isiZulu
Amakhamera e-AI Ezixazululweni Zokubala Abahamba Ngezinyawo Ezihlakaniphile: Ngaphezu Kokubala, Kubheke Ekuthathweni Kwezinqumo Ezihlakaniphile
Kwadalwa ngo 01.31
Esikhathini samadolobha akhanyayo kanye nezinhlelo ezisebenzisa idatha, ukubalwa kwabantu abahamba ngezinyawo sekuphenduke umsebenzi olula wezibalo ube yisisekelo sokuphathwa okuhlakaniphile. Izindlela zendabuko—kusukela ekubhaleni ngesandla kuya kumasensori we-infrared—zihlale zibhekene nezinkinga zokunembile ezimeni eziyinkimbinkimbi, zishiya amadolobha, abathengisi, nezindawo zomphakathi zingazi ngama-patterns abalulekile okuphuma. Namuhla, amakhamera anokuhlakanipha kwe-AI aguqula le ndawo, aguqula ukubalwa kwabantu abahamba ngezinyawo kube umthombo wezinsiza ezisebenzisekayo, hhayi inqubo yokuhlanganisa idatha engasebenzi. Le ndatshana ihlola ukuthi amakhamera e-AI aguqula kanjani izixazululo zokubalwa kwabantu abahamba ngezinyawo, ukuphumelela kwawo kwezobuchwepheshe, izicelo zangempela ezimbonini, kanye nezinto ezibalulekile zokuphumelela kokufakwa.
Izithiyo Zokubalwa Kwabantu Abahamba Ngemikhondo: Kungani Amakhamera e-AI Engumsebenzi Obalulekile
Ngaphambi kokucubungula intuthuko ye-AI, kubalulekile ukuqonda amaphutha ezindlela zokubala abahamba ngezinyawo ezivamile eziye zabangela ushintsho ezixazululweni ezihlakaniphile. Ukubala ngesandla, nakuba kunembile, kulula ukwenza amaphutha omuntu—ikakhulukazi ezindaweni ezinabantu abaningi njengezindawo ezinhle noma izikhungo zezokuthutha ngesikhathi samaholide. Ngesikhathi samaholide kaMeyi 2025, iHuangshan Scenic Area yabika ukubala okungaphansi cishe ngo-20% ngabalibali abayisithupha ngesandla emnyango omkhulu, okwaholela ekubambeaneni kwendawo isikhathi esingu-1 ihora nabavakashi abangaphezu kuka-800 abangabalwanga. Izinzwa ze-infrared namatafula okucindezela, phakathi naleso sikhathi, ayikwazi ukuhlukanisa abahamba ngezinyawo nezinto ezingaphili, okubangela amaphutha amakhulu ezindaweni eziyinkimbinkimbi. Isitolo esikhulu iWumart eZhongguancun sabika izinga lamaphutha okubala elidlula u-30% ngesikhathi samaholide ekuseni ngenxa yokubonakaliswa kwezindawo zokugcina ezibandayo, okwaholela ekuphelelwa ubisi nesinkwa njalo.
Lezi zikhala azikona nje kuphela izinkinga—zinomthelela oqondile ebhizinisini nasekuphepheni komphakathi. Abathengisi baphuthelwa amathuba okwenza imali ngenxa yokungahambisani kwedatha yokuhamba kwabantu, amadolobha abhekana nezinkinga zokwenza ngcono ukuhamba kwemoto, futhi izindawo zomphakathi zibhekene nezingozi zokugcwala. Amakhamera e-AI abhekana nalezi zikhala ngokusebenzisa ukubona kwekhompyutha nokufunda okujulile ukuze afinyelele ukunemba okungakaze kubonwe nokwaziswa kokwakhiwa, aguqule idatha yokubala engashintshi ibe yinzuzo yesu.
