Deep-learning-based pedestrian re-identification method

Abstract

The invention discloses a deep-learning-based pedestrian re-identification method. The method comprises the following steps: S1, bringing forward a deep network structure for pedestrian re-identification processing and obtaining similarity scores of pedestrians based on naked pixels of original images; S2, providing a learning sorting algorithm for guiding learning of the deep network; S3, carrying out sorting unit sampling on a training sample and training the deep network by using a stochastic gradient descent algorithm; and S4, after completion of the deep network training, for a pedestrian under one lens, calculating a score of similarity with a candidate image under another lens directly by a network, and obtaining a matching result. According to the invention, a mapping relation between original image pairs and corresponding similarity scores is established based on the deep convolutional neural network; and the network input is a pixel value of the original image and no pretreatment and design of hand-operated features are needed. Moreover, features with high discriminative and expressive properties can be learned based on large-scale data, thereby substantially improving the pedestrian re-identification effect.

Description

technical field [0001] The invention relates to the research field of pedestrian re-identification, in particular to a pedestrian re-identification method based on deep convolutional neural network for feature expression and similarity measure learning. Background technique [0002] At present, large-scale video surveillance networks have been popularized in various public places, such as railway stations, hospitals, airports and other places are the key areas of video surveillance. However, due to factors such as cost control and privacy rights, the monitoring network does not fully cover all areas, that is, the monitored areas are discontinuous. This brings great challenges to cross-camera video analysis (such as cross-camera pedestrian tracking, abnormal behavior detection and crowd flow analysis, etc.). To mine the high-level semantic information of pedestrians in the camera network through video surveillance technology, a key premise is to associate the same pedestrian...

AI Technical Summary

Problems solved by technology

[0003] The pedestrian re-identification problem is a very challenging problem

The second type of method uses labeled samples for supervised learning, and the effect is usually better than the first type of method, but there are still three main defects: (1) This type of method first extracts features, and then learns the distance measure for the extracted features, so , the performance of the learned distance metric is largely limited by the expressiveness and discriminativeness of hand-designed features; (2) this type of method treats the two important modules of feature extraction and distance learning in isolation, and usually only obtains a single (3) The distance metric learned by this type of method is specific to the current scene. When migrating to another new scene, the matching effect will be greatly reduced.

Images