Multi-target image retrieval method

An image retrieval, multi-target technology, applied in the field of multi-target image retrieval, can solve the problem of inability to detect the content of bricks, and achieve the effect of speeding up the retrieval speed

Inactive Publication Date: 2019-10-01
NORTHWEST UNIV
View PDF5 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a rapid detection method for brick content in recy

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
  • Multi-target image retrieval method
  • Multi-target image retrieval method
  • Multi-target image retrieval method

Examples

Experimental program
Comparison scheme
Effect test

Example Embodiment

[0031] Example 1:

[0032] This embodiment provides a multi-target image retrieval method, which includes the following steps:

[0033] Step 1. Input I images, take any one of the I images as image i, i=1, 2,...,I; extract M candidate targets in image i;

[0034] The deep network model constructed is: on the basis of a typical VGG-16, a number of convolutional layers of different scales (19×19×1024, 10×10×512, 5×5×256, 3×3× 256, 1×1×256), each convolutional layer extracts candidate targets through a 3×3 sliding window, where the stride is 1 and the padding is 0;

[0035] In this way, the extracted candidate targets are generated on the feature map instead of on the original image. Compared to extracting from the original image, their size is much smaller. This helps to increase efficiency and keep the model compact.

[0036] The method for obtaining candidate targets in the present invention is "W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C.-Y. Fu, and ACBerg, "Ssd: Single sh...

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 provides a multi-target image retrieval method. The multi-target image retrieval method comprises the following steps: 1, extracting M candidate targets in an image i; step 2, obtainingan initial prediction label of the candidate target m; step 3, updating the initial prediction tag of each candidate target in the M candidate targets to obtain an optimized prediction tag of each candidate target; step 4, obtaining target hash codes of all candidate targets in the I images according to the optimized prediction tags of the candidate targets, and constructing an index database; andstep 5, inputting an image to be retrieved by the user, extracting an interest target in the image to be retrieved, mapping the interest target into a target hash code, inputting the target hash codeinto an index database to find a key code Key corresponding to the target hash code, and returning an image ID in a linked list corresponding to the key code Key to the user. The invention designs anindexing method based on an inverted index and a quantization thought, so that the retrieval speed is further increased.

Description

technical field [0001] The invention belongs to the technical field of image retrieval, and in particular relates to a multi-target image retrieval method. Background technique [0002] An important content of user visual interest object retrieval is to accurately detect or even identify interest objects in visual images. Such as figure 1 As shown, the supervised target detection algorithm usually needs to manually mark the specific position of the target in the picture, and then use the classifier to learn. Because manual labeling is very time-consuming and labor-intensive, and is easily affected by subjective assumptions, it is difficult to promote and apply. The weakly supervised learning method only needs to know the type information at the image level, that is, only knows whether the image contains the target object, but does not need to know the specific location of the object in the image. This learning method usually measures the similarity between features by dis...

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
IPC IPC(8): G06K9/00G06K9/32G06K9/62G06T9/00
CPCG06T9/00G06V20/00G06V10/25G06V2201/07G06F18/23213G06F18/214
Inventor 赵万青元莉伟侯勇管子玉罗迒哉彭进业
Owner NORTHWEST 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