CN104376052A - Same-style commodity merging method based on commodity images - Google Patents
Same-style commodity merging method based on commodity images Download PDFInfo
- Publication number
- CN104376052A CN104376052A CN201410607607.6A CN201410607607A CN104376052A CN 104376052 A CN104376052 A CN 104376052A CN 201410607607 A CN201410607607 A CN 201410607607A CN 104376052 A CN104376052 A CN 104376052A
- Authority
- CN
- China
- Prior art keywords
- image
- money
- node
- commodity
- machine node
- 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
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
Abstract
The invention discloses a same-style commodity merging method based on commodity images. The same-style commodity merging method comprises the steps that by means of overall and local visual features, a distributed storage and calculation matching mode is adopted, firstly, the high-efficiency overall features are searched for similar repeated images, and if no similar repeated image is found in the overall features, the local features are searched for similar repeated images; an image library is searched for the most similar repeated image, a commodity is attributed to the same-style set to which the similar repeated image belongs, if no similar repeated image is found, the commodity is attributed into a new same-style set. Due to the overall and local feature index structure, concurrent access, real-time reading and writing, distributed storage and calculation can be supported, and real-time online same-style calculation of massive image data can be achieved rapidly.
Description
Technical field
The present invention relates to computer vision and technical field of information retrieval, particularly relate to a kind of based on same money commodity merging method that is overall and local feature.
Background technology
Approximate multiimage (Image Near-Duplicate) this concept is proposed in " Efficient Near-duplicate Detection and Sub-image Retrieval " by people such as Yan Ke the earliest.Approximate multiimage refers to the image that same object or scene obtain in different shooting situation (illumination, yardstick, angle, to block).Approximate multiimage retrieval, as a branch of image retrieval, has wide application scenarios.Such as, image infringement detection, image shopping search, link related web page, video frequency searching etc.Blocking in approximate multiimage, displacement, yardstick, light change etc. be all that automatic approximation multiimage is retrieved and brought challenge.
David G.Lowe propose in " Distinctive Image Features from Scale-Invariant Keypoints " a kind of based on metric space, to image scaling, rotate image local feature descriptor--the SIFT descriptor that even affined transformation maintains the invariance, its full name is Scale Invariant Feature Transform, i.e. Scale invariant features transform.Practice shows that SIFT descriptor not only has good yardstick and brightness unchangeability, also has certain robustness to affine deformation, visual angle change and noise simultaneously.
The feature that the characteristic that robustness, locality and resolving ability are strong makes local feature become multiple field such as image retrieval, video copy detection to favor.But the characteristic that its dimension is high, quantity is many is require high searching field to bring index to Time & Space Complexity and inquire about the large challenge of burden.For making full use of local feature and reducing calculated amount, researchist uses for reference word bag method (the Bag of Words in text retrieval field, BoW), the frequency histogram namely utilizing keyword to occur represents the method for one section of document, proposes visual dictionary method (Bag of Visual Words).First algorithm extracts local feature from a large amount of training sample image, then these unique points is carried out cluster, obtains vision code book, i.e. a visual dictionary, and each cluster centre in code book represents a vision word; Finally all local feature region is mapped to vision word for often opening image, the coupling of two images changes into the coupling being similar to two text documents, and the index of image also can adopt the inverted index being similar to document.
Summary of the invention
The object of this invention is to provide a kind of same money commodity merging method based on commodity image.
