isiZulu
Ikusasa Lezinhlelo Zekhamera: Indlela Ama-Chip e-AI Ase-Sensor Ashintsha Ngayo Imigomo
Kwadalwa ngo 2025.12.25
Imohla wekameraimboni ibheke endaweni yokuguquka. Ngaphezu kweminyaka eyishumi, intuthuko ibikhokhelwa yimpi ye-pixel, ukuhlanganiswa kwezibuko eziningi, kanye nokuthuthukiswa kwe-algorithm yangemuva—kodwa lezi zindlela ziqala ukuhamba kancane. Ama-smartphone manje anama-bumps ezithombe athatha u-25%–40% wendawo yomkhiqizo, nokho abasebenzisi abacabangi ngcono okuncane okukhona. Amakhamera emboni abhekene nezinkinga zokulibaziseka ekuhlaziyeni ngesikhathi sangempela, futhi amadivayisi e-IoT abhekene nezithiyo zamandla ezinciphisa amandla e-AI. Faka ama-chip e-AI asensori: ushintsho olukhulu oluhambisa ubuhlakani ukusuka efwini noma kumphakathi wedivayisi ngqo kum sensor yezithombe, uvula ukusebenza okungakaze kubonwe, isivinini, kanye nokuguquguquka.
Ukuvalwa Kwepharamitha Endala: Kungani Sasingadingi I-AI Ebonakalayo
Ukuze siqonde ukubaluleka kwe-AI ebonakalayo, kumele sokuqala siqonde amaphutha ezakhiwo zemakhamera ezijwayelekile. Ake sithathe umkhondo wokuguquguquka kwemboni:
• Isikhathi Sezithombe Zezibuko (2010–2016): Inqubekelaphambili yehlulekile ekuncikeni ezithombeni ezinkulu, ama-aperture amakhulu, kanye ne-megapixels aphezulu. Kodwa izimo zokwakheka kwefoni zafaka imikhawulo eqinile—awukwazi ukufaka isithombe esikhulu se-DSLR kudivayisi encane.
• Isikhathi Sezithombe Zokubala (2017–2023): Ama-algorithms afana ne-HDR, imodi yephupho, kanye nokuhlanganiswa kwemifanekiso eminingi akhokhela izinkinga zehardware. Nokho, lokhu kudale izinkinga ezintsha: ukuhamba kokuqhuba, ukusetshenziswa kwamandla okweqile, kanye nokuncika kakhulu kumithombo ye-ISP/NPU.
• Isikhathi Sokuhlanganiswa Kwamakamelo Amaningi (2021–2024): Abakhiqizi bafaka izithombe ezibanzi, ezikude, kanye nezithombe zokujula ukuze baphumelele ezinkingeni zezibuko. Nokho, isibani esengeziwe sikhuphula ubunzima be-algorithm ngendlela engapheli, kanti izinkinga zokushisa zinciphisa isikhathi sokurekhoda ividiyo.
Ngokweminyaka ka-2024, imboni ibhekene neqiniso elibuhlungu: ukuthuthukiswa kokusebenza kwakuncipha njengoba izindleko nezinkinga zikhula. Abathengi bangenza baphumelele ukushintsha impilo ye-bhathri noma ubukhulu bedivayisi ukuze bathole ukuthuthukiswa okuncane kwezithombe. Okudingekayo akukhona ukuphucula ukuhlanganiswa kwezinsiza, kodwa ukucabanga kabusha okuyisisekelo kokuthi izinhlelo zokuthwebula izithombe zisebenza kanjani. I-AI esesikhathini sokuthwebula izithombe iletha lokho ngokuhambisa ukubala emthonjeni wedatha—i-sensor uqobo.
