CN111382299A - Method, device, computer equipment and storage medium for accelerating image retrieval - Google Patents
Method, device, computer equipment and storage medium for accelerating image retrieval Download PDFInfo
- Publication number
- CN111382299A CN111382299A CN202010064425.4A CN202010064425A CN111382299A CN 111382299 A CN111382299 A CN 111382299A CN 202010064425 A CN202010064425 A CN 202010064425A CN 111382299 A CN111382299 A CN 111382299A
- Authority
- CN
- China
- Prior art keywords
- image
- node
- class
- target
- feature vector
- 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.)
- Withdrawn
Links
Images
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
-
- 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
- G06F16/55—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
Abstract
The invention discloses a method, a device, computer equipment and a storage medium for accelerating image retrieval, which respectively take the feature vectors of reference images as an initial class, determine two initial classes with the largest cosine similarity of the feature vectors as an intermediate class, determine an intermediate class root node and two subtrees, repeatedly execute the process of determining the two initial classes with the largest cosine similarity of the feature vectors as the intermediate class until only one initial class is left, so as to determine a binary clustering tree comprising the feature vectors of all the reference images, obtain a target feature vector, search the target root node in the binary clustering tree, search the child node with the smallest cosine similarity between the target root node and the target feature vector in the subtree of the target root node, determine the image difference which is most similar to the target image, and can improve the efficiency of retrieving the target image in the binary clustering tree, hardware cost for realizing corresponding image retrieval is reduced.
Description
Technical Field
The present invention relates to the field of image processing technologies, and in particular, to a method and an apparatus for accelerating image retrieval, a computer device, and a storage medium.
Background
At present, image identification on the internet can be classified into two problems, one is 'near-repeated retrieval', which is mainly used for retrieving the same source image through different deformations (including illumination, watermarking, zooming, local missing replacement and the like), or identifying generally similar objects, and is mainly applied to copyright protection, illegal identification, image duplication removal, basic similar retrieval and the like; the other is "local retrieval", which means that two pictures can be matched as long as some objects are repeated, for example, different models can be imagined, but as long as they span the same LV bag, they can be regarded as similar images, i.e., image retrieval in the true sense is realized.
In the conventional image retrieval, after the feature extraction is carried out on the image, the linear comparison of 1: N with the image library is required, the calculation amount is huge, the retrieval efficiency is easy to be low, and the hardware cost is high.
Disclosure of Invention
In view of the above problems, the present invention provides a method, computer device, and storage medium for speeding up image retrieval.
In order to achieve the purpose of the invention, the invention provides a method for accelerating image retrieval, which comprises the following steps:
s10, acquiring the feature vectors of a plurality of reference images, and taking each feature vector as an initial class; the initial class is a tree with only one node;
s20, determining two initial classes with the largest cosine similarity of the feature vectors as an intermediate class, determining the average vector of the two feature vectors corresponding to the intermediate class as the feature vector of the intermediate class root node, and taking the two initial classes corresponding to the intermediate class as two subtrees of the intermediate class root node;
s30, repeating the step S20 until only one initial class is left, and determining a clustering binary tree comprising the feature vectors of all the reference images according to each intermediate class and the remaining initial class;
s40, obtaining a feature vector of the target image, obtaining a target feature vector, searching an intermediate root-like node with the minimum cosine similarity difference between the intermediate root-like node and the target feature vector in the clustering binary tree, obtaining a target root node, searching a sub-node with the minimum cosine similarity difference between the sub-node and the target feature vector in a sub-tree of the target root node, and determining the image of the searched sub-node as the image most similar to the target image.
In one embodiment, obtaining feature vectors for a plurality of reference images comprises:
and extracting the last layer of convolution characteristics corresponding to each reference image by adopting the VGG16 as the characteristic vector of the corresponding reference image.
