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

A video target detection method based on a depth feature pyramid and tracking loss

A feature pyramid, target detection technology, applied in image data processing, instruments, computing and other directions, to achieve rich semantic information, increase detection speed, high dimensional effects

Active Publication Date: 2019-05-21
BEIJING INSTITUTE OF TECHNOLOGYGY
View PDF5 Cites 7 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] In view of this, the present invention provides a video target detection method based on deep feature pyramid and tracking loss, which solves the problems of stability and accuracy of video target detection

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 video target detection method based on a depth feature pyramid and tracking loss
  • A video target detection method based on a depth feature pyramid and tracking loss
  • A video target detection method based on a depth feature pyramid and tracking loss

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0041] The present invention will be described in detail below with reference to the accompanying drawings and examples.

[0042] Step 1. Construct a deep neural network structure based on deep feature pyramid and tracking loss. The network structure is a Siamese twin network. The Siamese twin network is two identical neural networks. Each neural network contains a basic neural network and a feature pyramid.

[0043] The way to obtain the feature pyramid is: select a certain convolutional layer in the basic network according to the size of the video target, convolve the subsequent layer based on the convolutional layer, obtain the second subsequent layer based on the convolution of the subsequent layer, and obtain the second subsequent layer based on the second subsequent layer. Convolute successively to obtain each subsequent layer, and obtain a feature pyramid based on the convolution layer and each subsequent layer;

[0044] Furthermore, the feature pyramid is composed of 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 video target detection method based on a depth feature pyramid and tracking loss, and solves the problems of video target detection stability and accuracy. According to the method, a multi-scale feature map method is adopted in the target detection process, and the bottom feature map is high in dimension, rich in detail information, low in high-level feature latitude and rich in semantic information, so that the spatial information of the video image can be better utilized, and the detection can adapt to multi-scale and multi-type targets. According to the method, a multi-scale candidate window generation mode is adopted, small targets are subjected to dense sampling, large targets are subjected to discrete sampling, and independent processing, small-scale target fine sampling, large-scale target coarse sampling and detection speed increase are carried out according to target scales in consideration of different precisions required when targets with different scales are detected.

Description

technical field [0001] The invention belongs to the technical field of video target detection, and relates to a video target detection method based on feature pyramid and tracking loss. Background technique [0002] With the popularization of image acquisition equipment, especially the large-scale deployment of monitoring equipment, the resolution of video data obtained is getting higher and higher, and the demand for video image processing caused by this is also getting higher and higher. Among them, target detection in video data is an important research direction. Video target detection is of great significance for applications such as security protection, crowd density statistics, and target re-identification. At the same time, compared with image data, video data has more time-dimensional data, and the amount of data increases geometrically compared with image data, which brings challenges to the target detection of video data. [0003] In the field of video target det...

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): G06T7/246
Inventor 赵保军赵博雅唐林波王文正邓宸伟
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