Indlela I-AI Esesikhathini Sokuthwebula Izithombe Iguqula Imodyuli Zekhamera
I-AI esesikhathini sokuthwebula ihlanganisa izixhumi ezikhethekile zokucubungula i-neural ngqo kumasensor wesithombe we-CMOS, ivumela ukuhlaziywa kwedatha ngesikhathi sokuthwebula. Le shintsho yokwakha iletha izinzuzo ezintathu ezishintshayo:
1. Ukubambelela Okuphansi Kakhulu Nokunciphisa Ukusetshenziswa Kwamandla
Izinhlelo zendabuko zidinga idatha yesithombe el raw ukuthi ihambe ivela kumsensor iye kumphakathi wesikhumbuzo (ISP/NPU), bese ibuyela emibonweni—kudala ukuhamba okuthile okuphazamisa izinhlelo ezisebenzayo ngesikhathi sangempela. I-Sony LYTIA 901, okuyisithombe sokuqala sokuhweba esinezikrini zokuhlola ezihlanganisiwe ze-AI, ikhipha le nkinga ngokucubungula idatha ngaphakathi. Isibonelo, i-QQBC (Quad Quad Bayer Coding) eyakhiwe nge-AI ibuyisela izithombe eziphezulu ngesikhathi sokukhulisa i-4x ku-30fps ngaphandle kokudinga amandla ebhethri.
Le msebenzi ubalulekile kumadivayisi asebenzisa ibhethri. I-NSF-ixhaswe i-Preventive Maintenance AI Chip isebenza ngamamicroamperes ambalwa, ivumela ukuqapha okungama-24/7 kwemishini yezimboni namadrone ngaphandle kwesidingo sokuphinda ukushaja kaningi. Kuma-smartphones, i-AI esesikhathini sokuhlola yehlisa umthwalo we-ISP ngaphezu kuka-60%, ikhulisa isikhathi sokurekhoda ividiyo futhi yehlisa ukukhiqizwa kokushisa.
2. Kusuka "Ukuthwebula Idatha" kuya ku "Ukuqonda Izimo"
Ukuphakama okukhulu nge-AI esesimweni sokuhlola kukhombisa ushintsho oluvela ekuqoqweni kwedatha okungaphansi kokubhekwa kuya ekuhloleni izigcawu. Imodyuli zekhanda ezedlule zazirekhoda lokho ezazikubona; ezamanje zihlaziya ngokushesha. Isikhumbuzo esizayo se-Samsung esine-"Zoom Anyplace" ubuchwepheshe silandelela izinto ngokuzenzakalelayo ngenkathi sithwebula kokubili okukhulisiwe nokuphelele - konke kuhlinzekwa ngqo esikhumbuzweni.
Ezindaweni zezimboni, i-Lucid Vision Labs’ Triton Smart Camera isebenzisa i-Sony IMX501 sensor ukwenza ukutholwa kwezinto nokuhlukaniswa offline, ngaphandle kokuxhumana nefu noma ama-prosesa angaphandle. Ukuklama kwayo kwe-dual-ISP kuhambisa i-AI inference nokucubungula izithombe ngesikhathi esisodwa, kuhambisa imiphumela ngemizuzwana—okubalulekile ekuzenzakaleni kwefektri lapho izinqumo eziphuthumayo zivikela isikhathi sokuphumula esibizayo.
3. Izinsiza Ezinembile, Amandla Aphakeme
I-AI esebenzisa isikhala sokukhanya inciphisa ukuncika kumasistimu amaningi wemakhamera ngokulingisa imiphumela yokukhanya ngokucubungula okuhlakaniphile. I-Sony’s LYTIA 901 ifinyelela ku-4x optical-quality zoom nge-lens eyodwa, kungenzeka inciphise amamojula wemakhamera ezithombe ze-smartphone eziphezulu zisuka kumalensi amathathu/nane ziye kumalensi amabili kuphela. Lokhu akunciphisi kuphela ubukhulu bezisetshenziswa kodwa futhi kunciphisa nezindleko zokukhiqiza ngokususa izingxenye eziphindiwe ezifana namalensi engeziwe kanye ne-VCM motors.