Izithombe Zobuchwepheshe: Indlela Amakhamera e-AI Ahlinzeka Ngokubala Okunembile, Ngesikhathi Sangempela
Amandla ayisisekelo wamakhamera e-AI ekubhaleni abantu ahlale ekwazi ukujolisa ezindaweni ezihlukahlukene nezinselele ngokusebenzisa ama-algorithms athuthukile nokuhlanganiswa kwezinsiza. Ngokwehlukana nezinhlelo zendabuko, amakhamera e-AI awaboni nje kuphela—awaqonda isimo, ahlukanisa abantu abahamba ngezinyawo kwezinye izinto, alandele ukuhamba kwabantu ngabanye, futhi alungiselela izinguquko zomhlaba ezifana nezinguquko zokukhanya, ukuvimbela, nokutholwa kwezinto ezincane.
1. Ama-Algorithm Athuthukisiwe Okuthola Nokulandela Umkhondo
Izinhlelo zesimanje ze-AI zokubala abahamba ngezinyawo zisebenzisa ukuhlanganiswa kwamamodeli athuthukile kakhulu okuthola izinto kanye nezinqubo zokulandelela izinto eziningi. Imodeli yakamuva ye-YOLOv11, ngokwesibonelo, isibe yinto eshintsha imidlalo ngomklamo wayo olula kanye nokunemba okuthuthukisiwe. Ngokwamukela i-GhostNet njengomgogodla wayo, i-YOLOv11 inciphisa izinombolo zamapharamitha ngo-40% ngenkathi igcina ukunemba kokuthola okungaphezu kuka-90% (mAP@0.5), ivumela ukucubungula ngesikhathi sangempela kwevidiyo ye-1080p ku-50 amafreyimu ngomzuzwana (FPS) enezinga lamaphutha angalungile angaphansi kuka-3%. Lapho ihlanganiswa nenqubo yokulandelela ye-DeepSORT, ehlanganisa ukuhlunga kwe-Kalman ukubikezela ukunyakaza kanye namamodeli e-ReID (Re-identification) wokufanisa okusekelwe ekubukeni, lezi zinhlelo zixazulula ngempumelelo inkinga yokushintsha kwe-ID nokufihla ezixukwini eziningi.
Ama-Feature Pyramid Networks (FPN) athuthukisa ukusebenza ngokuhlanganisa izici eziphakeme zokuchaza kanye nezici ezingezansi zemininingwane, aqinisekisa ukutholwa okunembayo kwezihloko ezinkulu nezincane—okubalulekile ezimeni ezifana nezikwele ezinabantu abaningi noma imishayo emincane yezitolo. Ezivivinyweni zemigwaqo yasemadolobheni, izinhlelo zamakhamera e-AI ezisebenzisa lezi zibuchwepheshe zithole ukunemba okumaphakathi kokutholwa okungaphezu kuka-95%, kufinyelela ku-98% ezimweni ezinhle zasemini.
2. Ukubala Emngceleni (Edge Computing): Isivinini, Ubumfihlo, Nokuqina
Olunye uthuthuko olubalulekile ukuhlanganiswa kwe-edge computing, okucubungula idatha endaweni yekhamera noma idivayisi eseduze kunokuthembela kumaseva efwini. Lokhu kususa izinkinga ze-latency ezihlotshaniswa nokudluliswa kwamafu, kuqinisekisa imiphumela yokubala yesikhathi sangempela—ebalulekile ezinhlelweni ezidinga isikhathi esifushane njengokulawula izixuku noma ukuphathwa kwezimoto. I-Edge computing iphinde ixazulule izinkinga zobumfihlo ngokugcina idatha ebonakalayo ezindaweni ezithintekayo, kunciphisa ubungozi bokwephulwa kwedatha ngesikhathi sokudluliswa. Izixazululo zehadiwe njenge-NVIDIA Jetson Orin Nano (amandla okubala angu-40 TOPS) noma i-Intel Movidius Myriad X zivumela ukucubungula okusebenzayo endaweni, ngisho nasezindaweni ezinomkhawulo wezinsiza.
