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

A multi-source image fusion method based on reinforcement learning

A technology of reinforcement learning and fusion method, applied in the field of multi-source image fusion based on reinforcement learning, can solve the problems of poor signal reconstruction stability, different fusion precision, reliability and accuracy need to be improved, etc., to improve reliability. and accuracy, resolving image quality degradation, and stabilizing the fusion and reconstruction process

Active Publication Date: 2020-12-15
中国航天电子技术研究院
View PDF7 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] 1) In UAV aerial reconnaissance, the images acquired by multi-image sensors are susceptible to external interference such as atmosphere, light and shaking, resulting in a decrease in the quality of the acquired images;
[0007] 2) Different fusion methods have different emphases on image data, and the accuracy of fusion is also different, and the fusion results may not obtain ideal results;
[0008] 3) In the process of image decomposition, it is easy to lose information or generate redundant information, and the signal reconstruction stability in the fusion process is poor;
[0009] In the process of image fusion and reconstruction, the reliability and accuracy of the weighted weight estimation of fusion coefficients need to be improved

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-source image fusion method based on reinforcement learning
  • A multi-source image fusion method based on reinforcement learning
  • A multi-source image fusion method based on reinforcement learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0042] The invention adopts the combination of enhanced learning and wavelet transform to complete the fusion of multi-source images of drones, and the images include visible light images, infrared images, SAR images and multi-spectral images acquired by drones. Among them, image enhancement technology is firstly used to enhance the multi-source image, and then the wavelet transform technology is used to decompose the multi-source image, and the Q learning algorithm in the reinforcement learning is used to learn the weights of the decomposed high and low frequency coefficients, and finally the high and low frequency The coefficients are weighted and reconstructed to obtain the fused image. The overall process of the algorithm is as follows: figure 1 shown.

[0043] The algorithm can be summarized in the following steps:

[0044] 1) Enhance multi-source images, the processing process includes dehazing processing and Gaussian filter denoising processing;

[0045] 2) Low-pass f...

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 relates to the field of image processing, in particular to a multi-source image fusion method based on reinforcement learning. The method decomposes a multi-source image to obtain a high-frequency coefficient signal and a low-frequency coefficient signal, and performs weighted reconstruction on the high-frequency coefficient signal and the low-frequency coefficient signal to obtain a fusion image; the method adopts Q-learning in reinforcement learning The algorithm trains the weights of the high-frequency coefficient signal and the low-frequency coefficient signal, takes the weight when the fusion evaluation criterion is optimal, performs weighted fusion to obtain the fusion coefficient signal, and reconstructs the fusion image according to the fusion coefficient signal. Based on the Q-learning algorithm, the weights of high-frequency coefficient signals and low-frequency coefficient signals from multi-source image decomposition are trained, which improves the reliability and accuracy of weight estimation.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a multi-source image fusion method based on reinforcement learning. Background technique [0002] In recent years, military drones have been widely used in combat reconnaissance and precision strikes due to their unique advantages. With the development of military high-tech, the scope of the UAV battlefield has expanded to land, sea, air and electromagnetic and other multi-dimensional spaces. The battlefield environment has become increasingly complex. More and more attention has been paid to the multi-source image fusion technology that integrates the information of multiple image sensors to obtain a comprehensive and detailed description of the same target or scene. [0003] Multi-source image fusion technology is an important content of UAV intelligence processing technology. Its core idea is to integrate multi-source images obtained by multiple sensors under different conditi...

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): G06T5/50G06K9/62
CPCG06T5/50G06T2207/20221G06T2207/20064G06T2207/10004G06T2207/10048G06T2207/10032G06T2207/20024G06F18/232
Inventor 吴国强包文龙尹中义黄坤李晓明赵甲
Owner 中国航天电子技术研究院
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