Kwe-IoT kanye nezisetshenziswa zokuhlala ezihlakaniphile, le ndlela yokwenza kube lula iyashintsha. I-SK Hynix’s on-sensor AI prototype ihlanganisa ukuqashelwa kobuso nezinto ngqo kumasensor amancane, ivumela amakhamera okuphathwa kalula, anokonga amandla, kanye nezinsimbi zokushaya emnyango.
Izicelo Zangempela Eziguqula Imboni
Umthelela we-AI on-sensor udlulela kude kakhulu kumafoni, udala izimo ezintsha zokusetshenziswa ezindaweni ezahlukene:
Izinto Zokusetshenziswa Kwabathengi: Ukukhula Kwe-"AI-Native" Imaging
Amakhamera ezithombe eziphonesi azogxila ekuguqulweni kwesimo okuhlakaniphile kunokubala kwamaphikseli. Cabanga ngekhamera ezizenzakalelayo ezilungisa imibala yesikhumba ekukhanyeni okuphansi, ikhiphe izinto ezingafunwanga ngesikhathi sangempela, noma ilungiselela ukuhloleka kwemibhalo—konke ngaphandle kokucubungula ngemva. Ibhendi ye-Sony ethi LYTIA ibonisa isikhathi esisha lapho i-AI ezingeni lesikhala iba yisici esijwayelekile, iguqula ukuncintisana kusuka kumaspecs wezinsiza kuya ekuhlanganiseni kwemvelo nasezinhlelweni ezithile zesimo.
Ukuziphatha Kwezimboni: Ukugcinwa Okubikezelayo 2.0
Izikhungo zokukhiqiza zisebenzisa amakhamera e-AI aqapha impilo yezinsiza. I-NSF’s Preventive Maintenance AI Chip ihlaziya ukuhamba nokwakheka kwezwi ukuze ithole izinkinga ngaphambi kokuba zenzeke, inciphisa isikhathi sokungasebenzi ngaphezu kwama-40%. I-Lucid’s Triton Smart Camera, enokuhlola kwe-IP67 kanye ne-range yokusebenza ye- -20°C kuya ku-55°C, iphumelela ezindaweni ezinzima zezimboni, inikeza ukuhlaziywa okuqhubekayo ngaphandle kokulibaziseka kwefu.
Ezokuthutha Nezokuthutha: Ukuqonda Okuphephile, Okuhlakaniphile
Izimoto ezizimele kanye ne-ADAS (Izinhlelo Zokwesekwa Kwezithuthi Ezithuthukisiwe) zidinga ukutholwa kwezingozi ngokushesha. I-AI esesikhathini sokuhlola ibhizinisi idlulisa idatha yokubona ngemizuzwana, ibona abantu, abahamba ngebhayisikili, nezithiyo ngokushesha kunezinhlelo zendabuko. Ngokunciphisa ukuthembela kumayunithi okucubungula amakhulu, lezi zikhala zithuthukisa ukwethembeka futhi zinciphisa ukusetshenziswa kwamandla—okubalulekile ezimotweni zikagesi lapho wonke amawatt abalulekile.
I-IoT Nezindawo Ezihlakaniphile: Njalo-Khona, Ukuhlola Okuphansi Kwamandla
Izicelo zedolobha ezihlakaniphile ezifana nokuhlola ukuhamba kwezithuthi nokuphepha komphakathi zidinga amakhamera asebenza amahora angama-24/7 ngamandla alinganiselwe. I-AI esesikhathini sokuhlola ivumela lezi zisetshenziswa ukuba zicubungule idatha endaweni, zithumele kuphela izaziso ezibalulekile esikhundleni sokudlulisa ividiyo eqhubekayo. Lokhu kwehlisa izindleko ze-bandwidth futhi kuthuthukisa ubumfihlo ngokugcina idatha ebucayi kudivayisi.
Indlela Eya Phambili: Izinselelo Nezinto Ezizayo Zokwakha
Ngenkathi i-AI esesikhathini sokuthwebula isivele ishintsha amamojula wekhamera, ukuthuthukiswa okuningi kuzokwenza ukuthi ibhande lesikhathi esizayo:
Ukuthuthuka Kwezobuchwepheshe
• Ukuxubana Kwezimodi Ezinhlobonhlobo: Amakhamera azayo azohlanganisa ukuhlela kwedatha yokubona, yokuzwa, kanye nedatha yemvelo, avumele ukuqonda okujulile kwezimo.