In one embodiment, the formula for determining the cosine similarity between two reference images comprises:
Q1＝[x1，x2，x3，…，x512]，
Q2＝[y1，y2，y3，...，y512]，
wherein cos (θ) represents cosine similarity, Q1Feature vector, Q, representing a reference image2Representing the feature vector of another reference image, and n has a value of 512.
An apparatus for expediting image retrieval, comprising:
the first acquisition module is used for acquiring the feature vectors of a plurality of reference images and taking each feature vector as an initial class; the initial class is a tree with only one node;
the determining module is used for determining two initial classes with the largest cosine similarity of the feature vectors as an intermediate class, determining the average vector of the two feature vectors corresponding to the intermediate class as the feature vector of the intermediate class root node, and taking the two initial classes corresponding to the intermediate class as two subtrees of the intermediate class root node;
the execution module is used for entering the determination module to repeatedly execute the corresponding process until only one initial class is left, and determining a clustering binary tree comprising the feature vectors of all the reference images according to each intermediate class and the remaining initial class;
and the second acquisition module is used for acquiring the feature vector of the target image, acquiring the target feature vector, searching an intermediate root-like node with the minimum cosine similarity difference between the intermediate root-like node and the target feature vector in the clustering binary tree to acquire a target root node, searching a sub-node with the minimum cosine similarity difference between the sub-node and the target feature vector in a sub-tree of the target root node, and determining the image of the searched sub-node as the image most similar to the target image.
In one embodiment, the first obtaining module is further configured to:
and extracting the last layer of convolution characteristics corresponding to each reference image by adopting the VGG16 as the characteristic vector of the corresponding reference image.
In one embodiment, the formula for determining the cosine similarity between two reference images comprises:
Q1＝[x1，x2，x3，…，x512]，
Q2＝[y1，y2，y3，...，y512]，
wherein cos (θ) represents cosine similarity, Q1Feature vector, Q, representing a reference image2Representing the feature vector of another reference image, and n has a value of 512.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the method for expediting image retrieval of any of the above embodiments when executing the computer program.
A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of accelerating an image retrieval of any of the above embodiments.
The method, the device, the computer equipment and the storage medium for accelerating the image retrieval are characterized in that the method comprises the steps of obtaining feature vectors of a plurality of reference images, respectively taking each feature vector as an initial class, determining two initial classes with the largest cosine similarity of the feature vectors as an intermediate class, determining an average vector of the two feature vectors corresponding to the intermediate class as the feature vector of a root node of the intermediate class, taking the two initial classes corresponding to the intermediate class as two subtrees of the root node of the intermediate class, repeatedly executing the process of determining the two initial classes with the largest cosine similarity of the feature vectors as the intermediate class until only one initial class is left, determining a clustering binary tree comprising the feature vectors of all the reference images according to each intermediate class and the remaining initial class, obtaining the feature vectors of the target images to obtain the target feature vectors, searching the root node of the intermediate class with the smallest cosine similarity difference between the clustering binary tree and the target feature vectors, and obtaining a target root node, searching a sub-node with the minimum cosine similarity difference with the target characteristic vector in a sub-tree of the target root node, and determining the image of the searched sub-node as the image most similar to the target image, so as to improve the efficiency of searching the target image in the clustering binary tree and reduce the hardware cost for realizing corresponding image search.
Drawings
FIG. 1 is a flow diagram of a method to expedite image retrieval according to one embodiment;
FIG. 2 is a diagram of an embodiment of an image clustering tree construction process;
FIG. 3 is a diagram of a final hierarchical clustering tree of an embodiment;
FIG. 4 is a block diagram of an apparatus for expediting image retrieval according to one embodiment;
FIG. 5 is a schematic diagram of a computer device of an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.
In one embodiment, as shown in fig. 1, there is provided a method for accelerating image retrieval, comprising the steps of:
s10, acquiring the feature vectors of a plurality of reference images, and taking each feature vector as an initial class; the initial class is a tree with only one node.
