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

Matching method for large-scale images

A matching method, large-scale technology, applied in the field of image processing, can solve the problem of not fully utilizing the application scene algorithm

Active Publication Date: 2019-04-05
SHANGHAI JIAO TONG UNIV
View PDF5 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the algorithm does not make full use of the characteristics of the application scenario to improve the efficiency of the algorithm

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
  • Matching method for large-scale images
  • Matching method for large-scale images
  • Matching method for large-scale images

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0087] The final result of image matching is generally used for 3D reconstruction of the target, and the quality of reconstruction can indirectly reflect the quality of image matching. In the experiment, COLMAP is used to process the matching results to obtain the final reconstruction results.

[0088] The following introduces the equipment used for the entire experiment, CPU: 2×Intel(R) Xeon(R) E5-2630v4@2.20GHz, graphics card: Tesla P100 16G, memory: 128G, hard disk: SSD 480G, mechanical 4T / 7200 rpm. The data set used in the experiment comes from Flicker. Each image set is actually retrieved from the Flicker website through keywords. The number of image sets ranges from 1497 to 15685, and the ratio of matching image pairs to the total number of image pairs ranges from 0.04 to 0.004. It covers a large range and can effectively test the performance of the algorithm. The experiment compares ThinnerMatch with the brute force search method and GraphMatch. As shown in Table 1, t...

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 matching method for large-scale images. The matching method comprises the step of 1, obtaining a pedestrian area in the image; Step 2, carrying out scale invariant feature transformation feature extraction on the non-pedestrian region; step 3, acquiring Fisher Vector of each image, and obtaining a distance matrix of the image set; step 4, sampling a preset number of imagepairs each time according to the distance matrix; step 5, performing pre-matching on the image pairs, performing accurate matching on the image pairs which are successfully pre-matched, and storing the image pairs in a matching graph; step 6, image pairs with the matching rate lower than a preset threshold value are filtered out; step 7, propagating according to the matching relation to obtain anexpanded image pair; And step 8, performing large-scale feature pre-matching on the expanded image pairs, performing accurate matching on the image pairs which are successfully pre-matched, and storing the image pairs in a matching graph. Repeating the steps 4-8 until all related images have sufficient matching relationships or the image pairs obtained through sampling propagation are unmatched image pairs.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular, to a large-scale image matching method. Background technique [0002] From a large-scale network image collection, the 3D structure can be reconstructed by the method of structure from motion (SFM). At present, common methods mainly include: global reconstruction method, staged reconstruction method and incremental reconstruction method. [0003] Incremental reconstruction has become very popular because of its excellent ability to remove outliers. Its main process is as follows: first, feature extraction is performed on each image, and then a matching map is obtained by establishing a matching relationship between image pairs. The nodes of the graph correspond to each picture, and the edges of the graph correspond to the matching relationship between images. Finally, according to this graph, the three-dimensional structure of the image set is restored by geometri...

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): G06K9/62
CPCG06F18/22
Inventor 李国武周越
Owner SHANGHAI JIAO TONG UNIV