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

A Manifold Regularization Correlation Filtering Target Tracking Method Based on Augmented Samples

A technology of correlation filtering and target tracking, applied in image enhancement, image analysis, instruments, etc., can solve problems such as inaccurate classification and unconsidered manifold space structure, so as to improve classification accuracy performance and discrimination ability , the effect of enriching the number of training samples

Inactive Publication Date: 2020-06-26
BEIJING INSTITUTE OF TECHNOLOGYGY
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0013] The purpose of the present invention is to solve the problem of inaccurate classification caused by the fact that the existing tracking method based on correlation filtering does not consider the manifold space structure and the tracking drift problem caused by the negative base samples around the target area. A Manifold Regularized Correlation Filtering Target Tracking Method Based on Samples

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 Manifold Regularization Correlation Filtering Target Tracking Method Based on Augmented Samples
  • A Manifold Regularization Correlation Filtering Target Tracking Method Based on Augmented Samples
  • A Manifold Regularization Correlation Filtering Target Tracking Method Based on Augmented Samples

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0041] S1. Extract positive base samples from the target area in the previous frame image, extract negative base samples from the non-target area, and extract unmarked base samples from the target area of ​​the previous frame image in the current frame image; these three types Base samples form an augmented base sample set;

[0042] Wherein, the positive base samples extracted from the target area in the previous frame image, and the negative base samples extracted from the non-target area are marked base samples; the extracted base samples can be gradient histogram features, grayscale features , depth features;

[0043] During specific implementation, the image frame comes from the RGB image collected by video surveillance; one image block of the target area in the previous frame image and two image blocks of the same size adjacent to the target on the left and right sides of the target area can be extracted, and in An image block is extracted from the area where the target ...

Embodiment 2

[0097] The target tracking method is still effective if the RGB image collected by video surveillance in Embodiment 1 is changed to an infrared image collected by an infrared device, or image data obtained by other specific video collection devices or continuous image generation devices.

Embodiment 3

[0099] The present invention is still effective if the gradient histogram feature in Embodiment 1 is changed to grayscale feature, Lab color feature, HSV color feature, feature map in the neural convolutional network process, and other global image features.

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 an augmented sample-based manifold regularization correlation filtering target tracking method. The method comprises the steps of S1, extracting a positive basic sample and a negative basic sample in a target region and a non target region of a previous frame respectively and extracting an unmarked basic sample to form an augmented basic sample set; S2, generating a mark matrix according to an output of S1; S3, by utilizing outputs of S1 and S2, in combination with block circulant structures of a kernel matrix and a Laplacian matrix, learning a least square correlation filtering classification model of manifold regularization; S4, judging whether a current frame is a second frame or not and performing corresponding operation; S5, determining marks of all samples, generated by performing operation on the unmarked basic sample in S1 by utilizing S4, by adopting a quick block detection algorithm, and determining a current target position; S6, judging whether the current frame is the last frame or not, deciding to jump to S1 or S7; and S7, outputting a target state of each frame. According to the method, the unmarked sample is predicted in a semi-supervised manner, so that the classification accuracy of the correlation filtering classification model is remarkably improved; and the method can be applied to a real-time system.

Description

technical field [0001] The invention relates to a manifold regularization correlation filtering target tracking method based on augmented samples, and belongs to the technical fields of computer vision, pattern recognition, human-computer interaction, video monitoring and image compression. Background technique [0002] Object tracking is an important frontier topic in the field of computer vision, and it is one of the focuses of academia and industry. It aims to locate the object of interest in the scene from a video or image sequence, and estimate the motion state of the object, including position, scale, rotation angle, etc. Robust and accurate object tracking can provide support and input for high-level computer vision tasks such as human motion analysis, event detection, behavior and scene understanding, thus promoting the development of computer vision itself. At the same time, in terms of practical applications, due to the rapid development of software and hardware t...

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 Patents(China)
IPC IPC(8): G06T7/20
CPCG06T2207/10016G06T2207/20056
Inventor 马波胡宏伟
Owner BEIJING INSTITUTE OF TECHNOLOGYGY
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