Domain generalization pedestrian re-identification method based on multi-level data disturbance strategy

A domain generalization pedestrian re-identification model training method based on a multi-level data perturbation strategy comprises the steps of inputting an original pedestrian image into an explicit random perturbation module, reconstructing a target pedestrian and a perturbed pedestrian background into a pedestrian image after background perturbation, and obtaining the pedestrian image after background random perturbation; combining the original pedestrian image and the pedestrian image after background perturbation into an image pair, inputting the image pair into a baseline network for feature extraction, performing implicit perturbation on the extracted features by an uncertain sampling standardization module, and obtaining an output feature pair after integration; and inputting the output feature pair into a loss calculation module, splitting the output feature pair into original pedestrian features and disturbance pedestrian features to calculate covariance loss, finally calculating total loss, car