Visual scene recognition method based on semantic gradient points and road power points

A technology of scene recognition and gradient points, applied in the field of visual scene recognition, to achieve the effect of improving performance, increasing weight, and eliminating interference

Pending Publication Date: 2022-04-01
ZHEJIANG UNIV
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, in extremely challenging scenes (combined with strong scene appearance changes and strong camera viewing angle changes), there is still a lot of room for improvement in its performance.

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
  • Visual scene recognition method based on semantic gradient points and road power points
  • Visual scene recognition method based on semantic gradient points and road power points
  • Visual scene recognition method based on semantic gradient points and road power points

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0038] The present invention will be further described below in conjunction with the accompanying drawings.

[0039] refer to figure 1 with figure 2 , a visual scene recognition method based on semantic gradient points and road power points, comprising the following steps:

[0040] Step 1, semantic feature extraction, using the semantic segmentation network to process the image, retaining the middle layer features, logits layer features and the final semantic label;

[0041] Step 2, semantic gradient point detection, calculate the absolute gradient sum of all channels of the logits layer feature, filter out the points with higher absolute gradient sum, and retain their position information in the image;

[0042] The steps of semantic gradient point detection are as follows:

[0043] Step 2-1, calculate the absolute gradient sum of all channels of the logits layer feature, each channel of the Logits layer (W×H×C) feature can be considered as the probability of the correspon...

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

A visual scene recognition method based on semantic gradient points and road power points utilizes the characteristic that the semantic gradient points can be repeatedly detected under strong scene appearance change, different advantages of different layer features of the semantic gradient points and the structure of the scene to improve the performance in a very challenging scene. The characteristic that the semantic gradient points can be repeatedly detected under the strong scene appearance change ensures that the intersection of the adopted characteristics between the semantic gradient points and the strong scene appearance change is correctly matched; the features of different layers of semantic gradient points are spliced together, and different characteristics of the features of different layers for scene appearance change and camera visual angle change can be utilized; the regions are divided according to the road power points, and the corresponding relationship between the regions is considered when the similarity is calculated, so that the weight of the visual overlapping part in the similarity calculation can be increased, the visual overlapping part is more focused, and the interference of irrelevant parts is eliminated.

Description

technical field [0001] The present invention relates to the technical field of Visual Place Recognition (VPR), in particular to a visual scene recognition method based on semantic gradient points and road power points. Background technique [0002] Visual Place Recognition (VPR) or loop closure detection technology is an important part of visual simultaneous localization and mapping (vSLAM). Visual scene recognition is the key to detecting closed loops, which helps to improve localization accuracy and reduce the uncertainty of the constructed maps. Traditional visual scene recognition can achieve satisfactory results in some specific scenarios, such as short-term tasks (revisiting the same place in a short period of time), a certain degree of scene appearance change, and a certain degree of perspective change. However, for challenging scenes under long-term tasks, such as scene appearance changes (different time of day, different seasons, and different lighting conditions) ...

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): G06V20/00G06V10/74G06K9/62
Inventor 潘赟包瑶琦杨哲朱怀宇
Owner ZHEJIANG UNIV
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
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap