Multi-target tracking method based on feature aggregation

A multi-target tracking and target tracking technology, applied in the field of multi-target tracking based on feature aggregation, can solve the problems of reducing network speed, lack of differentiation, and a large amount of calculation, and achieve the effect of accurate data association and improved differentiation.

Active Publication Date: 2021-09-17
ZHEJIANG LAB
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, most methods still use a separate single-object tracker for each object to capture motion information, which will lead to a large amount of calculation and greatly reduce the speed of the network when there are a large number of objects to be tracked in the video.
In addition, the appearance features of the target are also not sufficiently discriminative due to the target distortion

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
  • Multi-target tracking method based on feature aggregation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0017] The present invention will be described in detail below according to the accompanying drawings.

[0018] Such as figure 1 Shown, a kind of multi-target tracking method based on Single Shot target tracking and Identity-Aware feature aggregation of the present invention comprises the following steps:

[0019] Step 1: Read the video frame frame by frame, and the corresponding detection frame of each frame, and send them to the network.

[0020] Step 2: Use the multi-target tracking method of Single Shot target tracking and Identity-Aware feature aggregation to match and associate the detection frames between video frames. This step is the core of the present invention and is divided into the following sub-steps.

[0021] 1) Feature extraction:

[0022] Use the ResNet-34 model pre-trained on ImageNet as the basic feature extraction network to extract feature maps for adjacent frames, which includes five large convolutional layers Conv1-Conv5.

[0023] 2) Single Shot tar...

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 multi-target tracking method based on feature aggregation, and the method comprises the steps: dividing a multi-target tracking task into a Single Shot target tracking part and a Identity-Aware feature aggregation data association part, tracking all targets of a previous frame into a current frame through the Single Shot target tracking at a time, and enabling the Identity-Aware feature aggregation data association partto aggregate features of the same object of adjacent frames. Target tracking of Single Shot provided by the invention is sharing calculation of all targets, so that the speed of the multi-target tracking method is not influenced by the number of the targets; the data association of Identity-Aware feature aggregation proposed by the invention improves the discrimination of object features, so that the data association part is more accurate.

Description

technical field [0001] The invention belongs to the field of computer vision, and in particular relates to a multi-target tracking method based on feature aggregation. Background technique [0002] A recent research trend in the field of Multi-Object Tracking is to integrate motion models and appearance models into the same network framework, which allows the network to share computation and support end-to-end training. However, most methods still use a separate single-object tracker for each object to capture motion information, which will lead to a large amount of computation and greatly reduce the speed of the network when there are a large number of objects to be tracked in the video. In addition, the appearance features of objects are also not sufficiently discriminative due to object distortion. Contents of the invention [0003] The purpose of the present invention is to provide a multi-target tracking method based on feature aggregation to address the deficiencies...

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
IPC IPC(8): G06T7/246G06K9/32G06K9/62G06N3/04G06N3/08
CPCG06T7/246G06N3/08G06T2207/10016G06T2207/20076G06T2207/20081G06T2207/20084G06N3/047G06N3/045G06F18/2415G06F18/241
Inventor 储琪俞能海刘斌龚涛朱世强张鸿轩
Owner ZHEJIANG LAB
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

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap