CN102496033A - Image SIFT feature matching method based on MR computation framework - Google Patents

Image SIFT feature matching method based on MR computation framework Download PDF

Info

Publication number
CN102496033A
CN102496033A CN2011103972105A CN201110397210A CN102496033A CN 102496033 A CN102496033 A CN 102496033A CN 2011103972105 A CN2011103972105 A CN 2011103972105A CN 201110397210 A CN201110397210 A CN 201110397210A CN 102496033 A CN102496033 A CN 102496033A
Authority
CN
China
Prior art keywords
image
key
point
node
value
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN2011103972105A
Other languages
Chinese (zh)
Other versions
CN102496033B (en
Inventor
崔江涛
张佳琦
李林
蒋莲
王博
张国良
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Qingdao Institute Of Computing Technology Xi'an University Of Electronic Science And Technology
Original Assignee
Xidian University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xidian University filed Critical Xidian University
Priority to CN201110397210.5A priority Critical patent/CN102496033B/en
Publication of CN102496033A publication Critical patent/CN102496033A/en
Application granted granted Critical
Publication of CN102496033B publication Critical patent/CN102496033B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Image Analysis (AREA)

Abstract

The invention discloses an image scale-invariant feature transform (SIFT) feature matching method based on a mapreduce (MR) computation framework. The method comprises the following steps that: 1, a cluster is established; 2, a Hadoop platform is constructed; 3, an image library is established; 4, feature points are extracted; 5, key-key-value pairs of the feature points are established; 6, the feature points are uploaded to a distributed file system; 7, an image matching request is made; and 8, image matching is carried out. According to the invention, an MR computation framework is employed; and parallel calculation is realized during the image matching process; therefore, image matching efficiency is improved and time complexity for calculation on image matching can be reduced. Besides, an MR open source is employed to enable a cluster image processing mode of a framework Hadoop platform to be realized; and thus mass image data can be effectively dealt with.

Description

Calculate the image SIFT feature matching method of framework based on MR
Affiliated technical field
The invention belongs to the computer image processing technology field, further relate to a kind of graphical rule invariant features conversion (Scale-invariant feature transform) SIFT feature matching method that calculates framework based on MR (mapreduce) in the technical field of image matching.The present invention can make things convenient for, the identical and similar image in the efficient matching image storehouse, thereby effectively solves the excessive redundancy issue of data in the mass data.
Background technology
The develop rapidly of multimedia application technology, image processing techniques have obtained using widely, and wherein graphical rule invariant features conversion SIFT feature matching method is widely used as a kind of effective image matching method.
The patent of invention " based on the adaptive topography SIFT feature matching method of data clusters " of BJ University of Aeronautics & Astronautics's application (number of patent application: 201110185894.2, publication number: CN 102194133A).This patented claim at first proposes a kind of graphical rule invariant features conversion SIFT feature matching method; Through two phase characteristic clusters; Phase one is carried out cluster based on the k-d data structure to overall repeated characteristic collection; Subordinate phase is utilized vocabulary data structure organization cluster feature collection and the cluster feature that the whole nodes of vocabulary data structure comprise is purified, and guarantees that cluster feature is contained in the voronoi cell that node representes.This patented claim is through two stage graphical rule invariant features conversion SIFT characteristic matching, and the phase one is utilized BBF that cascade vocabulary tree carries out purifying based on ratio and double mode cluster feature match selection key images based on information entropy.The key images that subordinate phase adopts the phase one to select carries out characteristic matching between image, thereby improves the robustness of characteristic matching, strengthens the adaptability of characteristic matching.The main deficiency that this patented claim exists is: in actual use, what store in the image library is mass image data, and the technical scheme that adopts above-mentioned patented claim to provide still can not successfully manage magnanimity.In addition, the processing mode of serial is adopted in above-mentioned patented claim, and higher time complexity is still arranged when mating.
Another kind of method image SIFT feature matching method " David G.Lowe.Distinctive Image Feature from Scale-Invariant Keypoints [C] the .Interational Journal of Computer Vision; 60,2 (2004): 91-110 that extensively adopts in the yardstick invariant features conversion SIFT characteristic matching process." this method proposes Euclidean distance method between a kind of calculatings two width of cloth matching image unique points in graphical rule invariant features conversion SIFT Feature Points Matching process; confirm through calculating in the image to be matched in each unique point and image library the ratio of the Euclidean distance of two key points of the arest neighbors of image whether this unique point obtains to mate, and finds out the number of two width of cloth Image Feature Point Matching simultaneously and confirms whether two width of cloth images mate.The deficiency that causes this method to exist thus is: in matching process, the image in image to be matched and the image library calculates by the width of cloth, higher time complexity is still arranged, and can not successfully manage mass data.
A kind of calculating framework MR " the Dean J; Ghemawat S.MapReduce:simplified data processing on large clusters [J] .Communications of the ACM; 2005,51 (1): 107-113 that extensively adopts in the process of mass data processing." this calculating framework is to be invented by Google company; emerging in recent years Parallel Programming Models; it is placed on parallelization, fault-tolerant, DATA DISTRIBUTION, load balancing etc. in the storehouse; system all is summed up as two step: Map (mapping) stages and Reduce (abbreviation) stage to all operations of data, the developer of those few of parallel computation experiences also can be developed parallelly should be had, to carry out the parallel processing to mass data.
Summary of the invention
The objective of the invention is to overcome the deficiency of prior art, propose a kind of graphical rule invariant features conversion SIFT feature matching method that calculates framework based on MR.In matching process, the image in image to be matched and the image library calculates the coupling that walks abreast under the framework at MR and calculates.
Be to realize above-mentioned purpose, may further comprise the steps of the inventive method:
(1) all nodes in the cluster to be matched is connected in the same LAN cluster that foundation can intercom mutually;
(2) on each node through build the MR implementation framework Hadoop platform of increasing income by the method on bottom to upper strata;
(3) whole matching images are uploaded on the node of cluster to be matched, set up the image library on the node;
(4) utilize graphical rule invariant features conversion SIFT method for distilling, extract minutiae the image library from node;
(5) it is right to set up the key-key-value of each unique point
The unique point of 5a) in node, same width of cloth image being extracted adds identical figure number as first right key of key-key-value, is designated the unique point of same width of cloth image;
The unique point of 5b) same width of cloth image being extracted adds different periods as second right key of key-key-value, is designated the different characteristic point of same width of cloth image;
Generate the high dimensional feature vector during 5c) with extract minutiae as the right value of key-key-value;
(6) key-key-value of whole unique points is increased income in the distributed file system in the implementation framework Hadoop platform to being uploaded to MR;
(7) images match request utilizes graphical rule invariant features conversion SIFT method for distilling to extract the unique point of image to be matched in client;
(8) images match
8a) the Euclidean distance of the characteristic point value of the value of whole unique points of calculating image to be matched and all images in the image library;
8b) the Euclidean distance value with same reference numbers outputs on the same node, with Euclidean distance ascending sort as a result, again ranking results is pressed the period ascending sort, exports two nearest values of identical period Euclidean distance;
8c) in each node, calculation procedure 8b) ratio of two values of all identical periods of output in;
8d) judge ratio whether in threshold value, if ratio in thresholding, then shows the unique point of image to be matched and the Feature Points Matching of the matching image in the image library, the matched feature points of the matching image in the output image storehouse is carried out next step; Otherwise show the coupling failure, return step 8c);
8e) in each node with step 8d) output unique point measure the coupling number according to different figure numbers respectively through the method for sequential search, the result is exported;
Whether the ratio of number of 8f) judging each coupling number and image characteristic point to be matched is in threshold value; If in threshold value; Then show the image that exists in the image library with images match to be matched, otherwise show the image that does not exist in the image library with images match to be matched.
The present invention compared with prior art has the following advantages:
The first, aspect efficient,, adopt the method for serial computing to compare with prior art because the present invention adopts the parallel computation mode in the images match process, can improve the matching efficiency of image, therefore reduced the time complexity that calculates matching image.
The second, aspect image-capable, the present invention adopts MR to increase income the cluster diagram of implementation framework Hadoop platform as processing mode, only adopts the method for single cpu mode to compare with prior art, can successfully manage the view data of magnanimity.
Description of drawings
Fig. 1 is a process flow diagram of the present invention;
Fig. 2 is simulated effect figure of the present invention.
Embodiment
With reference to Fig. 1, performing step of the present invention is done concrete description.
Embodiments of the invention adopt Java language, can under the environment of supporting Java language, realize.Before the described method of embodiment of the present invention, utilize the MR implementation framework Hadoop platform of increasing income, build experimental situation.
Step 1 is connected to all nodes in the cluster to be matched in the same LAN cluster that foundation can intercom mutually.
Step 2; On each node through build the MR implementation framework Hadoop platform of increasing income by the method on bottom to upper strata: linux ubuntu10 operating system is installed on each node; After accomplishing linux ubuntu10 operating system installation; Hadoop0.20.2 and JDK1.6 plug-in unit are installed on system, are accomplished building of Hadoop platform through the file in configuration linux ubuntu10 system file and the Hadoop0.20.2 plug-in unit.
Step 3 is uploaded to whole matching images on the node of cluster to be matched, sets up the image library on the node.
Step 4 is utilized graphical rule invariant features conversion SIFT method for distilling, extract minutiae the image library from node: at first, set up metric space: set up metric space according to the Gaussian convolution formula.Comprising that foundation and the DoG of gaussian pyramid is pyramidal sets up two processes; Secondly, the metric space extreme value detects: in the metric space of having set up, each sampled point compares with its all consecutive point, obtains the maximum or minimum extreme point in the space; Once more, accurately locate extreme point: for the extreme point that has found, through three-dimensional quadratic fit method with remove the mobile rim response point and obtain accurate extreme point; Then, be the unique point assign direction: utilize the gradient direction distribution character of key point neighborhood territory pixel to be each key point assigned direction parameter, make operator possess rotational invariance; At last, unique point generates: according to the direction and the size of key point, and the generating feature point.
Step 5, key-key-value of setting up each unique point is right: at first, the unique point of in node, same width of cloth image being extracted adds identical figure number as first right key of key-key-value, is designated the unique point of same width of cloth image; Secondly, the unique point that same width of cloth image is extracted adds different periods as second right key of key-key-value, is designated the different characteristic point of same width of cloth image; Generate the high dimensional feature vector during once more, with extract minutiae as the right value of key-key-value.
Step 6 is increased income the key-key-value of whole unique points in the distributed file system in the implementation framework Hadoop platform to being uploaded to MR.
Step 7, the images match request utilizes graphical rule invariant features conversion SIFT method for distilling to extract the unique point of image to be matched in client; At first, set up metric space: set up metric space according to the Gaussian convolution formula.Comprising that foundation and the DoG of gaussian pyramid is pyramidal sets up two processes; Secondly, the metric space extreme value detects: in the metric space of having set up, each sampled point compares with its all consecutive point, obtains the maximum or minimum extreme point in the space; Once more, accurately locate extreme point: for the extreme point that has found, through three-dimensional quadratic fit method with remove the mobile rim response point and obtain accurate extreme point; Then, be the unique point assign direction: utilize the gradient direction distribution character of key point neighborhood territory pixel to be each key point assigned direction parameter, make operator possess rotational invariance; At last, unique point generates: according to the direction and the size of key point, and the generating feature point.
Step 8, images match: at first, calculate the Euclidean distance of characteristic point value of value and all images in the image library of whole unique points of image to be matched, the Euclidean distance computing formula is:
L = Σ J = 1 d ( q J - p J ) 2
Wherein, L representes Euclidean distance,
Figure BSA00000627819300052
The secondary square root is calculated in expression,
Figure BSA00000627819300053
Expression putting in marks entirely from j=1 to j=d, d representes the dimension of high dimension vector, q j, p jJ component of the proper vector of expression key point; Secondly; The Euclidean distance value of same reference numbers is outputed on the same node; With Euclidean distance ascending sort as a result, again ranking results is pressed the period ascending sort, export two nearest values of identical period Euclidean distance; In each node, calculate the ratio of two values of all identical periods of exporting in the last step; Once more, judge ratio whether in threshold value, described threshold value is 0.36; If ratio is in thresholding; Then show the unique point of image to be matched and the Feature Points Matching of the matching image in the image library, the matched feature points of the matching image in the output image storehouse is carried out next step; Otherwise show the coupling failure, return a step; Then, in each node, will go up the unique point of step output and measure the coupling number respectively through the method for sequential search, the result will be exported according to different figure numbers; At last; Whether the ratio of number of judging each coupling number and image characteristic point to be matched is in threshold value; Described ratio threshold value is 0.36; If the coupling number in threshold value, then shows the image that exists in the image library with images match to be matched, otherwise shows the image that does not exist in the image library with images match to be matched.
Below in conjunction with Fig. 2 effect of the present invention is done further description.
Fig. 2 representes to adopt image SIFT feature matching method and the efficient contrast synoptic diagram that adopts the present invention in the images match process.Fig. 2 is adding up the present invention respectively and SIFT feature matching method matching image in image library is respectively 250 width of cloth; 500 width of cloth; 1000 width of cloth are completed on the basis of required separately match time under four kinds of situation of 2000 width of cloth, wherein; Solid line representes to use efficiency curve of the present invention, and the actual situation line representes to use the efficiency curve of image SIFT feature matching method.The present invention's matching image in image library is increased in the process of 2000 width of cloth by 250 width of cloth; Required match time, increasing degree was less; And image SIFT feature matching method image in image library is increased in 2000 width of cloth processes by 250 width of cloth, and required match time, increasing degree was obvious.But there is tangible efficient to improve at the reply mass image data than image SIFT feature matching method by the invention of figure knowledge capital, can be good at tackling mass data.