The object of the invention is to be achieved through the following technical solutions: a kind of same money commodity merging method based on commodity image, comprises the steps:
(1) extraction of characteristics of image: extract the overall situation of each image in image library, the two kinds of visual signatures in local, wherein the integral color of global characteristics Description Image distributes and grain details, the gradient information of the crucial regional area of local feature description one group;
(2) global characteristics coupling, find approximate repeated matching image: adopt distributed storage and calculate match pattern, utilizing N number of machine node to manage feature database, be divided into following steps:
(2.1) the retrieval machine node of match query is determined: utilize the global image proper vector that step (1) obtains, carry out two differentiation to global characteristics data and combine obtaining a hash value, by carrying out grid section division to hash value, set up the corresponding relation of hash value with each machine node; Be specially: from global characteristics vector V, select CLD and EHD proper vector, wherein for these two category features data, the magnitude range of every category feature is CLD:0 ~ 63, EHD:0 ~ 7, two differentiation are carried out to each characteristic, exports 0/1 coding, composition hash value of being joined together by multiple coding; To each machine node uniform distribution part hash value, multiple combination of nodes is that overall hash shows together; When the approximate multiimage of needs inquiry, calculate feature by image and generate hash value by binaryzation, then find affiliated machine node by this hash value;
(2.2) global characteristics coupling: the picture in the global data base in the machine node determined in use step 2.1 carries out matching ratio comparatively with the global characteristics of retrieving images, analyzes whether there is same money, and carries out Data Update; Be specially: the global characteristics of this query image mates with the set of image characteristics of all images in the image library in this machine node; If similarity distance is less than 0.01, then think for approximate repeated matching image, namely with money commodity, return this approximate repeated matching image ID, as do not inquired identical money, then this retrieving images ID and characteristics of image are written in the global image feature set of this machine node, enter step 3; When writing proper vector to machine node, using locks carries out write-protect;
(3) approximate repeated matching image is looked in the choosing of local feature coupling: adopt distributed storage and calculate match pattern, using N number of machine node to manage local property data base and inquire about, be divided into following steps:
(3.1) retrieval and inquisition of individual node: for query image, utilize step 1, first local feature is calculated, obtain a series of vision word, different with global registration, the local feature vectors of query image is inquired about being sent to N number of storage computing node simultaneously, the local feature vectors of query image with in each node { image in the feature database in Local Data} mates, if phase knowledge and magnanimity are greater than 0.9, then think that this machine node exists the approximate multiimage of query image, namely with money commodity, return the ID of this approximate multiimage, otherwise nothing is with money commodity in this node diagnostic storehouse,
(3.2) merge the result for retrieval of each node: the Query Result of N number of calculating storage node is merged, if there is same money, then enter step (4), if do not retrieve approximate repeated matching image, enter step 3.3 characteristics of image stored record link,
(3.3) characteristic storage record: first the local feature vectors of query image compares with the characteristic write in buffer memory WriteCache, whether inquiry exists same amount of money certificate, if data cached middle existence is with money, enter step (4), characteristic is not preserved; If data cached middle nothing is with amount of money certificate, enter step (4), the local feature vectors of query image writes in WriteCache simultaneously, and whether the property data base size analyzing WriteCache exceedes the cache size threshold value of setting; If buffer memory is full, then adopt repeating query hit mode to be written in the local feature storehouse of this machine node to machine node the characteristic in WriteCache, the machine node be written into locks when writing and protects;
(4) obtain with money group number: if find approximate multiimage, the same money group number of approximate multiimage step (2) or step (3) matched distributes to new images, if step (2) and step (3) all do not find approximate multiimage, use the same money group number of image ID as input figure of input figure; New outcome record in same money database.
The invention has the beneficial effects as follows: utilize the overall situation and local visual signature, adopt Distributed Calculation to store and calculate match pattern, first in the global characteristics that efficiency is relatively higher, approximate multiimage is searched, if global search failure, then search approximate multiimage in relatively slow local feature; From image library, retrieve a most similar approximate multiimage, commodity are belonged to the same same money group belonging to approximate multiimage, if approximate multiimage, commodity are belonged to a new same money group.The first inquiry mode of the rear local of the overall situation, under the prerequisite not affecting recall ratio, improve the query performance of system, whole system can support Concurrency Access, in real time read-write, distributed storage and calculating, and the real-time online completing mass image data fast calculates with money.
Accompanying drawing explanation
Fig. 1 is that input picture calculates with money overall flow figure.
Embodiment
Describe the present invention in detail below in conjunction with accompanying drawing, object of the present invention and effect will become more obvious.
As shown in Figure 1, a kind of same money commodity merging method based on commodity image comprises the steps:
Step 1: the overall situation of image, local visual feature extraction.
