Group pedestrian re-identification method based on hybrid attention mechanism

A pedestrian re-identification and attention technology, applied in the field of computer vision, can solve the problems of time-consuming matching process, inapplicability to large data sets and real-world scenarios, etc.

CN110751018AInactive Publication Date: 2020-02-04SHANGHAI JIAO TONG UNIV
4 Cites 12 Cited by

Patent Information

Authority / Receiving Office
CN · China
Current Assignee / Owner
Publication Date
2020-02-04
Estimated Expiration
Not applicable · inactive patent

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention provides a group pedestrian re-identification method based on a hybrid attention mechanism. The method comprises the following steps: carrying out preliminary feature extraction on a group image by using a deep convolutional neural network backbone model; further extracting the preliminarily extracted features by using a hybrid attention mechanism model; and comparing and evaluatingthe features of the mixed attention model by using a least square residual distance. According to the invention, various challenges existing in a group pedestrian re-identification problem are fully considered; a hybrid attention model including spatial attention and channel attention is utilized, so that the network pays more attention to key areas and features of group images, a novel least square residual distance based on a least square algorithm is proposed, and measurement between group image pairs is learned better. The method can well adapt to various challenges brought by group pedestrian images, and has good diversity and general applicability.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The invention belongs to the technical field of computer vision, in particular to a group pedestrian re-identification method based on a mixed attention mechanism, and relates to group-focused pedestrian re-identification under non-overlapping surveillance cameras. Background technique

[0002] In recent years, people have paid more and more attention to public safety issues, and the degree of attention to safety issues has risen to a new height. The video surveillance network in the city is becoming more and more perfect, and video surveillance is spread all over the city. The quality and quality of video surveillance data The number has greatly increased. In all parts of the video surveillance network, considering the flexibility and variability of pedestrian activities and the significance of monitoring pedestrians for protecting personnel safety and criminal investigation and solving crimes, pedestrians are one of the key objects of surveillance vi...

Examples

Embodiment Construction

[0054] The embodiments of the present invention are described in detail below in conjunction with the accompanying drawings: the present embodiment is implemented on the premise of the technical solution of the present invention, and detailed implementation methods and specific operating procedures are provided, but the protection scope of the present invention is not limited to the following Example.

[0055] An embodiment of the present invention provides a method for group pedestrian re-identification based on a hybrid attention mechanism, including the following steps:

[0056] S1: Based on the deep convolutional neural network, the backbone model feature extraction network P for the group pedestrian re-identification task is formed, and the backbone model feature extraction network P obtained by applying the images on the entire group pedestrian re-identification dataset is applied to the group pedestrian re-identification task. For each image s of a group of pedestrians,...