Pedestrian re-identification method based on improved YOLOv3 network and feature fusion

A pedestrian re-identification and feature fusion technology, applied in the field of pedestrian re-identification, can solve problems such as large computational load, interference, and reduced detection accuracy, so as to improve accuracy, improve generalization ability and robustness, and reduce the amount of network parameters. Effect

Active Publication Date: 2020-10-16
XIDIAN UNIV
View PDF6 Cites 12 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] Traditional pedestrian re-identification methods rely on manual image features, such as color, Haar-like features, histogram HOG, scale-invariant feature transformation SIFT, local binary mode LBP and local maximum occurrence LOMO. These methods ignore the target sensitive area. Susceptible to interference from non-target areas, resulting in reduced detection accuracy and a large amount of computation

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
  • Pedestrian re-identification method based on improved YOLOv3 network and feature fusion
  • Pedestrian re-identification method based on improved YOLOv3 network and feature fusion
  • Pedestrian re-identification method based on improved YOLOv3 network and feature fusion

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0042] Embodiments and effects of the present invention will be further described in detail below in conjunction with the accompanying drawings.

[0043] refer to figure 1 The implementation steps of this example are as follows:

[0044] Step 1. Construct a dataset of pedestrian image training samples.

[0045] Collect at least 30,000 images with no less than 64×128 pixels, the images must contain pedestrian targets, and each pedestrian is captured by at least two cameras;

[0046] Manually mark the bounding box of pedestrians in each image containing pedestrians, and draw pedestrian detection rectangles;

[0047] The collected pedestrian images are constructed into a training data set and a verification data set in a ratio of 3:1.

[0048] Step 2, build an improved YOLOv3 pedestrian detection network.

[0049] 2.1) The feature extraction network Darknet-53 in the YOLOv3 network is pruned and optimized to obtain an improved feature extraction network Darknet-37:

[0050]...

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 an improved YOLOv3 network and feature fusion, and mainly solves the problems of low retrieval precision and low speed of specific pedestrians in a video monitoring scene in the prior art. According to the scheme, the method comprises the steps of 1) constructing a pedestrian picture data set; 2) establishing an improved YOLOv3network; 3) establishing a pedestrian re-identification network fusing global features and multi-scale local features; 4) training an improved YOLOv3 network and a pedestrian re-identification network by using the data set; 5) fusing the two networks trained in the step 2) and the step 3) to obtain a pedestrian re-identification system; and 6) inputting the monitoring video and the to-be-retrieved target pedestrian picture into a pedestrian re-identification system, retrieving the to-be-retrieved target pedestrian, and outputting a re-identification result of the target pedestrian. Accordingto the method, the sensitivity to pedestrians with different postures is enhanced, the retrieval speed and precision of pedestrian re-identification are improved, and the method can be used for regional security and protection, criminal investigation, video monitoring and behavior understanding.

Description

technical field [0001] The invention belongs to the field of computer vision, and in particular relates to a pedestrian re-identification method, which can be used for regional security, criminal investigation, video monitoring and behavior understanding. Background technique [0002] In recent years, more and more cameras have been deployed in public places. How to realize the intelligent analysis and application of massive video data with the help of artificial intelligence technology has become the key to building intelligent security. Among them, pedestrian re-identification ReID is one of the core topics. [0003] With the development of face recognition technology, the industry began to seek the possibility of more technical applications. Pedestrian re-identification technology is an important supplement and extension of face technology. More and more Internet giants and technology unicorns are beginning to realize its importance and gradually invest resources in the...

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): G06K9/00G06K9/46G06N3/04G06N3/08
CPCG06N3/08G06V40/20G06V20/40G06V10/464G06N3/045Y02T10/40
Inventor 姬红兵段育松张文博李林臧博
Owner XIDIAN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products