The following several visual signature of concrete employing in the present invention:
Global color feature have employed the color layout descriptors (Color layout Descriptor, CLD) of MPEG-7, have expressed the distributed intelligence of color in space.The present invention adopts partition strategy (be namely divided into full figure block and picture centre block, totally 2 block images extract feature respectively) to this operator, totally 12 dimension (2*6) color layout features.
Overall situation textural characteristics have employed the edge histogram descriptor (Edge Histogram Descriptor, EHD) of MPEG-7, is extracted 80 dimension data described from 16 sub-picture content.
The vision word of local feature based on SIFT feature; First use the SIFT feature cluster of samples pictures to generate visual dictionary that one comprises 1,000,000 rank vision word quantity, each SIFT key point of image is described as a vision word and comprises the auxiliary matched information such as volume coordinate, yardstick, direction.SIFT key point is by conspicuousness cutting, and a reserve part remarkable characteristic, to improve accuracy and the speed of coupling.
Present invention employs above 3 kinds of features and carry out feature extraction to commodity image, global characteristics comprises color layout descriptors CLD and edge histogram EHD, and local feature is the vision word that SIFT feature generates.
Step 2: global characteristics mates, finds approximate reproducible results;
Same money under large data sets calculates, approximate repeated matching image is searched in more than one hundred million width commodity image from commodity storehouse, in order to improve system performance and efficiency, because global characteristics characteristic matching speed is fast, first using global characteristics to search golden repeated matching result, in order to improve the inquiry velocity of system, utilizing distributed storage computation schema, adopt N number of calculating store machine rows of nodes to calculate and management to image overall feature database, concrete meter search procedure is as follows:
The 2.1 retrieval machine nodes determining match query, first the arbitrary commodity image in image library is chosen, extract overall Vision proper vector V, owing to adopting distributed storage and calculating, there is N number of storage and computing node, in order to accelerate matching primitives speed, the mode that we adopt grid section to divide, proper vector is utilized to produce a series of hash value, according to hash value, set up one-to-one relationship between hash value and machine node i (i=0...N), each storage and calculated value only carry out at a machine joint.
N represents N number of distributed machines node of the existence of system;
Specific features vector is set up as follows with machine node relationships:
From global characteristics vector V, select CLD and EHD proper vector, wherein for these two category features data, the magnitude range of every category feature is CLD:0 ~ 63, EHD:0 ~ 7, two differentiation are carried out to each characteristic, exports 0/1 coding, composition hash value of being joined together by multiple coding; To each machine node uniform distribution part hash value, multiple combination of nodes is that overall hash shows together;
When the approximate multiimage of needs inquiry, calculate feature by image and generate hash value by binaryzation, then find affiliated machine node by this hash value.,
2.2 global characteristics mates:
Use the picture in the global data base in the machine node determined in 2.1 to carry out matching ratio comparatively with the global characteristics of query image, analyze whether there is same money, and carry out Data Update, specific as follows:
The global characteristics of this query image mates with the set of image characteristics of all images in the image library in this machine node.If similarity distance is less than 0.01; then think for approximate repeated matching image; for same money commodity, return this approximate repeated matching image ID, as do not inquired identical money; then this retrieving images id and characteristics of image are written to this machine node; the characteristics of image being written to this machine node is concentrated, in order to ensure security and the integrality of data, when writing proper vector to machine node; carry out write-protect, a machine node synchronization does not allow the proper vector writing multiple image.
Step 3 local feature mates
The local feature of image is one group of vision word, and local feature adopts vision word to set up inverted index (setting up inverted index method is general conventional method, can check prior art document);
In order to improve the speed of retrieval, local feature coupling is with global characteristics coupling, and system adopts distributed multimachine device node calculate match pattern, adopts N number of calculating store machine rows of nodes to calculate and management to image local feature storehouse, the characteristic data set of each node is recorded as { Local Data}
In order to ensure the matching analysis completing any two images on single machine, multiple vision word in the local feature of every width image are stored in same machine node data, instead of are evenly distributed in multiple machine node according to single vision word.
