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

A Learning Visual Tracking Method Based on Appearance Model

An appearance model and visual tracking technology, applied in the field of computer vision, can solve problems such as affecting the accuracy of the results, and achieve the effects of excellent experimental results, improved computing efficiency, and improved accuracy.

Active Publication Date: 2018-12-18
NANJING SINCERE CREATIVE DIGITAL TECH CO LTD
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, in practical applications, there is inevitably some background noise inside the bounding box, especially for non-rigid objects, which affects the accuracy of the results.

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 Learning Visual Tracking Method Based on Appearance Model
  • A Learning Visual Tracking Method Based on Appearance Model
  • A Learning Visual Tracking Method Based on Appearance Model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0022] Attached below figure 1 The present invention will be further described.

[0023] A learning type visual tracking method based on an appearance model, characterized in that the method comprises the following steps:

[0024] Step 1: Learning the foreground background set: over-segment the frame in training, the whole process conforms to the (MIL) paradigm, use the positive bag to model the target inside the bounding box, and use the negative bag set to represent the background; through the positive Negative packet information, which can estimate the position of the target;

[0025] Assuming that the positive package set is independent from the negative package set, the division of confidence can be determined by the following method:

[0026] C(r i )=p(l(r i )=1|B + )p(l(r i )=1|B - ), where B + and B - Represent positive and negative envelope sets;

[0027] In order to remove the set of negative superpixels in the bounding box, it is necessary to calculate the...

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 provides a learning-type visual tracking method based on an appearance model. The method comprises the following steps: firstly, partitioning an image frame into a plurality of super-pixel regions; secondly, gathering the partitioned super-pixel regions into a positive packet set and a negative packet set, wherein the positive packet set and the negative packet set can be interpreted as accurate foreground partition and background partition; and lastly, learning the distribution of packets by a greedy search algorithm, wherein weights of the packets depend on the magnitude of significance. Compared with an existing modeling method based on super-pixels, the modeling method has the advantages that iteration is not used in a learning process, and the appearance model is modelled through a multi-instance learning task, so that limitations of a modeling method based on a bounding box are overcome; the calculation efficiency is increased; and the learning-type visual tracking method can be suitable for a real-time target tracking application. Moreover, a two-step process is provided to partition confidence, so that the effectiveness of confidence partition is ensured, and the target tracking accuracy is increased greatly.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to a learning-type visual tracking method based on an appearance model. Background technique [0002] Visual tracking has a wide range of applications, including video surveillance and human-computer interaction. Accurate and efficient visual tracking is still a challenging problem because of large non-rigid deformations, target appearance changes, severe occlusions, and unknown camera motions. Visual tracking falls into two categories: discriminative recognition and generative recognition. Discriminative recognition is to regard the tracking task as a binary classification problem of dividing foreground and background, learn a classifier according to the current frame and classify the subsequent frames. The generative recognition method learns the target appearance model according to the previous target, and when a new frame arrives, the candidate sample closest to the target model...

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): G06T7/246
CPCG06T7/20G06T2207/10004G06T2207/10016G06T2207/20081
Inventor 周瑜明安龙廖鸿宇孙放
Owner NANJING SINCERE CREATIVE DIGITAL TECH CO LTD
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