isiZulu
I-Camera Vision Ezimotweni Zokuhlanza Imigwaqo Ngamarobhothi: Iqhawe Elingazange Lizwakale Lokuhlanzeka Kwamadolobha Ahlakaniphile
Kwadalwa ngo 01.29
Njengoba ukuphuma kwelanga kubonakala emadolobheni esimanje, kuvela isizukulwane esisha sabahlanza imigwaqo—esithule, esisebenza kahle, futhi esihlome "ngamehlo" abona indawo yedolobha ngokunemba okungakaze kubonwe. Izimoto zokuhlanza imigwaqo ezisebenza ngomshini, ezake zaba umqondo wekusasa, manje sezijwayelekile ezinhlelweni zedolobha elihlakaniphile emhlabeni wonke. Inhliziyo yenguquko yazo itholakala kubuchwepheshe bokubona ngekhamera, ingxenye ekhule yaba yisenzisi esingeziwe yaba "injini eyinhloko yokwenza izinqumo" ezuza ukusebenza kahle, ukuphepha, nokusimama. Ngokungafani nezinhlelo ze-LiDAR ezibizayo ezibusa izingxoxo ekuhambeni okuzimele,ubuchwepheshe bekhameraukuthuthukisa kahle ukuhlinzeka ngempilo yomphakathi emadolobheni ngokunikeza isixazululo esingabizi, esinembile esihambisana nezinselelo ezihlukile zokuhlanza imigwaqo. Le ndatshana ihlola ukuthi kanjani ukubona ngekhamera okuhamba phambili kushintsha ukuhamba kwemigwaqo kwezimoto, kuhlukanisa izinguquko zayo zobuchwepheshe, umthelela wempilo yangempela, kanye nekusasa lobuchwepheshe obubalulekile bezixeko ezihlakaniphile.
Izinselelo Ezingavamile Zokuhlanza Imigwaqo: Kungani Ukubona Ngekhamera Kungabalulekile
Ukuhlanza imigwaqo yedolobha kunzima kakhulu kunezicelo eziningi ezizenzakalelayo. Ngokungafani nezindawo zomgwaqo ezilawulwayo noma izindawo zezimboni ezivalekile, imigwaqo yedolobha iyimvelo eguquguqukayo yezithiyo ezingalindelekile, izimo ezihlukahlukene zendawo, kanye nezinguquko ezihlala zikhona zemvelo. Isihlanzi serobhothi kumele sihambe ezindaweni eziwumngcingo, sithole imfucumfucu emincane njengezintwala zikagwayi kanye namacashana okudla, sigweme abahamba ngezinyawo nabagibeli bamabhayisikili, futhi sizivumelanise nokukhanya okushintshayo—kusukela elangeni eliqinile emini kuya ebusuku obumnyama kanye nobusuku bemvula. Izinobhothi zokuhlanza zakudala zazisebenzisa izinzwa eziyisisekelo noma imizila eyenziwe ngaphambili, okwaholela ekungasebenzi kahle: imfucumfucu eyenqanyuliwe, imizila engadingekile, kanye nokuphazamiseka okuvamile komuntu.
Umbono wekhamera ubhekana nalezi zinkinga ngokulingisa nokudlula amakhono okubona omuntu. Izinhlelo zesimanje zisebenzisa amakhamera e-high-definition (HD) kanye ne-RGB-D ukuthatha idatha ecebile yokubona, okuvumela amarobhothi ukuthi angagcini nje ngokubona kodwa futhi aqonde izindawo ezizungezile. Ngokwesibonelo, i-array yekhamera eyodwa ingakwazi ukuhlukanisa phakathi kweshidi (elidinga ukuswetshwa) nedwala elincane (elingadinga ukugwenywa), ihlukanise izinhlobo zodoti ukuze kuhlanzwe ngokuqondile, futhi izenzele imephu yezindawo ezinodoti omningi ukuze kuhlelwe umzila kahle. Leli zinga lokuqonda umongo alikwazi ukwenziwa ngezimbangi eziyisisekelo futhi libiza kakhulu ukukala nge-LiDAR kuphela—kwenza umbono wekhamera ube yisisekelo esihle sokuhlanza emigwaqweni okufinyelelekayo, nokusebenza kahle.