In order to improve the concurrency performance of system, one is arranged to characteristics of image storage and writes data cached space W rite Cache, the data writing buffer memory can be written in the local feature storehouse on different machine nodes that { Local Data} is uniformly distributed to make the characteristic of distributed machines in turn.
Concrete matching process is as follows:
The retrieval and inquisition of 3.1 individual nodes:
For query image, first-selected calculating local feature, obtain a series of vision word, different with global registration, the local feature vectors of query image is inquired about being sent to N number of storage computing node simultaneously, the local feature vectors of query image with in each node { image in the feature database in Local Data} mates, if phase knowledge and magnanimity are greater than 0.9, then think that this machine node exists for approximate multiimage, for same money commodity, return the ID of this approximate multiimage, otherwise without with money commodity in this node diagnostic storehouse.
The result for retrieval of 3.2 each nodes of merging:
The Query Result of N number of calculating storage node is merged, if there is same money, then enters step 4, if do not retrieve approximate repeated matching image, enter 3.3 characteristics of image stored record links.
3.3 feature store recordings:
First the local feature vectors of query image compares with the characteristic write in buffer memory WriteCache, whether inquiry exists same amount of money certificate, if data cached middle existence, with money, enters step 4, characteristic is not preserved, if data cached middle nothing is with amount of money certificate, then the local feature vectors of query image writes in WriteCache, and whether the property data base size analyzing WriteCache exceedes the cache size threshold value of setting.If buffer memory is full, then the characteristic in WriteCache is written in one of them machine node, the selection of this machine node adopts the mode of repeating query, to ensure that characteristic is evenly distributed in each machine node.Prevent the loss of data for preventing writing two simultaneously for the data characteristics with money image, the machine node be written into locks when writing and protects, to ensure the recall ratio of system.
Step 4: obtain same money group number.
The image ID of the approximate multiimage obtained according to step 2 and step 3 inquires affiliated same money group number in a database, using this with the same money group number of money group number as input figure;
When step 2 and step 3 all do not find approximate multiimage, use the same money group number of image ID as input figure of input figure; New same money outcome record in database, are same moneys with all images that money group number is identical.
The present invention, by extracting the overall situation and local feature to input picture, adopts Distributed Calculation storage mode, first searches in the global characteristics that efficiency is relatively higher, searches approximate multiimage again during global search failure in local feature; The index structure of the overall situation and local feature can both support Concurrency Access, in real time read-write, distributed storage and calculating, and the real-time online realizing mass image data calculates with money.
Claims (1)
1. one kind based on the same money commodity merging method of commodity image, it is characterized in that, comprises the steps:
(1) extraction of characteristics of image: extract the overall situation of each image in image library, the two kinds of visual signatures in local, wherein the integral color of global characteristics Description Image distributes and grain details, the gradient information of the crucial regional area of local feature description one group;
(2) global characteristics coupling, find approximate repeated matching image: adopt distributed storage and calculate match pattern, utilizing N number of machine node to manage feature database, be divided into following steps:
