With　the　acceleration　of　digitization　and　informatization,　graphic　visualization　has　already　become　an　indispensable　tool　and　medium　in　modern　society.　Electroluminescent　devices　(EL),　which　refer　to　certain　materials　that　release　photons　through　internal　electron　leaps　when　excited　by　an　electric　field,　can　construct　low-cost　and　flexible　multispectral　image　sensors.　In　this　paper,　we　propose　an　alternating　current　EL　device　based　on　a　pyramidal　conical　structure　luminescent　layer　and　design　a　luminescent　display　image　recognition　system　in　combination　with　a　convolutional　neural　network.　The　system　can　recognize　the　shapes　of　objects　made　of　different　materials　while　effectively　reducing　the　influence　of　environmental　factors　on　recognition　accuracy,　thus　achieving　a　more　efficient　and　reliable　image　recognition　function.　Multi-spectral　imaging　technology　provides　rich　spectral　information　for　the　robot,　which　can　provide　richer　and　more　comprehensive　environment　perception　capability　to　meet　the　needs　of　diverse　dynamic　application　scenarios.　With　the　significant　advantages　of　EL　technology-based　image　recognition　devices,　such　as　high　brightness,　high　contrast,　low　power　consumption,　long　life,　flexibility,　and　multispectral　imaging　capability,　robots　can　adapt　to　complex　dynamic　environments　and　achieve　higher　recognition　accuracy　and　operational　efficiency.　(Figure　presented.)　©　Science　China　Press　2025.