Claims (6)

1. calculate the image SIFT feature matching method of framework based on MR, may further comprise the steps:
(1) all nodes in the cluster to be matched is connected in the same LAN cluster that foundation can intercom mutually;
(2) on each node through build the MR implementation framework Hadoop platform of increasing income by the method on bottom to upper strata;
(3) whole matching images are uploaded on the node of cluster to be matched, set up the image library on the node;
(4) utilize graphical rule invariant features conversion SIFT method for distilling, extract minutiae the image library from node;
(5) it is right to set up the key-key-value of each unique point
The unique point of 5a) in node, same width of cloth image being extracted adds identical figure number as first right key of key-key-value, is designated the unique point of same width of cloth image;
The unique point of 5b) same width of cloth image being extracted adds different periods as second right key of key-key-value, is designated the different characteristic point of same width of cloth image;
Generate the high dimensional feature vector during 5c) with extract minutiae as the right value of key-key-value;
(6) key-key-value of whole unique points is increased income in the distributed file system in the implementation framework Hadoop platform to being uploaded to MR;
(7) images match request utilizes graphical rule invariant features conversion SIFT method for distilling to extract the unique point of image to be matched in client;
(8) images match
8a) the Euclidean distance of the characteristic point value of the value of whole unique points of calculating image to be matched and all images in the image library;
8b) the Euclidean distance value with same reference numbers outputs on the same node, with Euclidean distance ascending sort as a result, again ranking results is pressed the period ascending sort, exports two nearest values of identical period Euclidean distance;
8c) in each node, calculation procedure 8b) ratio of two values of all identical periods of output in;
8d) judge ratio whether in threshold value, if ratio in thresholding, then shows the unique point of image to be matched and the Feature Points Matching of the matching image in the image library, the matched feature points of the matching image in the output image storehouse is carried out next step; Otherwise show the coupling failure, return step 8c);
8e) in each node with step 8d) output unique point measure the coupling number according to different figure numbers respectively through the method for sequential search, the result is exported;
Whether the ratio of number of 8f) judging each coupling number and image characteristic point to be matched is in threshold value; If in threshold value; Then show the image that exists in the image library with images match to be matched, otherwise show the image that does not exist in the image library with images match to be matched.
2. the image SIFT feature matching method based on MR calculating framework according to claim 1, it is characterized in that: the described building method concrete steps by bottom to upper strata of step (2) are following:
The first step is for installing linux ubuntu10 operating system on each node;
Second step is for each node is installed the Hadoop0.20.2 plug-in unit;
The 3rd step is for each node is installed the JDK1.6 plug-in unit;
The 4th step, configuration linux ubuntu10 system file;
The 5th step, the file in the configuration Hadoop0.20.2 plug-in unit.
3. the image SIFT feature matching method that calculates framework based on MR according to claim 1 is characterized in that: the concrete steps of the described graphical rule invariant features conversion of step (4) and step (7) SIFT method for distilling are following:
The first step is set up metric space: set up metric space according to the Gaussian convolution formula, comprise that foundation and the DoG of gaussian pyramid is pyramidal to set up two processes;
In second step, the metric space extreme value detects: in the metric space of having set up, each sampled point compares with its all consecutive point, obtains the maximum or minimum extreme point in the space;
In the 3rd step, accurately locate extreme point: for the extreme point that has found, through three-dimensional quadratic fit method with remove the mobile rim response point and obtain accurate extreme point;
In the 4th step, be the unique point assign direction: the gradient direction distribution character that utilizes the key point neighborhood territory pixel makes operator possess rotational invariance for each key point assigned direction parameter;
In the 5th step, unique point generates: according to the direction and the size of key point, and the generating feature point.
4. the image SIFT feature matching method based on MR calculating framework according to claim 1, it is characterized in that: step 8a) said Euclidean distance computing formula is:
L = Σ J = 1 d ( q J - p J ) 2
Wherein, L representes Euclidean distance,
Figure FSA00000627819200022
The secondary square root is calculated in expression,
Figure FSA00000627819200023
Expression putting in marks entirely from j=1 to j=d, d representes the dimension of high dimension vector, q j, p jJ component of the proper vector of expression key point.
5. the image SIFT feature matching method based on MR calculating framework according to claim 1, it is characterized in that: step 8d) described ratio threshold value is 0.36.
6. the image SIFT feature matching method based on MR calculating framework according to claim 1, it is characterized in that: step 8f) described ratio threshold value is 0.6.
CN201110397210.5A 2011-12-05 2011-12-05 Image SIFT feature matching method based on MR computation framework Active CN102496033B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201110397210.5A CN102496033B (en) 2011-12-05 2011-12-05 Image SIFT feature matching method based on MR computation framework