(2.1) the retrieval machine node of match query is determined: utilize the global image proper vector that step (1) obtains, carry out two differentiation to global characteristics data and combine obtaining a hash value, by carrying out grid section division to hash value, set up the corresponding relation of hash value with each machine node; Be specially: from global characteristics vector V, select CLD and EHD proper vector, wherein for these two category features data, the magnitude range of every category feature is CLD:0 ~ 63, EHD:0 ~ 7, two differentiation are carried out to each characteristic, exports 0/1 coding, composition hash value of being joined together by multiple coding; To each machine node uniform distribution part hash value, multiple combination of nodes is that overall hash shows together; When the approximate multiimage of needs inquiry, calculate feature by image and generate hash value by binaryzation, then find affiliated machine node by this hash value;
(2.2) global characteristics coupling: the picture in the global data base in the machine node determined in use step 2.1 carries out matching ratio comparatively with the global characteristics of retrieving images, analyzes whether there is same money, and carries out Data Update; Be specially: the global characteristics of this query image mates with the set of image characteristics of all images in the image library in this machine node; If similarity distance is less than 0.01, then think for approximate repeated matching image, namely with money commodity, return this approximate repeated matching image ID, as do not inquired identical money, then this retrieving images ID and characteristics of image are written in the global image feature set of this machine node, enter step 3; When writing proper vector to machine node, using locks carries out write-protect;
(3) approximate repeated matching image is looked in the choosing of local feature coupling: adopt distributed storage and calculate match pattern, using N number of machine node to manage local property data base and inquire about, be divided into following steps:
(3.1) retrieval and inquisition of individual node: for query image, utilize step 1, first local feature is calculated, obtain a series of vision word, different with global registration, the local feature vectors of query image is inquired about being sent to N number of storage computing node simultaneously, the local feature vectors of query image with in each node { image in the feature database in Local Data} mates, if phase knowledge and magnanimity are greater than 0.9, then think that this machine node exists the approximate multiimage of query image, namely with money commodity, return the ID of this approximate multiimage, otherwise nothing is with money commodity in this node diagnostic storehouse,
(3.2) merge the result for retrieval of each node: the Query Result of N number of calculating storage node is merged, if there is same money, then enter step (4), if do not retrieve approximate repeated matching image, enter step 3.3 characteristics of image stored record link,
(3.3) characteristic storage record: first the local feature vectors of query image compares with the characteristic write in buffer memory WriteCache, whether inquiry exists same amount of money certificate, if data cached middle existence is with money, enter step (4), characteristic is not preserved; If data cached middle nothing is with amount of money certificate, enter step (4), the local feature vectors of query image writes in WriteCache simultaneously, and whether the property data base size analyzing WriteCache exceedes the cache size threshold value of setting; If buffer memory is full, then adopt repeating query hit mode to be written in the local feature storehouse of this machine node to machine node the characteristic in WriteCache, the machine node be written into locks when writing and protects;
(4) obtain with money group number: if find approximate multiimage, the same money group number of approximate multiimage step (2) or step (3) matched distributes to new images, if step (2) and step (3) all do not find approximate multiimage, use the same money group number of image ID as input figure of input figure; New outcome record in same money database.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410607607.6A CN104376052B (en) | 2014-11-03 | 2014-11-03 | A kind of same money commodity merging method based on commodity image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410607607.6A CN104376052B (en) | 2014-11-03 | 2014-11-03 | A kind of same money commodity merging method based on commodity image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104376052A true CN104376052A (en) | 2015-02-25 |
CN104376052B CN104376052B (en) | 2017-07-14 |
Family
ID=52554959
