Unlock instant, AI-driven research and patent intelligence for your innovation.

Cross-domain remote sensing image target detection method based on style content decoupling

A target detection and remote sensing image technology, applied in the field of cross-domain remote sensing image target detection, can solve the problems of poor model generalization and poor remote sensing image target detection performance, and achieve the effect of good generalization.

Pending Publication Date: 2021-11-16
DALIAN UNIV OF TECH
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

This requires sufficient multi-domain images and the use of transfer learning to improve the generalization of the model; the model obtained by this method has poor generalization and poor object detection performance for remote sensing images of different domains

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
  • Cross-domain remote sensing image target detection method based on style content decoupling
  • Cross-domain remote sensing image target detection method based on style content decoupling
  • Cross-domain remote sensing image target detection method based on style content decoupling

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0043] DETAILED DESCRIPTION OF THE INVENTION The specific embodiments of the present invention will be further described below with reference to the drawings and technical solutions.

[0044] First, the decoupling and reorganization of the content and style between different domain images are obtained by the normalized method of adaptive instance. The process of decoupled recombination of content and style needs to be supervised by the formula (12). An image of the content and style interchangeably can then be obtained, and the characteristic extractor of the expression capability is obtained, and the corresponding content feature encoding is obtained by this method. Then, the content content feature encoding the decoder portion of the YOLO network is subjected to target detection, which requires the use of the formula (7) to supervise.

[0045] In general, the method designs a multi-domain remote sensing image target detection method based on content-style decoupling, which can e...

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 belongs to the technical field of image information processing, and provides a cross-domain remote sensing image target detection method based on style content decoupling. According to the end-to-end training cross-domain remote sensing image target detection method, a multi-domain image is obtained by using an adaptive instance normalization mode, and content feature codes obtained by decoupling are further sent to a target detection network for target detection. The method has good generalization for remote sensing images of different domains, and an ideal target detection result can be obtained.

Description

Technical field [0001] The present invention belongs to the technical field of image processing information, particularly to a method of cross-domain Remote Sensing Target. Background technique [0002] Currently, the method related to the present patent includes two aspects: a first target detection algorithm is based on sensing image depth study; second depth learning algorithm is decoupled style content. [0003] Target detection method based on the depth of learning is mainly divided into two categories: a stage target detection (one-stage) and two-stage target detection (two-stage) categories. A target phase detection algorithm faster but relatively low accuracy, the two-stage target detection algorithm is slow but the accuracy is relatively high. Most of the design of this particular network architecture may be used to extract the target region, Ren et al. In: Network proposed "Faster R-CNN Towards Real-Time Object Detection withRegion Proposal Networks" in on the design fo...

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): G06K9/00G06N3/04G06N3/08
CPCG06N3/08G06N3/045
Inventor 赵文达杨瑞凯祝嘉文徐从安姚力波刘瑜何友卢湖川
Owner DALIAN UNIV OF TECH