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201110397210.5A CN102496033B (en) 2011-12-05 2011-12-05 Image SIFT feature matching method based on MR computation framework

Publications (2)

Publication Number Publication Date
CN102496033A true CN102496033A (en) 2012-06-13
CN102496033B CN102496033B (en) 2014-06-18

Family

ID=46187858

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201110397210.5A Active CN102496033B (en) 2011-12-05 2011-12-05 Image SIFT feature matching method based on MR computation framework

Country Status (1)

Country Link
CN (1) CN102496033B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103366173A (en) * 2013-07-10 2013-10-23 华中科技大学 Distributed parallel extraction method for local feature
CN104112136A (en) * 2013-04-19 2014-10-22 昆山鸿鹄信息技术服务有限公司 Image low-level visual feature extraction method
CN109801319A (en) * 2019-01-03 2019-05-24 杭州电子科技大学 Method for registering is grouped based on the Hadoop classification figure accelerated parallel
CN110633733A (en) * 2019-08-14 2019-12-31 中国平安财产保险股份有限公司 Intelligent image matching method and device and computer readable storage medium
CN111080525A (en) * 2019-12-19 2020-04-28 成都海擎科技有限公司 Distributed image and primitive splicing method based on SIFT (Scale invariant feature transform) features
CN111401482A (en) * 2020-04-29 2020-07-10 Oppo广东移动通信有限公司 Feature point matching method and device, equipment and storage medium

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101714254A (en) * 2009-11-16 2010-05-26 哈尔滨工业大学 Registering control point extracting method combining multi-scale SIFT and area invariant moment features

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101714254A (en) * 2009-11-16 2010-05-26 哈尔滨工业大学 Registering control point extracting method combining multi-scale SIFT and area invariant moment features