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410607607.6A Expired - Fee Related CN104376052B (en) | 2014-11-03 | 2014-11-03 | A kind of same money commodity merging method based on commodity image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104376052B (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106777167A (en) * | 2016-12-21 | 2017-05-31 | 中国科学院上海高等研究院 | Magnanimity Face Image Retrieval System and search method based on Spark frameworks |
CN106844381A (en) * | 2015-12-04 | 2017-06-13 | 富士通株式会社 | Image processing apparatus and method |
CN106951551A (en) * | 2017-03-28 | 2017-07-14 | 西安理工大学 | The cumulative index image search method of joint GIST features |
WO2017143575A1 (en) * | 2016-02-25 | 2017-08-31 | 华为技术有限公司 | Method for retrieving content of image, portable electronic device, and graphical user interface |
CN107273392A (en) * | 2016-04-06 | 2017-10-20 | 百度(美国)有限责任公司 | Computer implemented method, device and data handling system for searching for image |
CN110119460A (en) * | 2019-05-16 | 2019-08-13 | 广东三维家信息科技有限公司 | Image search method, device and electronic equipment |
CN110622156A (en) * | 2017-05-12 | 2019-12-27 | 华为技术有限公司 | Incremental graph computation for querying large graphs |
CN110675207A (en) * | 2018-07-03 | 2020-01-10 | 阿里巴巴集团控股有限公司 | Image display combination recommendation method, device and equipment |
CN111008210A (en) * | 2019-11-18 | 2020-04-14 | 浙江大华技术股份有限公司 | Commodity identification method and device, codec and storage device |
CN111552829A (en) * | 2020-05-07 | 2020-08-18 | 北京海益同展信息科技有限公司 | Method and apparatus for analyzing image material |
CN112182264A (en) * | 2020-10-10 | 2021-01-05 | 书行科技(北京)有限公司 | Method, device and equipment for determining landmark information and readable storage medium |
CN113128923A (en) * | 2020-01-15 | 2021-07-16 | 北京京东乾石科技有限公司 | Storage position recommendation method and device |
CN113763211A (en) * | 2021-09-23 | 2021-12-07 | 支付宝(杭州)信息技术有限公司 | Infringement detection method and device based on block chain and electronic equipment |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102339438A (en) * | 2010-07-22 | 2012-02-01 | 阿里巴巴集团控股有限公司 | Commodity information website publishing method, system and device |
CN102567543A (en) * | 2012-01-12 | 2012-07-11 | 北京搜狗信息服务有限公司 | Clothing picture search method and clothing picture search device |
CN102760144A (en) * | 2011-04-26 | 2012-10-31 | 乐活在线(北京)网络技术有限公司 | Information search method and system |
CN102890686A (en) * | 2011-07-21 | 2013-01-23 | 腾讯科技(深圳)有限公司 | Method and system for showing commodity search result |
CN103092861A (en) * | 2011-11-02 | 2013-05-08 | 阿里巴巴集团控股有限公司 | Method and system for selecting commodity representative picture |
US20140032359A1 (en) * | 2012-07-30 | 2014-01-30 | Infosys Limited | System and method for providing intelligent recommendations |
US20140052580A1 (en) * | 2012-08-17 | 2014-02-20 | Kallidus, Inc. | Product explorer page for use with interactive digital catalogs and touch-screen devices |
-
2014
- 2014-11-03 CN CN201410607607.6A patent/CN104376052B/en not_active Expired - Fee Related
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102339438A (en) * | 2010-07-22 | 2012-02-01 | 阿里巴巴集团控股有限公司 | Commodity information website publishing method, system and device |
CN102760144A (en) * | 2011-04-26 | 2012-10-31 | 乐活在线(北京)网络技术有限公司 | Information search method and system |
CN102890686A (en) * | 2011-07-21 | 2013-01-23 | 腾讯科技(深圳)有限公司 | Method and system for showing commodity search result |
CN103092861A (en) * | 2011-11-02 | 2013-05-08 | 阿里巴巴集团控股有限公司 | Method and system for selecting commodity representative picture |
CN102567543A (en) * | 2012-01-12 | 2012-07-11 | 北京搜狗信息服务有限公司 | Clothing picture search method and clothing picture search device |
US20140032359A1 (en) * | 2012-07-30 | 2014-01-30 | Infosys Limited | System and method for providing intelligent recommendations |
US20140052580A1 (en) * | 2012-08-17 | 2014-02-20 | Kallidus, Inc. | Product explorer page for use with interactive digital catalogs and touch-screen devices |
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106844381A (en) * | 2015-12-04 | 2017-06-13 | 富士通株式会社 | Image processing apparatus and method |
WO2017143575A1 (en) * | 2016-02-25 | 2017-08-31 | 华为技术有限公司 | Method for retrieving content of image, portable electronic device, and graphical user interface |
CN107533566A (en) * | 2016-02-25 | 2018-01-02 | 华为技术有限公司 | Method, portable electric appts and the graphic user interface retrieved to the content of picture |
CN107273392A (en) * | 2016-04-06 | 2017-10-20 | 百度(美国)有限责任公司 | Computer implemented method, device and data handling system for searching for image |
