Eureka AIR delivers breakthrough ideas for toughest innovation challenges, trusted by R&D personnel around the world.

Compressed object imaging method and system based on deep learning

A technology of deep learning and imaging methods, applied in image coding, image data processing, instruments, etc., can solve the problems of information transmission speed and information processing speed limit, restrict development, and affect the real-time performance of the system, so as to improve real-time performance and solve problems. Storage and transmission problems, the effect of saving network bandwidth and storage overhead

Active Publication Date: 2020-04-17
SOUTH CHINA NORMAL UNIVERSITY
View PDF5 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] However, most of the target recognition and detection methods currently proposed adopt the method of collecting first and then processing, which will cause the system to analyze and process a large amount of useless information, affecting the real-time performance of the system, and the massive data collected will be lost during storage and transmission. Put enormous pressure on system hardware and network bandwidth
Obviously, this method will be limited by the speed of information transmission and information processing, which will become a bottleneck restricting its development.

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
  • Compressed object imaging method and system based on deep learning
  • Compressed object imaging method and system based on deep learning
  • Compressed object imaging method and system based on deep learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0041] see figure 1Part1 and Part3, which are the deep learning network structure diagram based on the group full convolutional neural network of the embodiment of the present invention.

[0042] In this deep learning network, Part1 is the input of the image and the feature measurement matrix part of the extracted target object. The arrow with the word CS Layer indicates the convolutional layer designed according to the theory of compressed sensing. After the network training is completed, the convolutional layer’s The parameters will be used as the feature measurement matrix in the compressed sensing theory to achieve efficient measurement of specific target objects. The size and number of filters used as the first convolutional layer are manually adjusted according to the needs of samples and compression rates, which are hyperparameters. Part3 is the reconstructed sub-network of the target object and the output part of the image. The arrows with the words conv0, conv1, conv...

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 a compressed object imaging method and a system based on deep learning. The method comprises the following steps: training a deep learning network of a sample image to obtain afeature measurement matrix of a target object and a reconstruction sub-network of the target object, performing optical compression sampling on the target object, and importing the target object intothe reconstruction sub-network to perform target object reconstruction to complete compressed object imaging. According to the method, the limitation that only the whole shot scene is imaged by the traditional imaging technology is broken through; only the required specific target object is extracted; according to the technical scheme, the target object is separated from the background, other background objects are filtered out, optical compression imaging of the specific target object is completed, real-time monitoring, recognition and tracking of the target are facilitated, meanwhile, occupation of network bandwidth and storage expenditure of data in the transmission process are greatly reduced, the storage and transmission problems of mass data are solved, and the real-time performanceof the system is improved.

Description

technical field [0001] The present invention relates to technical fields such as optical imaging, computational imaging, target detection, recognition and tracking, and artificial intelligence applications, and in particular to a compression object imaging method and system based on deep learning. Background technique [0002] In complex physical scenes, how to simulate the biological visual system, quickly capture important image information, process and understand it, and actively screen out the target objects of interest has always been a research direction that scholars at home and abroad are keen on, and has important research value. . Especially in the digital age where a large amount of data streams are generated every day, how to make full use of the characteristics of high-speed computing of computers to quickly capture important image information and perform corresponding processing and understanding in the face of complex physical scenes, and actively Screening o...

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): G06T9/00G06N3/04
CPCG06T9/002G06N3/045Y02T10/40
Inventor 李军梁创学李玉慧王尚媛雷苗
Owner SOUTH CHINA NORMAL UNIVERSITY
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Eureka Blog
Learn More
PatSnap group products