Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

A pedestrian re-recognition method based on semantic region expression

A pedestrian re-identification and area technology, applied in the field of image processing and computer vision, can solve the problems of pedestrian misalignment, not considering two images, not considering interference factors, etc., to avoid interference, improve robustness and reliability, Improve indexability

Inactive Publication Date: 2019-02-15
TIANJIN UNIV
View PDF10 Cites 11 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] The methods in the prior art often compare the target similarity based on the global features of the image, but do not consider the background information and other interference factors contained in the global features; the existing methods usually calculate the similarity of the same position between image pairs, and then The sum of similarities is performed without considering the misalignment of pedestrians in the two images

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A pedestrian re-recognition method based on semantic region expression
  • A pedestrian re-recognition method based on semantic region expression
  • A pedestrian re-recognition method based on semantic region expression

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0030] The embodiment of the present invention proposes a pedestrian re-identification method based on semantic region expression, see figure 1 , the method includes the following steps:

[0031] 101: According to the high-level semantic information of pedestrians, extract the features of different body parts in the foreground of the image, and combine the global features and local features of the image to obtain a more complete feature description of pedestrians;

[0032] 102: Use the triplet loss to train the metric function, so that the distance between the features of the same pedestrian in the new mapping space is as small as possible, and the distance between the features of different new people is as large as possible.

[0033] This method mainly includes five parts: semantic region extraction, image feature extraction within the semantic region, image global and local region feature extraction, calculation of similarity between image samples, and training of metric lea...

Embodiment 2

[0036] The scheme in embodiment 1 is further introduced below in conjunction with specific calculation formulas and examples, see the following description for details:

[0037] 1. Image Feature Extraction in Semantic Regions

[0038] The pedestrian re-identification algorithm first uses the pedestrian detection algorithm to detect the pedestrians in the surveillance video, and then finds the correct matching target according to the similarity between the images. Since the rectangular frame obtained by the pedestrian detection algorithm contains certain background information in addition to the pedestrian, if the global feature of the image is used for matching, this feature will contain a lot of background information interference, which will bring a greater impact on the matching accuracy. influences.

[0039] In addition, due to factors such as viewing angle and pedestrian pose transformation, the positions of pedestrians between images are not aligned. It is possible that...

Embodiment 3

[0074] Combine below figure 1 The scheme in embodiment 1 and 2 is carried out feasibility verification, see the following description for details:

[0075] figure 2 The test results of this method on the general pedestrian re-identification dataset VIPeR are given. The vertical axis is the CMC cumulative matching score, and the horizontal axis is the ranking. Among them, Rank1 represents the probability that the correctly matched target ranks first, and Rank10 represents the probability that the correctly matched target ranks in the top ten. from figure 2 As can be seen in , by utilizing the semantic component information of pedestrians, this method achieves the best results on different evaluation metrics.

[0076] On Rank1, the current best performance results under this data set have been achieved, and on Rank10, an accuracy of more than 90% has been achieved, indicating that the vast majority of correctly matched samples will appear in the top ten of the retrieval res...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a pedestrian re-identification method based on semantic region expression, including the following steps of giving an image, detecting the positions of different parts of the pedestrian in the image to obtain a set comprising different component regions, according to the high-level semantic information possessed by pedestrians, extracting and cascading the features in eachcomponent region by using a scale-invariant local pattern descriptor and a color histogram descriptor, and then reducing the original features in dimension by principal component analysis to obtain the feature set of component region; according to the feature set of the component region, obtaining a more complete feature descriptor for pedestrians by combining the global feature and local featureof the image. By using the triple loss training function, the original eigenvector can be mapped into the new eigenspace and the separability between samples can be improved. The similarity between pedestrian image features is calculated according to the metric matrix obtained by learning, so that the pedestrian re-recognition can be realized. The method of the invention not only avoids the interference of the image background, but also realizes the comparison of similarities between corresponding regions, and effectively improves the robustness and reliability of the features.

Description

technical field [0001] The invention relates to the technical fields of image processing and computer vision, in particular to a pedestrian re-identification method based on semantic region expression. Background technique [0002] With the increasing demand for smart cities and public safety, intelligent video surveillance systems have received extensive attention and research, and have been applied to many fields such as security and industrial production, playing an important role in daily life. In the field of intelligent video surveillance, the main object of concern is pedestrians, and the most basic task is to carry out long-term and stable tracking of targets in a large-scale video surveillance network. However, due to considerations of privacy and maintenance costs, it is difficult for the camera monitoring network to cover the entire area, resulting in a monitoring blind spot. When the target passes through the monitoring blind spot, it is impossible to continuousl...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/46G06K9/62
CPCG06V10/56G06F18/22
Inventor 雷建军牛力杰郑泽勋彭勃罗晓维郭琰
Owner TIANJIN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com