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

Image scene recognition method and device

A scene recognition and scene technology, applied in the field of image processing, can solve the problems of reduced recognition accuracy, low applicability, and reduced scene recognition accuracy, and achieve the effect of improving accuracy and applicability

Active Publication Date: 2016-07-27
BEIJING QIYI CENTURY SCI & TECH CO LTD
View PDF3 Cites 19 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] In the prior art, scene recognition is performed by analyzing the global features of the image, but this method is only suitable for scene recognition with a small number of categories, such as identifying night scenes and non-night scenes
Once encountering multi-scene recognition or scene recognition in a complex background, the recognition accuracy of this scheme will drop significantly
For example, if the background of the image is relatively complex, some of the areas will not increase the salience of a certain scene category among the scene categories corresponding to the image as the scene label of the image, but will easily make the scene category different from other scene categories. categories are confused, thereby reducing the accuracy of scene recognition
[0004] Moreover, in practical applications, more and more scene recognition may involve various images or photos at night, backlighting, scenery, people, etc., which further leads to the increasing applicability of the above-mentioned existing scene recognition methods. Low

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
  • Image scene recognition method and device
  • Image scene recognition method and device
  • Image scene recognition method and device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0067] refer to figure 1 , which shows a flow chart of the steps of an embodiment of an image scene recognition method according to an embodiment of the present invention, which may specifically include the following steps:

[0068] Step 110, acquire the original image.

[0069] In practical applications, the image search engine can facilitate users to quickly search for the images they need by adding tags to the images. For example, a user can search for images by inputting keywords, and if the tags of some images in the image search engine are the same or close to the user's search keywords, then the corresponding images can be recommended to the user. Among them, tagging pictures is a very important process. What is disclosed in this application is an image scene recognition method, so it is first necessary to obtain the original image for scene recognition.

[0070] An image is the material reproduction of human visual perception, and the image is composed of a series o...

Embodiment 2

[0102] refer to Figure 6 , which shows a flow chart of the steps of an embodiment of an image scene recognition method according to an embodiment of the present invention, which may specifically include the following steps:

[0103] Step 210, acquire the original image.

[0104] Step 220, using the first classification model to calculate a first confidence level and an activation map corresponding to each scene category of the original image; the activation map includes an activation parameter of each pixel.

[0105] Step 230, sort the first confidence levels, and extract an activation map corresponding to the scene category to which at least one first confidence level ranked higher belongs.

[0106] Step 240, according to the preset threshold and the extracted activation parameters of each pixel in each activation map, calculate the first salient regions corresponding to different scene categories in the original image.

[0107] As described in Embodiment 1, each scene cat...

Embodiment 3

[0148] refer to Figure 7 , which shows a structural block diagram of an embodiment of an image scene recognition device according to an embodiment of the present invention, which may specifically include the following modules:

[0149] An image acquisition module 310, configured to acquire an original image.

[0150] The first classification module 320 is configured to use a first classification model to calculate a first confidence level and an activation map corresponding to each scene category of the original image; the activation map includes an activation parameter of each pixel.

[0151] The activation map extraction module 330 is configured to sort the first confidence levels, and extract the activation map corresponding to the scene category to which at least one first confidence level ranked higher belongs.

[0152] A salient area calculation module 340, configured to calculate a second salient area of ​​the original image according to the extracted activation map. ...

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 an image scene recognition method and device and relates to the technical field of image processing.The method includes the steps that an original image is obtained; first confidence degrees and activation images corresponding to various scene categories of the original image are calculated by means of a first classification model, wherein the activation images include activation parameters of pixels; the first confidence degrees are sequenced, and the activation images which rank higher in the sequence and correspond to the scene categories where at least one first confidence degree belongs are extracted; according to the extracted activation images, a second salient region of the original image is calculated; on the basis of the second salient region of the original image, the scene label of the original image is calculated by means of a second classification model.The method and device solve the problem that a scene recognition method in the prior art is low in accuracy and adaptability, scene recognition can be conducted according to features of salient regions of extracted images, accuracy of scene recognition is improved, and adaptability of scene recognition is further improved.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an image scene recognition method and an image scene recognition device. Background technique [0002] Scene recognition is a specific sub-problem in the image recognition problem in a broad sense. Its goal is to give an image or photo, and the computer automatically judges which specific scene the image or photo belongs to (including but not limited to outdoor Scenes such as beach, forest, city, etc., indoor scenes such as living room, bedroom, gym, etc., and event scenes such as football matches, parades, concerts, etc.). Nowadays, scene recognition technology plays a very important role in video surveillance, social network user behavior mining, etc., and therefore has received extensive attention and research. [0003] In the prior art, the scene recognition is performed by analyzing the global features of the image, but this method is only suitable for scene recogn...

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): G06K9/00
CPCG06V20/35G06V20/10G06V10/25G06V10/40
Inventor 刘玉明
Owner BEIJING QIYI CENTURY SCI & TECH CO LTD