In one embodiment, obtaining feature vectors for a plurality of reference images comprises:
and extracting the last layer of convolution characteristics corresponding to each reference image by adopting the VGG16 as the characteristic vector of the corresponding reference image.
Specifically, there are many ways to extract feature vectors of an image, and this embodiment may use VGG16 to extract the last layer of convolution features as feature vectors of an image, where Q is ═ x1，x2，x3，…，x512]。
And S20, determining the two initial classes with the largest cosine similarity of the feature vectors as an intermediate class, determining the average vector of the two feature vectors corresponding to the intermediate class as the feature vector of the intermediate class root node, and taking the two initial classes corresponding to the intermediate class as two subtrees of the intermediate class root node.
In one embodiment, the formula for determining the cosine similarity between two reference images comprises:
Q1＝[x1，x2，x3，…，x512]，
Q2＝[y1，y2，y3，…，y512]，
wherein cos (θ) represents cosine similarity, Q1Feature vector, Q, representing a reference image2Representing the feature vector of another reference image, and n has a value of 512.
In this embodiment, the above cosine similarity formula (cosine similarity determination formula) may be used to calculate two eigenvectors Q1And Q2The greater the cosine similarity value, the more similar the two eigenvectors are.
Optionally, two feature vectors Q1And Q2The similarity of (A) is as follows: sim (Q)1，Q2)。
And S30, repeatedly executing the step S20 until only one initial class is left, and determining a clustering binary tree comprising the feature vectors of all the reference images according to each intermediate class and the remaining initial class.
Specifically, the determining process of the clustering binary tree may be:
(a) taking the feature vector of each reference image as an initial class, wherein each initial class is a tree with only one node;
(b) two class trees with the maximum cosine similarity of the characteristic vectors of the root nodes between the class trees (initial classes) are searched, the two class trees are classified into one class, a new root node is defined, and the two class trees are respectively used as a left sub-tree and a right sub-tree of the root node. And the feature vector of the root node is represented as an average vector of the feature vectors of the two child nodes.
(c) And (c) repeating the step (b) until only one class is left to stop, and obtaining a clustering binary tree of the feature vectors of all the reference images.
S40, obtaining a feature vector of the target image, obtaining a target feature vector, searching an intermediate root-like node with the minimum cosine similarity difference between the intermediate root-like node and the target feature vector in the clustering binary tree, obtaining a target root node, searching a sub-node with the minimum cosine similarity difference between the sub-node and the target feature vector in a sub-tree of the target root node, and determining the image of the searched sub-node as the image most similar to the target image.
The steps can realize the retrieval of the target image in the clustering binary tree. The specific retrieval process may be:
(a) acquiring a feature vector D of a target image;
(b) in a binary clustering tree of feature vectors, hierarchical traversal is performed from the root node of the tree downwards
(c) In each layer, calculating the cosine similarity between the feature vector of each node and the feature vector D of the target image, searching for a node with higher cosine similarity, and then continuously traversing the subtree taking the node as a root node from the node;
(d) and (c) repeating the step (c) until a leaf node is reached, wherein the feature vector of the node is most similar to the feature vector D of the target image, and the image represented by the node is the most similar image of the target image.
Further, in the practical application process, if images similar to topN (target image) need to be searched, a search path for searching the most similar image may be recorded, leaves of the most similar image traverse along the search path in the reverse direction, and each time a node is traversed, all leaf nodes below the node may be considered as images similar to the target image. Nodes with deeper depths can be considered to be more similar.
