Multi-mode cooperative sensing system and method for intelligent network connection vehicle
The invention discloses a multi-modal cooperative sensing system and method for an intelligent networked vehicle, and the method comprises the steps: carrying out the single-modal fusion of LiDAR features and LiDAR features from different vehicles, and RGB image features and RGB image features throu...
Saved in:
Main Authors | , , , |
---|---|
Format | Patent |
Language | Chinese English |
Published |
19.04.2024
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | The invention discloses a multi-modal cooperative sensing system and method for an intelligent networked vehicle, and the method comprises the steps: carrying out the single-modal fusion of LiDAR features and LiDAR features from different vehicles, and RGB image features and RGB image features through employing a self-attention mechanism after the multi-modal vehicle features are extracted and preprocessed; feature alignment is carried out on the LiDAR-BEV features and the RGB-BEV features; and stacking the RGB-BEV feature and the LiDAR-BEV feature which are aligned to generate a pre-fused BEV feature, taking the pre-fused BEV feature as an initialized query Query of a cross attention mechanism, constructing a Key and Value matrix by using the LiDAR-BEV feature after feature alignment, and obtaining a BEV fusion feature by using the cross attention mechanism. Compared with the prior art, the method has the advantages that a more comprehensive and reliable sensing solution is realized, the robustness and relia |
---|---|
Bibliography: | Application Number: CN202410124929 |