Izithuthukisi Zobuchwepheshe: Indlela Izinhlelo Zokubona Ngekhamera Eziphucukayo Ngokuvumelana Nokuhlanza Imigwaqo
Ukusebenza kahle kokubona ngekhamera ekuhlanzeni imigwaqo kwezimoto kuvela ezintweni ezintathu ezibalulekile zokuthuthukiswa kobuchwepheshe: ukuhlanganiswa kwezinsiza eziningi, ama-algorithms e-AI alula, kanye nokucubungula okusheshayo okuzivumelanayo. Ndawonye, lezi zinguquko zishintshe ukubona ngekhamera kusuka kuthuluzi lokuthwebula elilula kube uhlelo lokwenza izinqumo oluqinile, oluzimele.
1. Ukuhlanganiswa Kwe-Multi-Modal: Ukuhlanganisa Amakhamera Nezinsiza Eziphumelelayo
Nakuba amakhamera ehlukanisa ngokuthwebula imininingwane ebonakalayo nolwazi lombala, asebenza kangcono lapho edidiyelwe nezinye izinzwa ezishibhile ohlelweni olunezindlela eziningi. Izihlanzi zemigwaqo zesimanje ezisebenzisa amarobhothi zihlanganisa amakhamera e-HD nezinzwa ze-ultrasonic kanye namayunithi okulinganisa okungaguquki (ama-IMU) ukuze zinqobe imikhawulo yendawo. Ngokwesibonelo, emvuleni enkulu noma enkungwini—izimo ezilimaza ikhwalithi yesithombe yekhamera—izinzwa ze-ultrasonic zinikeza idatha yebanga ukuze kugwenywe izithiyo, kuyilapho amakhamera eqhubeka nokuthola imfucumfucu enkulu. Lokhu kuhlanganiswa kuqinisekisa ukwethembeka kuzo zonke izimo zezulu, okuyisidingo esibalulekile sokusebenza kwezinsizakalo zasemadolobheni amahora angu-24 ngosuku.
Isibonelo esivelele yizimoto zokuhlanza ezizimele ezisetshenziswa eSuzhou Industrial Park, ezisebenzisa amakhamera angu-8 e-HD ahambisana namayunithi angu-5 e-LiDAR (ukuze kutholwe indawo enembayo kakhulu) ukuze kufezeke ukuhlanza onqenqemeni ngamasentimitha eduze kwemiphetho. Amakhamera agxila ekutholeni imfucumfucu nasekulandeleni izihambi, kuyilapho i-LiDAR iphethe ukuthola indawo—kwakha uhlelo olulinganiselayo olwenza izindleko nokusebenza kube kuhle. Emadolobheni amancane, izindlela ezonga kakhulu zisebenzisa amakhamera ahambisana nohlelo lokubeka indawo i-Beidou yaseChina ukuze kufezeke ukunemba okufanayo ngaphandle kwezindleko eziphelele ze-LiDAR.
2. Izindlela Zokubala Ezilula Ze-AI: Ukunika Amandla Ukwenziwa Kwezinqumo Ngokweqiniso Kumadivayisi Angaphandle
Inkinga enkulu kakhulu yokubona ngekhamera ekuhlanzeni okwenziwa imishini yayingamandla okubala. Izinhlelo zakuqala zazisebenzisa ukucubungula okusekelwe emafini, okuholele ekubambezelekeni okwenze ukwenziwa kwezinqumo ngesikhathi sangempela kungakwenzeki. Namuhla, izindlela zokubala ze-AI ezilula—eziklanyelwe amadivayisi angaphandle—zivumela amakhamera ukuthi acubungule idatha ebonakalayo endaweni, okunikeza imininingwane esheshayo.
