Izidingo zobuchwepheshe zamakhamera aku-General Humanoid Robots

Ulwazi olujulile lubalulekile ukuze amarobhothi avamile aqonde ukwakheka kwezinhlangothi ezintathu zendawo ezungezile, afinyelele ukuzulazula okunembile nokugwema izithiyo, futhi aqedele imisebenzi efana nokubamba. Ubuchwepheshe bokubona ukujula obujwayelekile, obufana namakhamera we-stereo umgomo we-parallax nokujula amakhamera kusetshenziswa amasu okukhanya okuhleliwe noma esikhathi sendiza (i-ToF), ihlinzeka ngamarobhothi ulwazi olujulile lwezinto. Uma irobhothi likhipha into, ubuchwepheshe bokubona ukujula bungakala ngokunembile indawo nokuma kwento, okwandisa izinga lempumelelo yokubamba. Ngesikhathi sokuzulazula, isiza irobhothi ibanga nendawo yezithiyo ezizungezile, lihlela indlela ephephile.

Ukuze kuzuzwe umbono obanzi wezemvelo, amarobhothi ajwayelekile e-humanoid avame ukusebenzisa ubuchwepheshe bamakhamera amaningi. Ngokuhlanganisa izinhlobo ezahlukene noma imibono ye amakhamera, njengamakhamera e-RGB namakhamera ajulile, irobhothi lingathola kokubili ulwazi lombala nokuthungwa kwezinto nolwazi olujulile, lithuthukise ukuqonda kwerobhothi kanye nombono walo ngendawo ezungezile. Amanye amarobhothi e-humanoid athuthukisiwe afakwe amakhamera amaningi, ezwa indawo ezungezile evela ema-engeli ahlukene, afinyelela ukumbozwa okubonakalayo, futhi athuthukisa ukwethembeka nokunemba kwesistimu yokubuka. Uma ikhamera yehluleka, amanye amakhamera asengakwazi ukuqinisekisa ukusebenza okuyisisekelo kwerobhothi, afeze ikhophi yasenqolobaneni yokuphinda umsebenzi.

Ngokuthuthukiswa kobuchwepheshe bobuhlakani bokwenziwa, amakhamera esejwayelekile amarobhothi e-humanoid awasewona nje amadivayisi okuthola izithombe kodwa ahlanganisa ama-algorithms ahlakaniphe kakhulu njengokutholwa kwento, ukubonwa kwesithombe, ukuhlukaniswa kwe-semantic, ukubonwa kokuthinta, nokubonwa kwesimo sobuso. Lawa ma-algorithms angenza ukuhlaziya kwesikhathi sangempela kanye nokucutshungulwa kwezithombe ekugcineni, ukunciphisa ukudluliswa kwedatha nokuthuthukisa ukusebenza kahle kokucubungula, ukufeza izinqumo ezisheshayo nezinembe kakhudlwana. Ngokusebenzisa ama-algorithms okuthola into, amarobhothi angakwazi ukubona ngokushesha okuhlosiwe okunjengabantu, izimoto nezinto; ukubonwa kokuthinta kanye nama-algorithms wokuqaphela isisho sobuso asiza amarobhothi ukuthi azuze ukusebenzisana okungokwemvelo okwengeziwe komshini womuntu.

Lapho enza imisebenzi, amarobhothi avamile e-humanoid adinga ukuthi amakhamera akwazi ukucubungula inani elikhulu ledatha yesithombe ngesikhathi sangempela kanye nemiphumela yokuhlaziya okukhipha ngokushesha. Lokhu kudinga ukuthi amakhamera abe namandla okwenza ikhompuyutha anamandla kanye nezakhiwo ze-algorithm ezisebenzayo ukuze ahlangabezane nezimfuneko zesikhathi. Amanye amakhamera afakwe izisheshisi zokufunda ezijulile ezakhelwe ngaphakathi, ezingasebenzisa ngokushesha amamodeli okufunda ajulile njengamanethiwekhi e-convolutional neural (CNN), azuze ukuhlaziya kwangempela nokuqonda izigcawu eziyinkimbinkimbi. Ekuqaphelweni kwezokuphepha, amakhamera erobhothi aqapha imisebenzi yabasebenzi ngesikhathi sangempela, futhi uma ukuziphatha okungavamile kutholakele, i-alamu icushwa ngokushesha encike emandleni ayo okucubungula ngesikhathi sangempela.