The method for accelerating the image retrieval comprises the steps of obtaining the feature vectors of a plurality of reference images, respectively taking each feature vector as an initial class, determining two initial classes with the largest cosine similarity of the feature vectors as an intermediate class, determining the average vector of the two feature vectors corresponding to the intermediate class as the feature vector of an intermediate class root node, taking the two initial classes corresponding to the intermediate class as two subtrees of the intermediate class root node, repeatedly executing the process of determining the two initial classes with the largest cosine similarity of the feature vectors as the intermediate class until only one initial class remains, determining a clustering binary tree comprising the feature vectors of all the reference images according to the intermediate classes and the remaining initial class, obtaining the feature vectors of the target images, searching the intermediate class root node with the smallest cosine similarity between the clustering binary tree and the target feature vectors, and obtaining a target root node, searching a sub-node with the minimum cosine similarity difference with the target characteristic vector in a sub-tree of the target root node, and determining the image of the searched sub-node as the image most similar to the target image, so as to improve the efficiency of searching the target image in the clustering binary tree and reduce the hardware cost for realizing corresponding image search.
In one embodiment, referring to FIG. 2, assuming A, B, C, D, E is 5 images, 5 feature vectors may be calculated for the 5 images.
The image clustering tree (clustering binary tree) construction process is as follows:
step1, each image as an independent class, for a total of 5 classes;
step2, finding two types A and B with the maximum similarity, and then totally 4 types, namely { (A, B), C, D, E };
step3, finding the two types D and E with the largest similarity, and then totally 3 types, namely { (A, B), C, (D, E) };
step4, calculating the average eigenvector of (A, B), calculating the average eigenvector of (D, E), then calculating the similarity of every two with C respectively, finding that the similarity of C and (A, B) is larger than the similarity of C and (D, E) is larger than the similarity of (A, B) and (D, E), so that (A, B) and C are classified into a new class, and then 2 classes are obtained in total, namely { ((A, B), C), (D, E) };
step5, leaving only two classes, so merge directly to get the final class { (((A, B), C), (D, E)) }.
The final hierarchical clustering tree (binary clustering tree) can be referred to as shown in fig. 3:
except for the root node, the feature vector representation of each cluster node (i.e., black nodes C1-C4 in the figure) is the average of the feature vectors of A and B with the method C1; c2 is the average of the eigenvectors of C1 and C, and so on.
When similar images of the target image F need to be searched (assuming that C1-C4 and A-E all represent feature vectors of all nodes), the specific process is as follows:
(1) calculating the feature vector of F as QF
(2) The cosine similarity of each subtree root node C2 and C3 of QF and C4 is calculated,
if sim (QF, C1) is larger than sim (QF, C2), the similarity of each subtree root node of QF and C1 is continuously compared
If sim (QF, C1) is smaller than sim (QF, C2), the similarity of each subtree root node of QF and C2 is continuously compared
(3) By analogy, until the last leaf node is found, assuming that the leaf node most similar to QF is C, then the most similar image of image F is C, and our search path is C4- > C2- > C
(4) When needing to search images similar to topN, defining a similar image set as L, reversely traversing from a leaf node C along a retrieval path, then adding leaf nodes of subtrees of all the nodes into the L, and sorting the similarity, namely respectively calculating the similarity between QF and each leaf node, and then sorting
Traversing to C2 in reverse, sim (QF, a) and sim (QF, B) need to be calculated respectively, and assuming sim (QF, a) > sim (QF, B), L ═ C, a, B } (the order in the set is similarity ranking);
when backward traversing to C4, sim (QF, D) and sim (QF, E) need to be calculated respectively, and assuming sim (QF, E) > sim (QF, D), L ═ C, a, B, E, D } (the order in the set is similarity ranking);
then, finally, the image top1 ═ { C }, top2 ═ C, a }, which is similar to F;
top3 ═ { C, a, B }, and so on.