3. I-Hardware Nesoftware Ezivumelana Nemvelo
Amakhamera e-AI ahlome ngezici zehadiwe eziklanyelwe izimo eziyinselele, okuhlanganisa isinqumo esiphezulu, ububanzi obubanzi bokuguquguquka, nokuzwela ekukhanyeni okuphansi. Lezi zici ziqinisekisa ukuthwebula kwezithombe okucacile ezimeni ezisukela elangeni eliqhakazile kuya emigwaqweni yasebusuku noma esimweni sezulu esinegwebu. Ukuthuthukiswa kwesoftware njenge-adaptive histogram equalization (CLAHE) kuthuthukisa kakhulu ikhwalithi yesithombe ezindaweni ezinokukhanya okuphansi, kanti izindlela zokwandisa idatha zithuthukisa ukuqina kwemodeli ezimeni ezahlukene zokukhanya nezizinda.
Ngale Kokubala: Izicelo Zangempela Zokubala Abahamba Ngezinyawo Nge-AI Camera
Inani langempela lamakhamera e-AI ekubaleni abahamba ngezinyawo likumandla awo okukhiqiza imininingwane ewusizo kuzo zonke izimboni. Kusukela emadolobheni ahlakaniphile kuya ezitolo nasekuvikelekeni kwezimboni, lezi zixazululo zishayela ukusebenza kahle, zithuthukisa ukuphepha, futhi zithuthukisa izinto ezithinta abasebenzisi.
1. Amadolobha Ahlakaniphile: Ukuthuthukisa Ukuhamba Kwethrafikhi Nokuphepha Komphakathi
Emadolobheni, ukubalwa kwabantu abahamba ngezinyawo ngamakhamera e-AI kuyisisekelo sokuphathwa kwethrafikhi okuhlakaniphile. Ngokuhlaziya ukuhamba kwabantu ngesikhathi sangempela ezindaweni eziphambanayo, izindawo zokuwela, nezindawo zokuhamba zomphakathi, abaphathi bedolobha bangalungisa izikhathi zamalambu ethrafikhi ngendlela eguquguqukayo, kunciphisa ukuminyana nokuthuthukisa ukuphepha kwabahamba ngezinyawo. Ngokwesibonelo, iShanghai Hongqiao Metro Station isebenzisa idatha yekhamera ye-AI ukulungisa izikhathi zezitimela phakathi namahora aphezulu, yandisa umthamo wokuhamba ekuseni ngo-25%.
Izindawo ezinhle ezibukwayo nazo zithola izinzuzo ezinkulu kulezi zixazululo. Indawo yeziNtaba zaseHuangshan yafaka amakhamera e-AI ezindaweni ezibalulekile eziyi-12 ngesikhathi sehlobo sikaMeyi 2025, okuvumela ukubalwa kwabantu ngesikhathi sangempela ezindaweni ezithile. Lapho inani lezivakashi eXihai Grand Canyon lidlula izinkulungwane ezimbili, uhlelo lwazishukumisa ngokuzenzakalelayo ukuthi kuqale ukusakazwa kwezaziso zokukhipha izixuku, kunciphisa izikhalazo mayelana nokuminyana ngo-60%. Amakhamera omphakathi, lapho ahlanganiswa nezinhlobo ze-AI, nawo asebenza njengezinsiza ezibalulekile zedatha yocwaningo lwezokuthutha, ahlinzeka ngokubalwa okuthembekile kwabantu nezimoto ezindaweni ezikhanyisiwe kahle.
