshenziswa njengobuchwepheshe obuyisihluthulelo sokuthuthukisa ukuphepha kokubonwa kobuso, okungamelana ngempumelelo nokukhwabanisa kokwenziwa mbumbulu njengezithombe, amavidiyo, amamodeli e-3D, kanye nama-prosthetic masks, futhi kunqume ngokuzimela ukuthi bangobani abasebenzisi abasebenzayo. Njengamanje, ngentuthuko esheshayo yobuchwepheshe bokubonwa kobuso, izinhlelo zokusebenza eziningi ezintsha ezifana namadivayisi ahlakaniphile, ezezimali eziku-inthanethi, kanye nokukhokha ubuso kuye kwaba yinto ethandwa kakhulu, okuletha isivinini kanye nokulula empilweni nasemsebenzini wawo wonke umuntu.

Ukuqashelwa kwesandla

Ukuqashelwa kwe-Palmprint uhlobo olusha lobuchwepheshe bokuqashelwa kwe-biometric, olusebenzisa i-palmprint yomzimba womuntu njengesici esiqondiwe, futhi luqoqa ulwazi lwezinto eziphilayo ngokusebenzisa ubuchwepheshe bokuthwebula izithombe ze-multispectral. Ukuqashelwa kwe-palmprint ye-multi-spectral kungabhekwa njengemodeli yobuchwepheshe bokuqashelwa kwe-biometric obuhlanganisa izici ze-multi-modality kanye nezici eziningi eziqondiwe. Lobu buchwepheshe obusha buhlanganisa izici ezintathu ezibonakala ze-skin spectrum, i-palmprint kanye ne-vein veins ukuze kunikezwe ulwazi oluningi ngesikhathi esisodwa futhi kwandiswe ukuhlukaniswa kwezici eziqondiwe.

Kulo nyaka, ubuchwepheshe bokuqaphela izintende zesandla ze-Amazon, obubizwa ngekhodi i-Orville, buqalile ukuhlola. Isithwebuli siqala ngokuthola isethi yezithombe zokuqala ezifakwe i-infrared polarized, sigxile ezicini zangaphandle zesundu, njengemigqa nama-fold; lapho sithola isethi yesibili yezithombe ezifakwe i-polarized futhi, sigxila esakhiweni sesundu kanye nezici zangaphakathi, njengemithambo yegazi, amathambo, izicubu ezithambile, njll. Izithombe ezingavuthiwe zicutshungulwa ekuqaleni ukuze zinikeze isethi yezithombe eziqukethe izandla. Lezi zithombe zikhanya kahle, zigxile, futhi zibonisa intende ngendlela ethile, ngendlela ethile, futhi zibhalwe njengesandla sobunxele noma sokudla.

Njengamanje, ubuchwepheshe bokuqashelwa kwesandla se-Amazon bungaqinisekisa ubunikazi bomuntu siqu kanye nokukhokha okuphelele ngama-millisecond angu-300 kuphela, futhi abudingi abasebenzisi ukuthi babeke izandla zabo kudivayisi yokuskena, bamane baphakamise isandla bese beskena ngaphandle kokuthintana. Izinga lokuhluleka kwalolu buchwepheshe licishe libe ngu-0.0001%. Ngesikhathi esifanayo, ukuqashelwa kwesandla sesandla kuwukuqinisekiswa okuphindwe kabili esigabeni sokuqala - okokuqala ukuthola izici zangaphandle, kanye nesikhathi sesibili ukuthola izici zangaphakathi zenhlangano. Uma kuqhathaniswa nezinye ubuchwepheshe be-biometric maqondana nokuphepha, kuthuthukisiwe.

Ngaphezu kwezici ze-biometric ezingenhla, ubuchwepheshe bokuqashelwa kwe-iris buyathandwa. Izinga lokuqashelwa okungamanga lokuqashelwa kwe-iris liphansi njengo-1/1000000. Isebenzisa kakhulu izici zokuguquguquka kokuphila kwe-iris kanye nomehluko ukuhlonza ubuwena.

Njengamanje, ukuvumelana embonini ukuthi ukuqashelwa kwendlela eyodwa kunezingqinamba kokubili ekusebenzeni kokuqashelwa kanye nokuphepha, futhi ukuhlanganiswa kwezindlela eziningi kuyindlela ebalulekile yokuqashelwa kobuso ngisho nokuqashelwa kwe-biometric—hhayi kuphela ngokusebenzisa izindlela eziningi. Indlela yokuthuthukisa ukunemba kokuqashelwa ingathuthukisa nokuzivumelanisa nezimo kwendawo kanye nokuphepha kobumfihlo bobuchwepheshe be-biometric ngezinga elithile. Uma kuqhathaniswa ne-algorithm yendabuko yemodi eyodwa, ingahlangabezana kangcono nezinga lokuqashelwa okungamanga kwezinga lezezimali (eliphansi njengeyodwa kwezigidi eziyishumi), okuyindlela eyinhloko yokuthuthukiswa kokuqashelwa kwe-biometric.

Uhlelo lwe-biometric olunezinhlobo eziningi

Izinhlelo ze-biometric ze-multimodal zisebenzisa izinzwa eziningi noma ama-biometric ukuze zinqobe ukulinganiselwa kwezinhlelo ze-biometric ze-unimodal. Isibonelo, izinhlelo zokuqaphela i-iris zingaphazamiseka ngenxa yama-iris agugayo kanti ukuqaphela iminwe ye-elekthronikhi kungenziwa kubi kakhulu ngenxa yeminwe egugile noma esikiwe. Ngenkathi izinhlelo ze-biometric ze-unimodal zinqunyelwe ubuqotho besihlonzi sazo, akunakwenzeka ukuthi izinhlelo eziningana ze-unimodal zizoba nemikhawulo efanayo. Izinhlelo ze-biometric ze-multimodal zingathola amasethi olwazi oluvela kumaka afanayo (okungukuthi, izithombe eziningi ze-iris, noma ukuskena komunwe ofanayo) noma ulwazi oluvela kuma-biometric ahlukene (okudinga ukuskena kweminwe ye-fingerprint kanye, ngokusebenzisa ukuqaphela izwi, iphasikhodi ekhulunywayo).

Izinhlelo ze-biometric ze-multimodal zingahlanganisa lezi zinhlelo ze-unimodal ngokulandelana, ngasikhathi sinye, inhlanganisela yazo, noma ngochungechunge, okubhekisela ezindleleni zokuhlanganisa ezilandelanayo, ezihambisanayo, ezilandelanayo kanye nezilandelanayo, ngokulandelana.

I-CHANCCTVusungule uchungechunge lweamalensi e-biometricukuqashelwa kobuso, ukuqashelwa kwesandla kanye nokuqashelwa kweminwe kanye nokuqashelwa kwe-iris. Isibonelo, i-CH3659A ilensi yokuphambuka ephansi engu-4k eyenzelwe izinzwa ezingu-1/1.8''. Iqukethe yonke imiklamo yengilazi kanye neyincane ene-TTL engu-11.95mm kuphela. Ibamba insimu yokubuka evundlile engama-degree angu-44. Le lensi ilungele ukuqashelwa kwesandla.