Izixazululo ezihamba phambili zisebenzisa izinguqulo eziguquliwe ze-algorithm ye-YOLO (You Only Look Once), njenge-YOLOv8 elula, ehambisa isivinini nokunemba ekutholeni imfucumfucu. Lezi algorithm ziqeqeshwa kumadatha amakhulu emfucumfucu yasemadolobheni—okuhlanganisa amabhodlela epulasitiki, izicucu zamaphepha, namachaphaza kawoyela—ngaphansi kwezimo ezahlukene zokukhanya nesimo sezulu. Ukuze kuthuthukiswe ukusebenza kakhudlwana, abathuthukisi bahlanganisa izindlela zokunakwa eziqondisa ukugxila kwe-algorithm ezindaweni ezibaluleke kakhulu, njengezindlela zemoto namaphasishi lapho imfucumfucu iqoqana kakhulu. Umphumela: ukunemba kokuqashelwa kudlula u-95% wezinhlobo ezivamile zemfucumfucu, nezinga lamaphutha angamanga elingaphansi kuka-5%—umkhawulo osusa izitobhi zokuhlanza ezingadingekile futhi unciphise ukuchitha amandla.
Olunye ubuchule ukufunda ukudlulisa, okuvumela ama-algorithm ukuba azivumelanise nezindawo ezintsha ngokushesha. Irubhoti efakwe edolobheni elisogwini lolwandle (lapho isihlabathi nolwandle oluluhlaza luvamile ukuba yimfucumfucu) ingalungisa imodeli yayo isebenzisa idatha yendawo ngaphandle kokuqeqeshwa kabusha ngokuphelele, okwenza izinhlelo zokubona ngekhamera zikwazi ukukala kakhulu ezindaweni ezahlukene zasemadolobheni.
3. Ukucubungula Okuzivumelanisayo: Ukunqoba Ukukhanya Nokuhlukahluka Kwemvelo
Ushintsho lokukhanya luyisitha sezinhlelo ezisebenzisa amakhamera, kodwa intuthuko yakamuva ekucubunguleni okuguquguqukayo ixazulule le nkinga. Izinhlelo zanamuhla zokubona ngekhamera zisebenzisa ukuhlaziywa kwe-spectral yeziteshi eziyi-16 ukuthola izimo zokukhanya zesikhathi sangempela—kusukela ekukhanyeni okunamandla kwelanga emini kuya ekukhanyeni okubuthakathaka kwemigwaqo—futhi zilungise amapharamitha esithombe ngokushesha. Ngokwesibonelo, ezindaweni ezinokukhanya okuphansi, uhlelo lwandisa isikhathi sokuchayeka futhi lusebenzisa ama-algorithm okunciphisa umsindo ukuze kugcinwe ukucaca kwesithombe, kuqinisekisa ukuthi ukutholwa kwemfucumfucu kuhlala kunembile ngemva kokuba sekumnyama. Lokhu kuguquguquka kubalulekile emadolobheni afana neHangzhou, lapho izihlanzi ezisebenzisa amarobhothi zisebenza amahora angama-24 ngosuku, zishintsha phakathi kokukhanya kwasekuseni, ilanga lasemini, nokukhanya kwemigwaqo ebusuku ngaphandle kwezinkinga.
Ukucubungula okwenziwe ngokwezifiso kuphinde kubhekane nokuphazamiseka kwangemuva, njengemibala ehlukahlukene yephavini noma amaphethini. Ngokusebenzisa izindlela zokukhipha okungemuva, uhlelo luhlukanisa izinto ezihambayo noma ezingajwayelekile (izinsalela, abantu) ezivela emvelweni ezinzile, kuqinisekisa ukusebenza okuqhubekayo phezu kwezindawo ze-concrete, asphalt, kanye nezitini.
Umthelela Wangempela: Umbono weKhamera Esemsebenzini Emadolobheni AseMhlabeni
Ukuqhubeka kobuchwepheshe embonini yekhamera kuguqulela ekuthuthukisweni okubonakalayo ekuhlanzweni kwamadolobha. Kusuka eShenzhen kuya eSuzhou, amadolobha asebenzisa izihlanzi zemigwaqo ezisebenzisa ubuchwepheshe bekhamera, zithola izinzuzo ezibalulekile ekusebenzeni kahle, ukonga izindleko, kanye nokuphepha kwezisebenzi.
