In response to the problems of insufficient detail capture and poor adaptability to multi-view changes in current three-dimensional(3D)reconstruction models

a 3D reconstruction model that fuses multi-view camera sensors and multimodal features is proposed.By using multi-view camera sensor data and through multimodal feature fusion and a global-local attention mechanism

the reconstruction quality of complex scenes is effectively improved.In addition

multimodal features such as color

depth

and semantics are extracted from multi-view two-dimensional(2D)images

and key regions are accurately captured by dynamically adjusting the importance of these features.Experimental results show that the proposed model outperforms existing mainstream models in 3D reconstruction on the local light field fusion(LLFF)and Danish Technical University(DTU)datasets.The peak signal-to-noise ratio(PSNR)indicators of the model reach 20.01

23.56

and 24.58 for 3—view

6—view

and 9—view inputs

respectively.The proposed model shows strong robustness in complex scenes and under multi-view changes

which verifies the effectiveness and reliability of the model.