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
王贤伟: "基于Hadoop的外观专利图像检索系统的研究与实现", 《中国优秀硕士学位论文全文数据库,中国学术期刊(光盘版)电子期刊》 *

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104112136A (en) * 2013-04-19 2014-10-22 昆山鸿鹄信息技术服务有限公司 Image low-level visual feature extraction method
CN103366173A (en) * 2013-07-10 2013-10-23 华中科技大学 Distributed parallel extraction method for local feature
CN103366173B (en) * 2013-07-10 2016-05-25 华中科技大学 A kind of distributed parallel extracting method of local feature
CN109801319A (en) * 2019-01-03 2019-05-24 杭州电子科技大学 Method for registering is grouped based on the Hadoop classification figure accelerated parallel
CN110633733A (en) * 2019-08-14 2019-12-31 中国平安财产保险股份有限公司 Intelligent image matching method and device and computer readable storage medium
CN110633733B (en) * 2019-08-14 2024-05-03 中国平安财产保险股份有限公司 Image intelligent matching method, device and computer readable storage medium
CN111080525A (en) * 2019-12-19 2020-04-28 成都海擎科技有限公司 Distributed image and primitive splicing method based on SIFT (Scale invariant feature transform) features
CN111401482A (en) * 2020-04-29 2020-07-10 Oppo广东移动通信有限公司 Feature point matching method and device, equipment and storage medium
CN111401482B (en) * 2020-04-29 2024-03-19 Oppo广东移动通信有限公司 Feature point matching method and device, equipment and storage medium

Also Published As

Publication number Publication date
CN102496033B (en) 2014-06-18

Similar Documents

Publication Publication Date Title
CN102496033B (en) Image SIFT feature matching method based on MR computation framework
CN110659693B (en) K-nearest neighbor classification-based power distribution network rapid topology identification method, system and medium
CN109033170B (en) Data repairing method, device and equipment for parking lot and storage medium
CN103116766A (en) Increment neural network and sub-graph code based image classification method
CN104750973B (en) Unit load (standard) steady state condition clustering algorithm based on data smoothness function
CN102708327A (en) Network community discovery method based on spectrum optimization
CN104346481A (en) Community detection method based on dynamic synchronous model
US20150169723A1 (en) Three-dimensional image file searching method and three-dimensional image file searching system
KR101182683B1 (en) A Visual Shape Descriptor Generating Method Using Sectors and Shape Context of Contour Lines and the Recording Medium thereof
CN108595631B (en) Three-dimensional CAD model double-layer retrieval method based on graph theory
CN107392239A (en) A kind of K Means algorithm optimization methods based on Spark computation models
Bidoki et al. Comparison of several clustering methods in the case of electrical load curves classification
CN101908215B (en) Spatial data fusing method
CN111626311B (en) Heterogeneous graph data processing method and device
CN109934866A (en) A kind of extracting method, copy method and the device of figure spot partition mode
Zou et al. Parallel in situ detection of connected components in adaptive mesh refinement data
CN115658809A (en) Data distributed clustering method and device based on local direction centrality
CN112258566B (en) Pavement collection point identification method and device and server
CN107391871A (en) A kind of space lattice deformation method based on parallelization RBF
CN110991083B (en) Photovoltaic power station model determining method, device, equipment and storage medium
CN113128006A (en) High-voltage power distribution method and device based on neural network
CN106789285B (en) Online social network multi-scale community discovery method
CN105404892A (en) Ordered fuzzy C mean value cluster method used for sequence data segmentation
CN105260415A (en) Hierarchical multi-granularity community discovering method
CN117056088B (en) Multi-mode mapping data distributed parallel computing method and system based on MapReduce

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
TR01 Transfer of patent right
TR01 Transfer of patent right

Effective date of registration: 20210202

Address after: Building 28 and 29, Tian'an Digital City, No.88 Chunyang Road, Chengyang District, Qingdao City, Shandong Province

Patentee after: Qingdao Institute of computing technology Xi'an University of Electronic Science and technology

Address before: 710071 No. 2 Taibai South Road, Shaanxi, Xi'an

Patentee before: XIDIAN University