2. Ezokudayisa: Ukuthuthukisa Okuhlangenwe Nakho Kwamakhasimende Nokusebenza Kahle Kwezisebenzi
Kubathengisi, idatha eqondile yokubala abantu (footfall) ibalulekile ukuze kuthuthukiswe ukuqashwa kwezisebenzi, ukuphathwa kwezimpahla, namasu okumaketha. Amakhamera e-AI angaphezu kokubala nje abantu ukuze ahlaziye izindlela zokuziphatha zamakhasimende, njengokuchitha isikhathi emashalofini athile noma amazinga okuguquguquka kusuka ekubalweni kwabantu kuya ekuthengisweni. Amakhamera okubala e-AI akwa-Hikvision, isibonelo, avumela abathengisi ukuthi babeke imingcele yobude bemigqa, bavule izaziso lapho izikhathi zokulinda zidlula imingcele echazwe ngaphambili.
Isitolo se-Hema Fresh sikhishwe abasebenzi abathathu abasebenza isikhathi esigcwele ngokwamukela izixazululo zamakhamera e-AI, saveshe ngaphezu kuka-42,000 RMB ngonyaka ezindlekweni zabasebenzi. Ngokuhlaziya idatha yezinyathelo, isitolo silungise abasebenzi bamarejista ezimali, kunciphisa isikhathi sokulinda esiqongweni sasekuseni kusuka emizuzwini engu-18 saya kwengu-7. Ngaphezu kwalokho, ukuhlanganisa idatha yezinyathelo nedatha yokuthengisa kwavumela isitolo ukuthi sibeke izinto zokukhangisa ezindaweni ezinabantu abaningi, kwandisa inani lokuthenga elijwayelekile ngo-12%. Izinkampani zokuqina ezifana ne-Leke Fitness zisebenzisa amagrafu okugeleza kwamalungu avela kumakhamera e-AI ukuhlela izikhathi zokuqeqeshwa komuntu siqu ngesikhathi esiphakeme (7-9 PM), kwandisa izinga lokubhuka ngo-35%.
3. Ezokuphepha Emisebenzini Nasesikhungweni
Ezindaweni zezimboni, ukubalwa kwabantu ngamakhamera e-AI kusiza ukuqinisekisa ukuthobela imithetho yokuphepha ngokuqapha ukuminyana kwabasebenzi ezindaweni ezinomkhawulo. I-SF Express's Shenzhen Industrial Park yahlanganisa izinhlelo zokubala ze-AI nokulawula ukufinyelela, yenza kusebenze izexwayiso lapho inani labasebenzi emsebenzini lidlula umkhawulo wokuphepha (isibonelo, abantu abangu-30), kunciphisa ukwephulwa ngo-70%. Ngokufanayo, izimboni zikagesi e-Suzhou Industrial Park zisebenzisa amakhamera e-AI angangenisi nothuli futhi angavuthi ukubheka ukuhamba kwabasebenzi, ziqinisekisa ukuthobela izimiso zokuphepha ezindaweni eziyingozi.
Izinto Ezibalulekile: Ukuthobela Ubumfihlo Nokusebenzisa Ngokuziphatha
Njenganoma yimuphi ubuchwepheshe bokugadwa obusekelwa yi-AI, ukuthobela ubumfihlo kanye nezingqikithi zokuziphatha kubaluleke kakhulu ekusebenziseni ngempumelelo izixazululo zokubala abahamba ngezinyawo ngekhamera ye-AI. Ohulumeni kanye nemibutho elawulayo emhlabeni wonke baye bamisela imithetho eqinile yokuvikela idatha, okubandakanya i-GDPR ye-EU, i-CCPA yaseCalifornia, kanye ne-Personal Information Protection Law yaseChina.
Ukuqinisekisa ukuthobela, izinhlangano kumele zinamathele ezimisweni eziningana: umkhawulo wenhloso (ukuqoqa idatha ngezinhloso ezichaziwe, ezisemthethweni kuphela), ukunciphisa idatha (ukuqoqa idatha edingekayo kuphela), kanye nokucaca (ukwazisa umphakathi ngobukhona bekhamera nokusetshenziswa kwedatha). I-Edge computing idlala indima ebalulekile lapha ngokunika amandla ukucubungula kwedatha kudivayisi kanye nokungaziwa, kunciphisa isidingo sokudlulisa noma ukugcina ulwazi lomuntu siqu olubucayi. Izibuyekezo ezijwayelekile zokuthobela kanye nokubuyekezwa kokuziphatha nakho kubalulekile ukuqinisekisa ukuthi izinhlelo zisetshenziswa ngendlela efanele nangaphandle kokubandlulula.