• Ukuklama Kwenhliziyo: Ukuphindaphinda ukwakheka kobuchopho bomuntu kuzokwehlisa kakhulu ukusetshenziswa kwamandla ngenkathi kuthuthukisa ukunemba kokuhlonza amaphethini.
• Ama-AI Cores Angalungiswanga: Izinsiza ezifana ne-chip ye-NSF enezinhlelo ezilungisekayo zizovumela abathuthukisi ukuthi baphathe imodeli ezenzakalelayo zezimo ezithile ngaphandle kokushintsha izinto zokusebenza.
Ukushintsha Kwezimakethe
Imakethe ye-smart sensor yomhlaba ibonakala ukuthi izokhula ngokushesha eminyakeni ezayo, lapho ukuzenzakalela kwezimboni nezinto zokusebenza zezimoto zenza ngaphezu kwama-40% wezidingo ngonyaka ka-2026. Ukuncintisana kuzokwandisa njengoba i-Samsung ne-SK Hynix zikhankasa i-54% yendawo yemakethe ye-Sony ngokusheshisa imikhiqizo yabo ye-AI esesimweni. Sizobona futhi ushintsho oluvela ekuthengiseni izinsiza zokuqala kuya kumamodeli e-"sensor-as-a-service", lapho izinkampani zenza imali ephindaphindayo ngokuvuselelwa kwe-algorithm kanye nokuhlaziywa kwedatha.
Izinto Ezibalulekile Zokulawula Nezokuziphatha
Njengoba ama-module wekhamera ethola ubuhlakani obukhulu, ukukhathazeka ngokuqinisekiswa kuzokhula. Ukucubungula ku-sensor kusiza ngokugcina idatha endaweni, kodwa izindinganiso zokuphatha idatha nokucaca kwe-algorithm kuzoba kubaluleke kakhulu. Ohulumeni sebesakha imithetho yezinsiza ze-AI ezisendaweni, ezizoshintsha ukuthuthukiswa kwemikhiqizo eminyakeni ezayo.
Isiphetho: Isikhathi Esisha Sokuthwebula Izithombe Ngokuhlakanipha
Ama-chips e-AI esesikhathini sokuthwebula izithombe awawona nje ukuthuthukiswa okuncane—bamele ushintsho olukhulu endleleni imodyuli zekhamera ziqopha, zisebenza, futhi ziqonda idatha yezithombe. Ngokuhambisa ubuhlakani kwi-sensor, imboni ixazulula izinkinga eziyisisekelo phakathi kokusebenza, amandla, nobukhulu ezibambezele ubuchwepheshe iminyaka.
Kusukela kumafoni anencane anempilo engcono yebhethri kuya kumakhamera ezimboni avimbela ukuwa kwezinsiza, izicelo azinakuqedwa. Njengoba i-Sony’s LYTIA 901 kanye ne-Lucid’s Triton Smart Camera zikhombisa, ikusasa lezi zinsiza zamakhamera alikho ematheni amaningi noma kumamegaphikseli aphezulu—kukhona ngama-sensors aqondile aqonda umhlaba ngesikhathi sangempela.
Kubakhiqizi, abathuthukisi, kanye nabathengi, le mpumelelo kusho ukuthi izinsiza zamakhamera ngeke zibe amathuluzi kuphela okuthwebula imizuzu—zizoba izinhlelo ezihlakaniphile ezithuthukisa ukwenza izinqumo, zithuthukise ukuphepha, futhi zivule amathuba amasha kuzo zonke izimboni. Isikhathi sokuthwebula ngezithombe ezisebenzisa i-AI sesifikile, futhi kusazoqala.
Uxhumane
Sicela uxhumane nathi uhambele
WhatsApp
WeChat