SIFT feature matching method based on OpenCL parallel acceleration

A feature matching and feature point technology, applied in the field of remote sensing mechanism, can solve the problems of long image matching calculation time, incompatible hardware platform, slow matching speed, etc., to achieve platform independence, obvious acceleration effect, and high total acceleration ratio. Effect

Inactive Publication Date: 2015-06-24
ZHENGZHOU NORMAL UNIV
View PDF2 Cites 16 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0002] In the past few decades, researchers at home and abroad have mainly focused on purchasing servers, workstations and other equipment with high-speed single-core CPUs as the computing core, and the application of multi-core CPU computing platforms, using single instruction multiple data flow technology. However, due to the constraints of the manufacturing process technology and the limitations of the design goals of the CPU architecture itself, the performance improvement soon encountered technical barriers. At the same time, the SIFT feature matching algorithm has high complexity. As a result, the image matching calculation time is long and the matching speed is slow. Therefore, on the one hand, traditional computing resources and computing acceleration methods have been difficult to meet the real-time processing requirements of remote sensing image SIFT feature matching. In applications with high real-time requirements, applications are subject to On the other hand, due to the incompatibility between hardware platforms, the problem of poor cross-platform portability of software is prominent

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
  • SIFT feature matching method based on OpenCL parallel acceleration
  • SIFT feature matching method based on OpenCL parallel acceleration
  • SIFT feature matching method based on OpenCL parallel acceleration

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0032] Such as figure 1 , figure 2 , image 3 , Figure 4 and Figure 5 As shown, a SIFT feature matching method based on OpenCL parallel acceleration, which includes SIFT feature matching parallel computing, OpenCL-based checkerboard convolution operation and task assignment and mapping of working nodes, the SIFT feature matching parallel computing includes the following step:

[0033] Step 1: First read the input image to the main memory, transfer the Gaussian kernel data of different scales from the main memory to the constant memory of the GPU, perform step-by-step continuous filtering in the GPU to accelerate the construction of the Gaussian scale space pyramid, and transfer the Gaussian kernel data to the constant memory of the GPU. The pyramid is stored in the global memory, and then the Gaussian pyramid image is read back to the CPU. Two Gaussian images of adjacent scales are subtracted to obtain the multi-scale space representation of the DOG pyramid. After uploa...

Embodiment 2

[0053] Such as figure 1 , figure 2 , image 3 , Figure 4 and Figure 5 As shown, a SIFT feature matching method based on OpenCL parallel acceleration, which includes SIFT feature matching parallel computing, OpenCL-based checkerboard convolution operation and task assignment and mapping of working nodes, the SIFT feature matching parallel computing includes the following step:

[0054] Step 1: First read the input image to the main memory, transfer the Gaussian kernel data of different scales from the main memory to the constant memory of the GPU, perform step-by-step continuous filtering in the GPU to accelerate the construction of the Gaussian scale space pyramid, and transfer the Gaussian kernel data to the constant memory of the GPU. The pyramid is stored in the global memory, and then the Gaussian pyramid image is read back to the CPU. Two Gaussian images of adjacent scales are subtracted to obtain the multi-scale space representation of the DOG pyramid. After uploa...

Embodiment 3

[0076] Such as figure 1 , figure 2 , image 3 , Figure 4 and Figure 5 As shown, a SIFT feature matching method based on OpenCL parallel acceleration, which includes SIFT feature matching parallel computing, OpenCL-based checkerboard convolution operation and task assignment and mapping of working nodes, with preprocessed images of different image sizes The image pair filter window is 21×21 to carry out SIFT feature matching comparison experiment, run SIFT feature matching CPU system, single GPU system and multi-GPUs system respectively, and record the processing time, see Table 1, with 489×561 image frame image pair The comparison experiment of SIFT feature matching with different filter window sizes is carried out, and the CPU system, single GPU system and multi-GPUs system of SIFT feature matching are run respectively, and the processing time is recorded, as shown in Table 2.

[0077] Table 1 Comparison of serial and parallel performance of SIFT feature matching in di...

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 relates to an SIFT feature matching method based on OpenCL parallel acceleration. The SIFT feature matching method includes the steps of SIFT feature matching parallel calculation, chessboard type convolution operation based on OpenCL and task distribution and mapping of working nodes. Plenty of calculation is partitioned to be conducted separately between a CPU and a GPU, the calculation advantages of the CPU and the GPU are played, the strong ability of CPU plus GPU heterogeneous calculation is fully reflected, real-time performance is high, and a Sift feature matching parallel algorithm with transportability achieves platform independence.

Description

technical field [0001] The invention belongs to the technical field of remote sensing mechanism and method, in particular to a SIFT feature matching method based on OpenCL parallel acceleration. Background technique [0002] In the past few decades, researchers at home and abroad have mainly focused on purchasing servers, workstations and other equipment with high-speed single-core CPUs as the computing core, and the application of multi-core CPU computing platforms, using single instruction multiple data flow technology. However, due to the constraints of manufacturing process technology and the limitations of the design goals of the CPU architecture itself, the improvement of performance soon encountered technical barriers. At the same time, the complexity of the SIFT feature matching algorithm is high. As a result, the image matching calculation time is long and the matching speed is slow. Therefore, on the one hand, traditional computing resources and computing accelerat...

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): G06K9/00G06K9/46
Inventor 肖汉肖诗洋周霄云肖波冯娜
Owner ZHENGZHOU NORMAL 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