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

Deep learning image segmentation method and system based on edge feature extraction

A deep learning and image segmentation technology, which is applied in image analysis, image enhancement, image data processing, etc., can solve the problems of difficult application, low portability, and increased network calculation, and achieve the effect of preventing danger and high segmentation accuracy

Pending Publication Date: 2021-10-22
DALIAN NATIONALITIES UNIVERSITY
View PDF2 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The above networks use complex convolutional network structures to extract edge features by fusing features of different depths, which can easily cause a large increase in network calculations, resulting in low portability and difficult application of these methods.

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
  • Deep learning image segmentation method and system based on edge feature extraction
  • Deep learning image segmentation method and system based on edge feature extraction
  • Deep learning image segmentation method and system based on edge feature extraction

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0092] Example 1: High Complexity Environment

[0093] In this embodiment, for a high-complexity environment (a visual environment with many objects and complex spatial positional relationships between objects, and a lot of edge information, such as theaters, classrooms, streets, etc.), the high-complexity traffic scene is input to the edge feature-based The extracted deep learning image segmentation system obtains the image segmentation results as shown in 6.

Embodiment 2

[0094] Example 2: Low Complexity Environment

[0095] In this embodiment, aiming at low-complexity environments (visual environments with few objects and simple spatial relationship between objects, and less edge information, such as scenes such as desks, sky, grass, etc.), low-complexity traffic scenes are input into the edge-based feature-based The extracted deep learning image segmentation system obtains image segmentation results such as Figure 7 shown.

Embodiment 3

[0096] Example 3: Common environment

[0097] In this embodiment, aiming at common environments (daily work and life scenes, such as offices, basketball courts, parks, etc.), the traffic scenes under common environments are input to the deep learning image segmentation system based on edge feature extraction, and the image segmentation results are obtained as follows: Figure 8 shown.

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 deep learning image segmentation method and system based on edge feature extraction, and the system comprises an edge convolutional network which is used for extracting the edge features of an input image, and comprises a semantic edge convolutional network and an instance edge convolutional network, the semantic edge convolutional network processes background filling objects such as sky and roads in the image, and the instance edge convolutional network processes foreground instance objects such as figures and vehicles in the image; the feature extraction network comprises a same-side convolutional feature extraction network and a different-side convolutional feature extraction network, and the same-side convolutional feature extraction network realizes layer-by-layer extraction of features, retains feature attributes under different levels, and forms a complete feature level system; and according to the different-side convolution feature extraction network, cross-level feature fusion is realized through a connection method of changing a convolution structure and a shortcut, so that the learning potential of the feature extraction network is stimulated, and the feature extraction capability is improved.

Description

technical field [0001] The invention relates to the field of image segmentation in computer vision applications, in particular to a deep learning image segmentation method and system based on edge feature extraction. Background technique [0002] Image segmentation is a basic technology for scene understanding, and it plays a vital role in unmanned systems such as intelligent driving, autonomous navigation at the cognitive level of robots, drone landing systems, and intelligent security monitoring. The purpose of image segmentation is to divide each pixel in the image into a semantic category and an instance label, generate a global and unified segmented image, and realize pixel-level segmentation. The key to image segmentation is to identify the edge pixels of objects in the image and distinguish different objects, so the extraction of edge features is crucial to the image segmentation task. [0003] The traditional edge feature extraction method focuses on the texture gra...

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): G06T7/12G06T7/13
CPCG06T7/12G06T7/13G06T2207/20081G06T2207/20084G06T2207/30196G06T2207/30232G06T2207/30236
Inventor 杨大伟任凤至毛琳张汝波
Owner DALIAN NATIONALITIES UNIVERSITY
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