Shielded pedestrian re-identification method based on teacher and student learning framework

Abstract

The invention discloses a shielded pedestrian re-identification method based on a teacher and student learning framework. The shielded pedestrian re-identification method comprises the steps that firstly, a teacher network is trained, and the existing large-scale complete pedestrian data are used for simulating the training process of shielding pedestrian re-identification, and the process is achieved through a joint significant detection network with a cross-domain simulator, and the process is the teacher teaching process; then the teacher network is transmitted to the student network, and the student network is enabled to utilize the model of the teacher network to continue to train on the real small-scale pedestrian shielding data, and the process is the student practice process; and finally, through the teacher teaching and student practice process, a model with pedestrian discrimination and shielding robustness is obtained through training, and the model can be used for shieldingpedestrian re-identification. The shielded pedestrian re-identification method can greatly improve the effect performance of an existing pedestrian re-identification shielding task, and has a wide application value.

Description

technical field [0001] The invention relates to a pedestrian re-identification method for occlusion, in particular to a multi-stage cross-domain learning method and a teacher-student learning framework-based re-identification method for occluded pedestrians. Background technique [0002] The pedestrian re-identification task refers to searching for pedestrians with the same identity across different cameras at different times, angles, lighting and other conditions. With the rapid development of intelligent surveillance systems, pedestrian re-identification technology is used in various public practical applications, aiming to find specific pedestrians in different cameras, such as criminals, children and missing persons. However, in the application of real scenes, the pedestrians captured by the camera are often blocked by static or dynamic occluders in the surrounding environment, such as other pedestrians, moving vehicles, buildings, flowers and trees, etc., resulting in t...

AI Technical Summary

Problems solved by technology

Although this scheme can achieve a certain effect, since it is necessary to perform feature extraction on each small block, the complexity of the calculation is greatly increased, and it is also prone to block alignment problems, which will affect the effect of pedestrian re-identification

Due to the limited research work on the re-identification of occluded pedestrians in recent years, in addition to the limitation of insuf

Images