A pedestrian re-identification method based on uncertainty optimization

Abstract

The invention relates to the field of computer vision, adopts a deep learning framework, and particularly relates to a pedestrian re-identification method based on uncertainty optimization, which comprises the following steps of 1) using a twin network structure to respectively use two original images belonging to the same or different pedestrians as the input of two isomorphic networks; 2) usingan inception network and a Dropout layer superposition mode to realize the Bayesian convolutional neural network with uncertainty optimization as a feature extraction network; 3) calculating binary classification loss and multi-classification loss of respective networks according to feature output of the twin network, and superposing the binary classification loss and the multi-classification lossfor back propagation and parameter optimization of the networks; 4) inputting the pedestrian image to be identified and all comparison images into the trained model, and extracting image features; 5)obtaining a final distance between the pedestrian image to be identified and the comparison image by using a Euclidean distance calculation formula; and 6) performing sorting according to the featuresimilarity distance to obtain a comparison image matching sorting corresponding to the to-be-identified pedestrian. Compared with the prior art, the method has the advantages of high accuracy, high robustness, rapidness, simplicity, convenience and the like under the conditions of all samples and few samples.

Description

technical field [0001] The present invention relates to the field of computer vision and adopts a deep learning framework. More specifically, the present invention relates to a pedestrian re-identification method based on uncertainty optimization. Background technique [0002] Pedestrian re-identification is a key part of intelligent video analysis to break through the bottleneck of mass surveillance video technology application, which has attracted the attention of many researchers in recent years. Pedestrian re-identification refers to matching the same pedestrian target at different times and under different cameras. Pedestrian re-identification is a very challenging research problem. In the large-scale video surveillance network system in real life, the complexity and unpredictability of the environment and equipment will bring various uncertainties to the realization of pedestrian re-identification. First, due to different hardware conditions and parameters of differen...

AI Technical Summary

Problems solved by technology

However, this method does not take into account the overfitting and uncertainty problems in the process of pedestrian re-identification. The samples, parameters, and structural uncertainties in deep learning will have a greater negative impact on the results of pedestrian re-identification, especially when pedestrians are re-identified. When the number of samples is small, the recognition rate is low

Images