A multi-pedestrian online tracking method based on spatio-temporal attention mechanism

A technology of attention and mechanism, applied in computer parts, instruments, biological neural network models, etc., can solve the problem of unbalanced positive and negative samples of tracking algorithms, and achieve the effect of improving accuracy and effectiveness

Active Publication Date: 2019-02-15
SHANGHAI JIAO TONG UNIV
View PDF4 Cites 36 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

First of all, to solve the problem of unbalanced positive and negative samples in the tracking algorithm, the present invention designs an objective function that fuses adaptive sa

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 multi-pedestrian online tracking method based on spatio-temporal attention mechanism
  • A multi-pedestrian online tracking method based on spatio-temporal attention mechanism
  • A multi-pedestrian online tracking method based on spatio-temporal attention mechanism

Examples

Experimental program
Comparison scheme
Effect test
No Example Login to view more

PUM

No PUM Login to view more

Abstract

A multi-pedestrian online tracking method based on spatio-temporal attention mechanism comprises the steps of pedestrian detection, pedestrian tracking and data association. The invention provides a multi-pedestrian online tracking method based on a spatio-temporal attention mechanism: Aiming at the problem of imbalance between positive and negative samples in the existing online tracking algorithm, a tracking objective function is proposed, which fuses the adaptive sample weight terms and reallocates the sample weights according to the sample loss values calculated in the training process, which improves the effectiveness of the tracking model updating; Aiming at the problem that the data association is easily disturbed by the noise samples which are occluded or shifted, In this paper, adeep neural network matching model based on temporal and spatial attention mechanism is proposed, which focuses on the correlation region and ignores the non-correlation region in the spatial domain,and focuses on the positive samples in the historical trajectory and ignores the noise samples in the temporal domain, so as to improve the accuracy of multi-pedestrian tracking.

Description

technical field [0001] The invention relates to the technical field of computer video processing, in particular to an online tracking method for multiple pedestrians based on a spatio-temporal attention mechanism. Background technique [0002] The multi-pedestrian tracking task is to calculate and track the trajectory of each pedestrian target in a video containing multiple pedestrians. This algorithm has a wide range of applications in practical scenarios, such as unmanned driving, intelligent video surveillance, ball sports analysis, etc. The challenge lies in (1) the number of pedestrian targets in the video is uncertain, and they may enter and leave the field of view at any time; (2) different pedestrians are prone to frequent interaction and occlusion, which interferes with the tracking of the target. [0003] Existing multi-pedestrian tracking methods can be divided into two categories: offline and online. The offline method takes the overall video content as input, ...

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): G06K9/00G06N3/04
CPCG06V20/42G06V20/53G06N3/045
Inventor 杨华朱继
Owner SHANGHAI JIAO TONG 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