isiZulu
Umbono weKhamera vs LiDAR: Yikuphi Okusebenza Kangcono Enkungu?
Kwadalwa ngo 01.13
Inkungu ingenye yezitha ezinzima kakhulu ezimotweni ezizihambelayo kanye nezinhlelo zokusiza abashayeli ezithuthukisiwe (ADAS). Iphazamisa ukukhanya, ihlakaza izimpawu, futhi inciphisa ukwethembeka kokubona imvelo—amakhono ayisisekelo agcina abashayeli nabahamba ngezinyawo bephephile. Ingxoxo phakathi kombono wekhamera ne-LiDAR (Ukuthola Ukukhanya Nokubanga) isibe khona iminyaka, kodwa izimo zenkungu zikhipha imikhankaso yokumaketha futhi ziphoqa ukugxila ekusebenzeni okuyisisekelo: Yikuphi ubuchwepheshe obuletha ngempela lapho ukubonakala kuncipha?
Le ndatshana idlula ukwahlukaniswa okujwayelekile kwe-"hardware vs. software". Esikhundleni salokho, sihlela ukuqhathanisa phakathi kwemigomo emibili ehlukene ye-"safety philosophies": ukubona kwekhameraukuthembela kumqondo we-algorithmic ukuze kudlule izithiyo zomzimba, kanye nokusetshenziswa kwe-LiDAR kokuphindaphinda kwehardware ukuze kuqinisekiswe isisekelo sokuthembeka. Sisebenzisa idatha yokuhlola yezenzakalo yangempela ka-2025, ukuthuthuka kwezobuchwepheshe, kanye nezifundo zamacala embonini, sizophendula umbuzo obalulekile: Yikuphi okusebenza kangcono efu?
Uhlukaniso Oluyinhloko: Izingqikithi Ezimbili Zokuphepha Ngaphansi Komhwamuko
Ukuze siqonde ukuthi kungani umhwamuko wembula amandla nobuthakathaka bobuchwepheshe ngabunye, kuqala sidinga ukuchaza izimiso zabo zokusebenza eziyisisekelo—nezingqikithi zokuphepha ezishukumisa ukwamukelwa kwazo.
Izinhlelo zokubona ngekhamera zisebenza njenge "amehlo anikwe amandla ubuchopho." Zisebenzisa amakhamera anencazelo ephezulu (ngokuvamile angu-8-10 kumalungiselelo athuthukile) ahambisana namachips anamandla e-AI kanye nemininingwane emikhulu ukuze kulinganiswe ukubona kwamehlo komuntu. Ifilosofi lapha ukunciphisa izinto: sebenzisa isofthiwe ukukhokhela ihadiwe elinganiselwe, usebenzise ukufunda komshini ukuguqulela idatha yokubuka engu-2D ukuze uqonde imvelo engu-3D. I-Tesla ne-Xpeng yizona ezikhuthaza kakhulu le ndlela, ekhanya ezimweni ezicacile lapho izimpawu eziningi zokubuka zivumela ama-algorithm ukuthi achume.
I-LiDAR, ngokungafani, iy "ingqwele eqala ngezinto zokwakha." Ikhupha izigidi zemifutho ye-laser njalo ngomzuzwana ukudala ifu lamaphuzu e-3D elinembayo elizungezile, elikala amabanga, izimo, namacebo ngokunemba okungajwayelekile. Ifilosofi lapha ukubuyekezwa: sebenzisa amakhono okuzwa ngokomzimba ukusungula umkhawulo wokuphepha, noma ngisho noma izimo ezizungezile zifiphaza imininingwane ebonakalayo. I-Huawei, i-BYD, kanye nabahlinzeki abaningi be-ADAS abanethezekile bamukela le "LiDAR + ikhamera + i-millimeter-wave radar" trinity, bephambili ukusebenza okungaguqukiyo phezu kokonga izindleko.
Inkungu ihlupha zombili izinhlelo—kodwa ngezindlela ezihlukene kakhulu. Emakhamereni, inkungu ihlakaza ukukhanya, yenza imiphetho ibe yisicefe, futhi yenze umehluko ube buthakathaka, ivimbele izindlela zokubala ukuthi zithole izici ezibonakalayo ezidingekayo ukuze zibone izithiyo. Ku-LiDAR, izinhlayiya zenkungu zihlakaza imiqeqezo ye-laser, kudale "umsindo wephuzu lamafu" ongacisha izinhloso zangempela noma udale izinto ezingekho. Umbuzo akukhona ukuthi yikuphi "okungahlupheki"—ngokuthi yikuphi okungabuyela esimweni ngokushesha, okugcina izindlela ezibalulekile zokusebenza, futhi kugcine abashayeli bephephile lapho ukubonakala kubi kakhulu.