Esifundeni sasePingshan eShenzhen—indawo yokuqala egcwele yobuchwepheshe be-AI yokuhlanza edumile eChina—izihlanzi eziyi-robotic ezinamakhamera angama-59 zenza umsebenzi wokuhlanza imigwaqo amahora angama-24 ngosuku, okunciphisa izidingo zabasebenzi abangaba-60%. Amakhamera enza ukuthi kube nokukhomba okunembayo udoti, ngakho-ke ama-robotic aqala amabhulashi awo okuhlanza kuphela lapho kudoti utholakele—kunciphisa ukusetshenziswa kwamandla ngo-30% uma kuqhathaniswa nezihlanzi ezijwayelekile ezisebenza njalo. Kwenye indawo yokuhlola, uhlelo lunciphise isilinganiso sokudlula kukadoti ngo-70%, abahlali bekhala ngokuthuthuka okubonakalayo ekuhlanzweni kwemigwaqo phakathi nenyanga yokuqala yokusetshenziswa.
Izihlanzi ezisebenza ngobuchwepheshe zaseSuzhou, ezihlome ngamakhamera angu-8 e-HD, zibonisa amandla e-camera vision ekuhlanzeni imiphetho—inselelo eqhubekayo emisebenzini eyenziwa ngesandla. Amakhamera athola imiphetho ngokunemba okungaphansi kwe-centimeter, avumela ama-robot ukuthi ahambe ngaphakathi kuka-3-5 cm ukusuka emaphethelweni futhi athwebule udoti ezindaweni ezincane zezitini abahlanzi abantu abangazitholi njalo. Lawa ma-robot abhalise ngaphezu kwamakhilomitha angu-2,000 okuphatha ngokuphepha, ngaphandle kokushayisana ngokuqondile ngenxa yokutholwa kwezihambi nezimoto ngesikhathi sangempela ngama-camera arrays abo.
E-Hangzhou, isihlanzi samalwesi esenziwe ngobuchwepheshe esithi "Blue Fatty" S330 sisebenzisa ubuhlakani bokubona obuthuthukisiwe ukubhekana nezindawo eziyinkimbinkimbi zasemadolobheni njengeWulin Square. Amakhamera aso aqaphela izibani zomgwaqo, izigwamu zomgwaqo, kanye nezithiyo zesikhashana ezifana nezithiyo zokwakha, kulungise indlela ngesikhathi sangempela. Amandla esistimu okusebenza ezindaweni ezinokukhanya okuphansi asho ukuthi ingakwazi ukuhlanza ngesikhathi esingaphansi kwesidingo (ekuseni kakhulu kanye nasebusuku kakhulu), igweme ukuminyana kwabantu futhi ithuthukise ukusebenza kahle. I-S330 eyodwa ithatha indawo yezihlanzi eziyi-16 ezenziwa ngesandla, ihlanganisa amamitha ayi-8,000 esikwele ngemizuzu engu-40—kaphinda kane ngokushesha kunamaqembu abantu.
Izinselele Neziqondiso Zekusasa: Umngceleli Olandelayo Wokubona Ngekhamera
Ngaphandle kokuphumelela kwayo, ukubona ngekhamera ekuhlanzeni imigwaqo okwenziwa imishini kusabhekene nezinselele ezizobumba ubuchule bekusasa. Okubaluleke kakhulu ukubhekana nezimo zezulu eziqine kakhulu, njengeqhwa eliningi noma inkungu eminyene, lapho ngisho nokucubungula okuzivumelanisa nezimo kunzima ukugcina ukucaca kwesithombe. Abacwaningi bahlola amakhamera okushisa ukwengeza amakhamera e-RGB kulezi zimo, njengoba ukuthwebula okushisayo kungathola udoti ngokungafani kwezinga lokushisa kunemininingwane ebonakalayo.