CN106777167B (en) * | 2016-12-21 | 2020-05-12 | 中国科学院上海高等研究院 | Massive human face image retrieval system and retrieval method based on Spark framework |
CN106777167A (en) * | 2016-12-21 | 2017-05-31 | 中国科学院上海高等研究院 | Magnanimity Face Image Retrieval System and search method based on Spark frameworks |
CN106951551A (en) * | 2017-03-28 | 2017-07-14 | 西安理工大学 | The cumulative index image search method of joint GIST features |
CN110622156B (en) * | 2017-05-12 | 2022-05-31 | 华为技术有限公司 | Incremental graph calculation method and device for querying large graph |
CN110622156A (en) * | 2017-05-12 | 2019-12-27 | 华为技术有限公司 | Incremental graph computation for querying large graphs |
CN110675207A (en) * | 2018-07-03 | 2020-01-10 | 阿里巴巴集团控股有限公司 | Image display combination recommendation method, device and equipment |
CN110119460A (en) * | 2019-05-16 | 2019-08-13 | 广东三维家信息科技有限公司 | Image search method, device and electronic equipment |
CN111008210A (en) * | 2019-11-18 | 2020-04-14 | 浙江大华技术股份有限公司 | Commodity identification method and device, codec and storage device |
CN111008210B (en) * | 2019-11-18 | 2023-08-11 | 浙江大华技术股份有限公司 | Commodity identification method, commodity identification device, codec and storage device |
CN113128923A (en) * | 2020-01-15 | 2021-07-16 | 北京京东乾石科技有限公司 | Storage position recommendation method and device |
CN111552829A (en) * | 2020-05-07 | 2020-08-18 | 北京海益同展信息科技有限公司 | Method and apparatus for analyzing image material |
WO2021223629A1 (en) * | 2020-05-07 | 2021-11-11 | 京东数科海益信息科技有限公司 | Method and device for analyzing image material |
CN111552829B (en) * | 2020-05-07 | 2023-06-27 | 京东科技信息技术有限公司 | Method and apparatus for analyzing image material |
CN112182264A (en) * | 2020-10-10 | 2021-01-05 | 书行科技(北京)有限公司 | Method, device and equipment for determining landmark information and readable storage medium |
CN113763211A (en) * | 2021-09-23 | 2021-12-07 | 支付宝(杭州)信息技术有限公司 | Infringement detection method and device based on block chain and electronic equipment |
Also Published As
Publication number | Publication date |
---|---|
CN104376052B (en) | 2017-07-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104376052A (en) | Same-style commodity merging method based on commodity images | |
Schönberger et al. | A vote-and-verify strategy for fast spatial verification in image retrieval | |
US11120068B2 (en) | Media fingerprinting and identification system | |
Lopez-Antequera et al. | Appearance-invariant place recognition by discriminatively training a convolutional neural network | |
Galvez-Lopez et al. | Real-time loop detection with bags of binary words | |
Philbin et al. | Lost in quantization: Improving particular object retrieval in large scale image databases | |
Gao et al. | Database saliency for fast image retrieval | |
Osokin et al. | Os2d: One-stage one-shot object detection by matching anchor features | |
Xu et al. | Near duplicate identification with spatially aligned pyramid matching | |
Gavves et al. | Visual synonyms for landmark image retrieval | |
Zheng et al. | Multi-scale attention vehicle re-identification | |
He et al. | Deep feature embedding learning for person re-identification based on lifted structured loss | |
Peng et al. | Eliminating cross-camera bias for vehicle re-identification | |
Romberg et al. | Bundle min-Hashing: Speeded-up object retrieval | |
Wu et al. | Self-similarity-based partial near-duplicate video retrieval and alignment | |
Hong et al. | Realtime and robust object matching with a large number of templates | |
Kordopatis-Zilos et al. | Finding near-duplicate videos in large-scale collections | |
Huo et al. | Person re-identification based on multi-directional saliency metric learning | |
Mei et al. | Instance-level object retrieval via deep region CNN | |
Li et al. | Fast key-frame image retrieval of intelligent city security video based on deep feature coding in high concurrent network environment | |
Li et al. | Instance image retrieval with generative adversarial training | |
Apostolidis et al. | Fast object re-detection and localization in video for spatio-temporal fragment creation | |
Cheng et al. | Stratification-based keyframe cliques for effective and efficient video representation | |
Ufer et al. | Object retrieval and localization in large art collections using deep multi-style feature fusion and iterative voting | |
Zhao et al. | Extracting representative motion flows for effective video retrieval |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170714 Termination date: 20201103 |