Pedestrian re-identification method based on global distance scale loss function

Abstract

The invention discloses a pedestrian re-identification method based on a global distance scale loss function. The method comprises the following specific steps: carrying out data enhancement processing on training data of a pedestrian re-identification data set; randomly selecting each batch of data; constructing a deep neural network based on human body components and initializing the network; using a cross entropy loss function and a loss function based on a global distance scale to simultaneously supervise training of the deep neural network; and performing feature extraction on the targetpedestrian image and the pedestrian image in the pedestrian image library, performing cosine similarity calculation, and performing sorting to obtain an identification result. According to the loss function of the global distance scale based on the statistical characteristics, noise interference can be effectively avoided, the risk of overfitting can be effectively reduced, and the robustness andthe generalization capability of the model can be improved.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a pedestrian re-identification method based on a global distance scale loss function. Background technique [0002] With the development and advancement of deep learning and the increasing popularity of video surveillance technology, pedestrian re-identification technology has become more and more important because of its ability to search for pedestrians to be found among a large number of pedestrians. [0003] In the process of the development of person re-identification technology from traditional metric learning to deeper and wider deep neural network learning, the measurement of distance is inseparable. Pedestrian re-identification technology can effectively shorten the distance between the same kind and increase the distance between the different kinds in the feature space. Based on distance metrics, researchers have proposed many loss functions for supervised netw...

AI Technical Summary

Problems solved by technology

[0004] However, the loss function based on the distance scale in the existing technology needs to rely on the measurement relationship between the two classes or between classes, and lacks a consideration of using global statistical properties.

Images