Olunye uhlobo lokugxila ukuhlanza ngokubambisana—ukusebenzisa ukubona ngekhamera ukuvumela amarubhoti amaningi ukuba asebenze ndawonye. Izinhlelo zesikhathi esizayo zizovumela amarubhoti ukuba abelane ngedatha ebonakalayo ngesikhathi sangempela nge-5G, ngakho irubhoti elithola inqwaba enkulu yemfucumfucu lingazisa amayunithi aseduze ukuba ashintshe indlela futhi asize. Le ndlela yokubambisana izothuthukisa ukusebenza kahle ezindaweni ezinkulu zasemadolobheni, njengezindawo zezindiza noma izindawo zamabhizinisi.
Ekugcineni, ukuhlanganiswa kwe-camera vision namapulatifomu amadolobha ahlakaniphile kuvula amathuba amasha okuphathwa kwezindaba zokuhlanzeka okusekelwe kudatha. Amakhamera angakwazi ukuqoqa idatha ngezindawo ezinodoti, izinhlobo zodoti, kanye nokuphindaphinda kokuhlanza, okungasetshenziswa omkhandlu ukuthuthukisa imizila yokuqoqa udoti nokukhomba imizamo yokuvimbela (isibonelo, ukubeka iziqukathi zodoti eziningi ezindaweni ezinodoti). E-Longgang District yaseShenzhen, le ndlela esekelwe kudatha yehlise isikhathi sokugcinwa kukadoti emoyeni ovulekile ngo-30%, ithuthukisa impilo yomphakathi futhi yehlisa izikhalazo ngephunga elibi.
Kungani Ukubona NgeKhamera Kuyisikhathi Esizayo Sokuhlanza Imigwaqo Ngamarubhoti
Umbono weKhamera uvele njengomqhele ongadumile wokuhlanza imigwaqo nge-robot, unikeza ukuhlanganiswa okukhethekile kokusebenza kahle kwezindleko, ukunemba, kanye nokwandiswa okungafinyelelwa ezinye izobuchwepheshe bokuhlola. Ngokunqoba izinselelo zemvelo ezishintshashintshayo zedolobha ngokusebenzisa ukuhlanganiswa kwemodi eminingi, i-AI elula, kanye nokucubungula okwenziwe ngokwezifiso, umbona weKhamera ushintshe abahlanzisi be-robot ukusuka kumathuluzi okuhlola abe izingxenye ezibalulekile zezakhiwo zedolobha ezihlakaniphile.
Njengoba amadolobha emhlabeni wonke ebhekene nezisebenzi ezindala emisebenzini yokuhlanzeka, izindleko eziphakeme zemisebenzi, kanye nezidingo ezikhulayo zezindawo ezihlanzekile, izihlanzi ezisebenzisa ubuchwepheshe bekhamera zinikeza isixazululo esizinzile. Azigcini nje ngokuthuthukisa ukusebenza kahle kokuhlanza kodwa futhi zithuthukisa ukuphepha kwezisebenzi ngokusingatha imisebenzi ephindaphindayo, enokubonakala okuphansi (isibonelo, ukuhlanza ekuseni kakhulu noma ebusuku) engozi enkulu kubantu.
Ikhanda lekhamera esikhundleni esizayo libukeka kahle kakhulu. Ngokuqhubeka kwezinguquko ku-AI, i-edge computing, kanye ne-sensor fusion, singalindela izihlanzi ezisebenza ngamandla kakhulu, eziphumelelayo, futhi ezibambisanayo—konke kunikwa amandla "amehlo" abona imizi yethu kangcono kunanini ngaphambili. Ezimaspala, abahlinzeki bezobuchwepheshe, kanye nezakhamuzi ngokufanayo, ikhamera vision ayiyona nje ubuchwepheshe obusha; kuyisikhuthazi sokudala izixeko ezihlakaniphile ezihlanzekile, eziphilile, futhi ezingcono kakhulu ukuhlala kuzo.
Uxhumane
Sicela uxhumane nathi uhambele
WhatsApp
WeChat