Font Size: a A A

Research On SOPC Design Technology Of Image Matching Algorithm

Posted on:2018-07-04Degree:MasterType:Thesis
Country:ChinaCandidate:T G SuFull Text:PDF
GTID:2348330542453186Subject:Electronic and communication engineering
Abstract/Summary:PDF Full Text Request
The image matching is a hot research field in digital image processing.With the development of computer hardware,image matching has been applied to many fields,such as object recognition,environment monitoring,3D modeling,image stitching,robot positioning and navigation and so on.SIFT matching algorithm is one of the most effective methods in matching algorithm,which can guarantee the rotation invariance,scale invariance and brightness invariance.But the image data is too large,the use of software for SIFT matching is difficult to meet the real-time requirements.So it is an important research direction to solve the real-time problem of image matching system by using SOPC design of image matching system,which is based on the combination of hardware and softwar,using the parallel processing capability of FPGA to shorten the computing time,and using software to control the system flexibly.In this paper,we focus on the SOPC design of image matching system for real-time and flexibility of image matching.Aiming at the requirement of image matching real-time processing,the improved algorithm is improved by combining wavelet transform and clustering analysis.According to the different sizes of images,the matching system needs to be dealt with flexibly.Combined with the hardware and software co-design technology,the hardware and software framework of SIFT-based image matching system is designed,and the image matching system is built by ZC702 evaluation board.In the improved part of algorithm,we use wavelet transform to extract the low frequency components of the original image,and them carry out rough matching.Then we use the clustering analysis method to estimate the matching position,and then get the final image matching relation by truncating the images near the center of the two matching images.Compard with the direct matching method,the two matching methods of rough matching and fine matching can greatly reduce the matching time and improved the efficiency of the image matching system under the premise of ensuring the accuracy of matching algorithm.In software and hardware co-design,the SOPC design of SIFT image matching system is based on FPGA.On the other hand,combining with software design,software design is needed in the system which needs flexible processing,and the hardware and software are combined to make the matching system have better overall performance and more flexible image processing.The image matching system designed in this paper combines the hardware system and the software system,and uses the ZC702 evaluation board of Xilinx to communicate with the host computer through Ethernet to build the SOPC design of the image matching system.The experimental results show that the image matching system possesses certain real-time property,certain rotation invariance and illumination invariance,and also has some ressitance to noise.
Keywords/Search Tags:Image matching, SIFT, SOPC technology, Wavelet transform, Cluster analysis
PDF Full Text Request
Related items