Idatha Yangempela: Indlela Ezisebenza Ngayo Emhwamukeni (Izivivinyo Zakamuva Zango-2025)
Ubufakazi obuphawulekayo buvela ku-2025 "Intelligent Driving Extreme Scenario Test White Paper," ekhishwe ngokubambisana yi-China Automotive Engineering Research Institute (CAERI) kanye ne-Dongchedi. Le study ebalulekile ihlole amamodeli angama-36 ajwayelekile ezindaweni ezingu-15km ezinomoya omningi kanye nezimo ezingu-216 zokuhlangana, ikala umehluko wokusebenza ngedatha eqinile. Ake sihlukanise imiphumela eyinhloko ngokubhekisela ebubanzini bomoya.
1. I-Efu Elincane (Ukubona: 200-500m)
Efu elincane—elivamile ekuseni kwasekuqaleni noma ezindaweni zasogwini—kokubili ubuchwepheshe busebenza kahle, kodwa izikhala ezincane zivele. Izinhlelo zokubona zekhamera, ezixhaswe ama-algorithms athuthukile okukhulula i-dehazing, zigcina izinga lazo ekuboneni izithiyo ezilula. I-Tesla’s FSD V12.5, ngokwesibonelo, yafinyelela ku-90% yokunemba kokubonwa kwezithiyo efu elincane, ngenxa yamathuluzi ayo okukhulula amanzi nezinkanyezi aqeqeshwe kumakhilomitha ayizigidi edatha yeqiniso.
Izinhlelo ze-LiDAR, phakathi naleso sikhathi, zagcina ukunemba okuseduze kokuphelele (98%+) ngomsindo omncane. I-Hesai ATX Lidar, imodeli entsha eyethulwe esikhathini eside, ibonise ikhono layo lokuhlunga u-99% womsindo ohlobene ne-fog ezingeni le-pixel, igcina amafu amaphuzu acacile ezimotweni nabahamba ngezinyawo ezizungezile. Umehluko lapha mncane, kodwa inzuzo ye-LiDAR ilele ekungaguquguquki: ngenkathi izinhlelo zekhamera zingase zibhekane nobunzima uma ukujiya kwe-fog kushintsha ngokungazelelwe, ukuzwa komzimba kwe-LiDAR kuhlala kuzinzile.
2. Moderate Fog (Visibility: 100-200m)
Uma ukubonakala kwehle ngaphansi kuka-200m, imikhawulo ye-algorithmic yekhamera iyabonakala. Ukuhlolwa kwe-CAERI kukhombise ukuthi amamodeli ekhamera ahlanzekile abhekane nokwanda okuphindwe ka-3 ezingeni lezithiyo ezingabonakali uma kuqhathaniswa nezimoto ezine-LiDAR. Ibanga lokubona izihambi le-Xpeng G6 lehla lisuka ku-150m esibhakabhakeni esicacile laya ku-65m enkungwini emaphakathi, kanti i-Tesla Model Y yehla yaya ku-78m. Lesi yisikhala esibalulekile: ngesivinini somgwaqo omkhulu (100km/h), ibanga lokuthola elingu-65m linikeza uhlelo imizuzwana engu-2.3 kuphela ukuthi lisebenze—okunganele ngisho nokubopha ngenxa yokuphuthuma.
Izinhlelo ze-LiDAR, ngokuphambene, zigcine ibanga lokuthola elisebenzayo ngaphezu kuka-80m. I-Huawei ADS 3.0, efakwe i-LiDAR enemigqa eyi-192, yafinyelela ibanga elijwayelekile lokubona abahamba ngezinyawo lika-126m enkungwini emaphakathi, inikeza iwindi lokuphendula elingamasekhondi angu-4.5. Umehluko uvela ekubeni nekhono le-LiDAR lokungena enkungwini kusetshenziswa amagagasi amade (1550nm) ancipha kancane kunokukhanya okubonakalayo okusetshenziswa amakhamera. Ngisho nalapho kuncipha, imisindo ye-laser igcina amandla anele ukubuyela kusenzisi futhi ibale amabanga ngokunembayo.
3. Dense Fog/Advection Fog (Visibility: <100m)
Efuzeni elikhulu—lapho ukubonakala kwehla ngaphansi kuka-100m, noma ngisho no-50m ezimweni ezinzima—ukwahlukaniswa kuba yichashaza. Lokhu kuyisimo se-"make-or-break" kumasistimu azimele, futhi idatha ye-CAERI ibonisa kahle: amasistimu wokubona ngekhamera abhekane nezinga lokuthathwa ngesandla elingu-15%, ngokuphindaphindiwe "ukwehluleka kokubona" izexwayiso. Ezimweni lapho ifu livalela izimpawu zomgwaqo, amalambu ezithuthi, futhi ngisho nezithiyo ezinkulu, ama-algorithms awanalo ulwazi olwaneleyo lokubona ukwenza izinqumo eziphephile.
Izimoto ezihlome nge-LiDAR, nokho, zigcine izinga lokuthatha isivinini elingu-3% kuphela. I-Huawei ADS 3.0 yaze yabonisa ikhono lokukhomba ngokunembayo izimoto ezimile futhi yenze iminyakazo yokugwema ekuboneni okungu-30m—izimo lapho abashayeli abantu bangalubona khona ngalé kwamalambu abo angaphambili. Okubalulekile kule nkulumo kukhona ama-algorithm athuthukisiwe okuhlunga inkungu, njengalawo athuthukiswa yi-LSLidar. Lawa ma-algorithm ahlaziya izici zamagagasi elaser ahlakazeke inkungu, ahlukanisa umsindo kudatha ye-point cloud evumelekile ukuze kugcinwe ulwazi olubalulekile lwezithiyo. Umphumela uhlelo olungagcini nje ngokuthi "lubone" ngenkungu—lugcina ukuqonda kwesimo lapho umbono wekhamera uhluleka ngokuphelele.
Ukuphumelela Kobuchwepheshe: Ukunciphisa Umehluko?
Ngenkathi i-LiDAR inamandla ezimweni zomhwamuko, ubuchwepheshe bobubili buyathuthuka ngokushesha. Ake sihlolisise izinto ezintsha zakamuva eziguqula ukusebenza kwazo emhwamukeni.
Umbono Wekhamera: Intuthuko Ye-Algorithmic
Izinyathelo ezinkulu ekusebenzeni kwefog ye-khamera ziza kuma-algorithms e-dehazing asekelwe ku-AI kanye nezinhlu ezinkulu, ezihlukahlukene. I-Tesla's FSD V12.5, isibonelo, isebenzisa inhlanganisela yokufunda ephethwe kanye nokufunda okungaphethwe ukuze "iphinde ikheze" imiphumela ye-fog, ibuyisela ukujula ezithombeni eziphukile. Ngokufundisa ngama-kilomitha ayizigidi eziyi-10 ze-data yezenzakalo zasesheshini nezimo zezulu ezinzima, uhlelo luthuthukisile isivinini sokulandela izinto eziphilayo ngama-40% ezimweni zokubona eziphansi.
Nokho, lezi zintuthuko zinezikhawulo. Zithembele ekubeni khona kwezici ezithile zokubona ukuze zisebenze—okuthile okuphuma emfog enkulu. Ngisho ne-algorithm ye-dehazing engcono kakhulu ayikwazi ukudala ulwazi olungekho, okwenza kube nzima ukweqa izikhawulo zomzimba zokubona kwekhamera.
LiDAR: Ukuhlanganiswa Kwezinsiza Ne-Algorithm
Ukuvela kwe-LiDAR kugxile ekuthuthukiseni ukungena, ukunciphisa umsindo, nokwehlisa izindleko. Enye yezinto ezithokozisayo kakhulu yi-single-photon LiDAR, ubuchwepheshe besizukulwane esilandelayo obuthuthukiswe ngokubambisana kososayensi base-UK nase-US. Lolu hlelo lusebenzisa izithwebuli ze-superconducting nanowire single-photon detectors (SNSPDs) ezizwelayo kakhulu nama-laser we-1550nm wavelength ukuthwebula izithombe eziphakeme kakhulu ze-3D nge-fog—ngisho nasemabanga angu-1 kilometer. Ngokuthola ama-photon ngamanye nokukala isikhathi sawo sokundiza ngokunemba kwe-picosecond (enye trillionth yesekhondi), uhlelo lungahlukanisa phakathi kwezinhlayiya ze-fog nezinto zangempela ngokunemba okungakaze kubonwe.
Izinhlelo ze-LiDAR zezentengiso nazo zithuthuka ngokushesha. I-algorithm ye-LSLidar yokuhlunga uthuli/imvula/inkungu, ehambisana nawo wonke amamodeli ayo (kuhlanganise ne-1550nm fiber kanye ne-905nm hybrid solid-state LiDAR), yehlisa kakhulu umsindo we-point cloud ngenkathi igcina ukutholwa kwezinhloso. I-ATX Lidar ye-Hesai, enenkambu ebanzi kakhulu engu-140° kanye nebanga lokuthola elingu-300m, ingakwazi ukukhomba futhi imake inkungu, umusi, namaconsi amanzi ngesikhathi sangempela, iqinisekisa idatha ye-point cloud ehlanzekile yesistimu. Lezi zinto ezintsha zenza i-LiDAR iqine kakhulu enkosini ngenkathi yehlisa izindleko—okwake kwaba isithiyo esikhulu ekwamukelweni—ngamanani ka-2025 ehla afinyelele ebangeni lika-$300-$450.
