Efficient pedestrian re-identification method based on neural network unsupervised contrast learning

Abstract

The invention discloses an efficient pedestrian re-identification method based on neural network unsupervised contrast learning. The method comprises the following steps: step 1, preparing a data settrained by a pedestrian re-identification model; step 2, selecting a convolutional neural network with relatively high feature extraction capability as a backbone network of the unsupervised contrastlearning model, the comparison learning in the unsupervised comparison learning model taking a feature vector as a starting point, i.e., constraining the feature vector extracted by the unsupervised comparison learning model, and correcting network parameters of the unsupervised comparison learning model by using an optimizer and a back propagation principle according to a loss function mode; step3, storing the feature vectors of training pictures into a cache region; and step 4, performing quantitative evaluation through the quantitative indexes. According to the method, standard data and non-standard data are used for training at the same time, high accuracy of a model is guaranteed, the characteristic that the model is easy to train is achieved, good model expansibility can be achieved, and better expansibility is achieved for new samples in a new environment.

Description

technical field [0001] The invention relates to intelligent security services covering the fields of social security, personnel monitoring, anti-terrorism and stability maintenance, etc., and provides an efficient pedestrian re-identification method based on neural network unsupervised comparative learning. A universal pedestrian recognition method with strong scene adaptability, cross-camera multi-angle, and high recognition accuracy. Background technique [0002] Pedestrian re-identification technology is an important experimental technology for video tracking. Through the extraction of pedestrian pictures from multiple cameras, the recognition and matching technology of the same person in different postures, different angles, and different environments is carried out. This technology can recognize pedestrians based on their clothing, posture, hairstyle and other information. At the same time, this technology can be used as an important supplement to face recognition techn...

AI Technical Summary

Problems solved by technology

[0007] The technology of pedestrian re-identification is greatly affected by environmental factors, and the risk of stable operation of the system is also great. In particular, the model with supervised training is not universal, and it is easy to be attacked and cause the system to fail to operate normally.

Images