Amathrendi Esikhathi Esizayo: Yini Okulandelayo Ekubalweni Kwabantu Nge-AI?
Ukuthuthukiswa kwamakhamera e-AI ekubalweni kwabantu akukapheli. Amathrendi amaningana asakhula athembisa ukuthuthukisa kakhulu amakhono abo futhi sandise izicelo zawo:
• Ukuhlanganiswa kwe-3D Perception Fusion: Ukuhlanganisa i-millimeter-wave radar noma amakhamera e-ToF (Time of Flight) nombono we-AI kuzovumela ukubala kwezikhala ze-3D, kuthuthukise ukunemba ezixukwini eziningi kakhulu nasezindaweni eziyinkimbinkimbi.
• Ukufunda Okuhlanganisiwe (Federated Learning): Le ndlela ivumela izinhlangano eziningi ukuthi ziqeqeshe amamodeli e-AI ngokubambisana ngaphandle kokwabelana ngedatha ebucayi, kuthuthukiswe ukujwayela kwemodeli ngenkathi kuvikelwa ubumfihlo.
• Ukucabanga Okubangela Izimbangela nge-GNN: Amagrafu e-Neural Networks (GNN) azovumela izinhlelo ukuthi zihlaziye izinhloso zokuhamba zabahamba ngezinyawo, zibikezele izindawo ezingase zibe nenkinga noma izingozi zokuphepha ngaphambi kokuba zenzeke.
• Ama-Chip we-ASIC Angokwezifiso: Ama-chip akhethekile e-AI enzelwe izindlela zokubala abahamba ngezinyawo (isibonelo, i-YOLOv11-DeepSORT) azonciphisa izindleko zehadiwe futhi athuthukise ukusebenza kahle kwamandla, okwenza kube lula ukusetshenziswa kabanzi.
Isiphetho: Amakhamera e-AI Njengengxenye Ebalulekile Yokuphathwa Okuhlakaniphile Kwabahamba Ngezinyawo
Amakhamera e-AI ashintshe ukubalwa kwezihambi kusukela emsebenzini okhathazayo nowonakalayo kwaba ithuluzi elinamandla lokwenza izinqumo ezihlakaniphile. Amandla abo okuletha ukunemba okuphezulu ezindaweni eziyinkimbinkimbi, ukukhiqiza imininingwane ewusizo kuzo zonke izimboni, kanye nokuqinisekisa ukuthobela imithetho yobumfihlo kubenza bangadingeki enkathini yamadolobha ahlakaniphile kanye nemisebenzi esekelwe kudatha. Njengoba ubuchwepheshe buqhubeka nokuthuthuka—ngokuthuthuka ekubambeni izinto nge-3D, ukufunda okuhlanganisiwe, kanye nezingxenyekazi zekhompyutha ezikhethekile—izixazululo zokubala izihambi ze-AI zizoba namandla amaningi futhi zibe nomthelela omkhulu.
Ema-nhlanganweni afuna ukwenza imisebenzi ihambe kahle, athuthukise ukuphepha, noma athuthukise okuhlangenwe nakho kwamakhasimende, ukutshala imali kwezixazululo zokubala abahamba ngezinyawo nge-AI camera akuseyona inketho kodwa kuyisidingo. Ngokugxila kwezobuchwepheshe, ukuzivumelanisa nezimboni ezithile, kanye nokuhambisana nezimiso zokuziphatha, lezi zixazululo zizoqhubeka nokushayela inqubekela phambili ekuphathweni okuhlakaniphile iminyaka ezayo.
Uxhumane
Sicela uxhumane nathi uhambele
WhatsApp
WeChat