Ukukhetha Okusebenzayo: Nini Ukuze Uqhakambise Itekhnoloji Ethile?
Impendulo ethi "yikuphi okusebenza kangcono emoyeni" incike ekusetshenzisweni kwakho nasekutheni ungakwamukela kanjani ubungozi. Nansi isikh framework sokwenza izinqumo:
Ngemoto Zokuthenga (ADAS)
Uma uhlala endaweni enezimo ezivamile zomoya (isb., izindawo ezisemanzini, imifula, noma izimo ezibandayo ezinezinhlangothi zokushisa), i-LiDAR iyikhetho eliphephile. Idatha ye-CAERI ikhombisa ukuthi ikhono layo lokugcina ukuqonda kwesimo emoyeni omningi kunikeza isikhala esibalulekile sokuphepha. Nakuba ukubona ngekhamera kuthuthuka, ukuhlinzekwa kwe-LiDAR kwehardware kusebenza njenge "neti yokuphepha" engakwazi ukukopishwa ngama-algorithms.
Ezindaweni ezinefu elincane, ukubona kwekhamera okuhlanzekile kungase kube kuhle—ikakhulukazi uma izindleko zibalulekile. Imodeli ezifana neTesla Model Y neXpeng G6 zinikeza ukusebenza okuhle kwe-ADAS ezimeni ezicacile nezinefu elincane, ngokuthuthukiswa okuqhubekayo kwe-OTA okuthuthukisa njalo ama-algorithms azo ngokuhamba kwesikhathi.
Ngokuzimela Kwezamabhizinisi (Ama-Robotaxi, Ukuthutha)
Kwezicelo zezentengiselwano lapho ukuphepha nokwethembeka kuyinto okungadingidwa khona (futhi ukuthobela imithetho kuyimpoqo), i-LiDAR ayikhethwa nje kuphela—iyadingeka. Amatekisi ezimoto ezisebenza ezindaweni zasemadolobheni ezinemicimbi ye-fog engalindelekile, noma amaloli ahamba ibanga elide ahamba emigwaqweni evame ukuba ne-fog, awakwazi ukubekezelela izinga lokuthatha elingu-15% wezinhlelo zekhamera ezingenasizo. Izinga lokuthatha elingu-3% le-LiDAR ku-fog ewugqinsi luyinto ehlukile phakathi kokusebenza kahle kwezinto kanye nezingozi zokuphepha.
Ikusasa: Ukusebenzisana, Hhayi Ukuncintisana
Indlela yokucabanga phambili kakhulu akukona ukukhetha ubuchwepheshe obunye kunobunye—ukuhlanganisa. Izinhlelo zesimanje ze-ADAS (njengeHuawei ADS 3.0) zisebenzisa amafu amaphuzu athembekile e-LiDAR angu-3D ukugcwalisa idatha yesithombe enencazelo ephezulu yekhamera. Enkosini, i-LiDAR ihlinzeka ngokutholwa kwezithiyo eziyinhloko, kanti amakhamera asiza ukukhomba imininingwane efana nemibala yokukhanya kwethrafikhi noma izimpawu zezinyawo (lapho zibonakala). Lokhu "kuhlanganiswa kwezinzwa" kusebenzisa amandla omabili ubuchwepheshe, kudala uhlelo oluqine kakhulu kunanoma iyiphi yazo eyedwa.
Conclusion: LiDAR Leads in Fog, But Camera Vision Isn’t Out
When it comes to foggy conditions, the data is unambiguous: LiDAR outperforms camera vision across all fog severity levels, with a particularly wide gap in dense fog. Its hardware-driven approach to perception—penetrating fog with laser pulses and filtering noise with advanced algorithms—establishes a safety baseline that camera vision’s software-centric model cannot match, at least for now.
That said, camera vision is evolving rapidly. AI dehazing algorithms and larger datasets are improving its performance in light to moderate fog, making it a viable choice for regions with minimal extreme fog events. For most drivers and commercial operators, however, LiDAR’s ability to "see through fog" and reduce manual takeovers is a safety advantage that is hard to ignore.
Ekugcineni, ikusasa lokubona okuzenzakalelayo emhwamukeni lincike ekuhlanganisweni kwezinzwa. Ngokuhlanganisa ukwethembeka kwe-LiDAR nemininingwane yokubona kwamakhamera, singakha izinhlelo eziphephile, eziphumelelayo, futhi ezivumelana nezimo ezimweni zezulu ezinzima kakhulu. Okwamanje, uma ukuphepha komhwamuko kuyinto ebaluleke kakhulu kuwe, i-LiDAR iyona ewina kucace bha—kodwa ungayibali amandla okubona kwamakhamera njengoba izindlela zokusebenza ziyaqhubeka nokuthuthuka.
Uxhumane
Sicela uxhumane nathi uhambele
WhatsApp
WeChat