Vehicle re-identification method based on cross-context and feature response attention mechanism

The invention relates to the technical field of vehicle re-identification, in particular to a vehicle re-identification method based on a cross-context and feature response attention mechanism, and the method comprises the steps: firstly providing a local mixed cross-context-feature response attention mechanism, and designing a channel attention module and a space attention module based on the attention mechanism; the global information and the local information are integrated into a unified system structure in an efficient mode to improve the attention learning efficiency and performance. A channel grouping reduction method and a spatial resolution reduction method are respectively used in the channel and the spatial attention module to reduce the parameter quantity and the calculation burden of the network; and finally, designing to remove similarity constraints, forcing space attention modules on multiple branches to pay attention to different semantic information, and realizing vehicle re-identification.