In one embodiment, referring to fig. 4, there is provided an apparatus for accelerating image retrieval, including:
a first obtaining module 10, configured to obtain feature vectors of multiple reference images, and use each feature vector as an initial class; the initial class is a tree with only one node;
a determining module 20, configured to determine two initial classes with the largest cosine similarity of the feature vectors as an intermediate class, determine an average vector of the two feature vectors corresponding to the intermediate class as a feature vector of an intermediate class root node, and use the two initial classes corresponding to the intermediate class as two subtrees of the intermediate class root node;
an executing module 30, configured to enter the determining module to repeatedly execute the corresponding process until only one initial class remains, and determine a binary clustering tree including feature vectors of all reference images according to each intermediate class and the remaining initial class;
the second obtaining module 40 is configured to obtain a feature vector of the target image, obtain a target feature vector, find an intermediate root-like node with a smallest cosine similarity difference between the clustering binary tree and the target feature vector, obtain a target root node, find a child node with a smallest cosine similarity difference between the child node and the target feature vector in a subtree of the target root node, and determine an image of the found child node as an image most similar to the target image.
In one embodiment, the first obtaining module is further configured to:
and extracting the last layer of convolution characteristics corresponding to each reference image by adopting the VGG16 as the characteristic vector of the corresponding reference image.
In one embodiment, the formula for determining the cosine similarity between two reference images comprises:
Q1＝[x1，x2，x3，…，x512]，
Q2＝[y1，y2，y3，...，y512]，
wherein cos (θ) represents cosine similarity, Q1Feature vector, Q, representing a reference image2Representing the feature vector of another reference image, and n has a value of 512.
For specific limitations of the apparatus for accelerating image retrieval, reference may be made to the above limitations of the method for accelerating image retrieval, which are not described herein again. The modules in the device for accelerating image retrieval can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 5. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of expediting image retrieval. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the architecture shown in fig. 5 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
Based on the examples described above, there is also provided in one embodiment a computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor, when executing the program, implements any of the methods of accelerating image retrieval as described in the embodiments above.
It will be understood by those skilled in the art that all or part of the processes in the methods of the embodiments described above may be implemented by a computer program, which may be stored in a non-volatile computer-readable storage medium, and in the embodiments of the present invention, the program may be stored in the storage medium of a computer system and executed by at least one processor in the computer system to implement the processes including the embodiments of the method for accelerating image retrieval described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
Accordingly, in an embodiment, a computer storage medium and a computer readable storage medium are also provided, on which a computer program is stored, wherein the program, when executed by a processor, implements any of the methods for expediting image retrieval as described in the embodiments above.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
It should be noted that the terms "first \ second \ third" referred to in the embodiments of the present application merely distinguish similar objects, and do not represent a specific ordering for the objects, and it should be understood that "first \ second \ third" may exchange a specific order or sequence when allowed. It should be understood that "first \ second \ third" distinct objects may be interchanged under appropriate circumstances such that the embodiments of the application described herein may be implemented in an order other than those illustrated or described herein.
The terms "comprising" and "having" and any variations thereof in the embodiments of the present application are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product, or device that comprises a list of steps or modules is not limited to the listed steps or modules but may alternatively include other steps or modules not listed or inherent to such process, method, product, or device.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (8)
1. A method for accelerating image retrieval is characterized by comprising the following steps:
s10, acquiring the feature vectors of a plurality of reference images, and taking each feature vector as an initial class; the initial class is a tree with only one node;
s20, determining two initial classes with the largest cosine similarity of the feature vectors as an intermediate class, determining the average vector of the two feature vectors corresponding to the intermediate class as the feature vector of the intermediate class root node, and taking the two initial classes corresponding to the intermediate class as two subtrees of the intermediate class root node;
s30, repeating the step S20 until only one initial class is left, and determining a clustering binary tree comprising the feature vectors of all the reference images according to each intermediate class and the remaining initial class;
s40, obtaining a feature vector of the target image, obtaining a target feature vector, searching an intermediate root-like node with the minimum cosine similarity difference between the intermediate root-like node and the target feature vector in the clustering binary tree, obtaining a target root node, searching a sub-node with the minimum cosine similarity difference between the sub-node and the target feature vector in a sub-tree of the target root node, and determining the image of the searched sub-node as the image most similar to the target image.
2. The method of expediting image retrieval of claim 1, wherein in one embodiment, obtaining feature vectors for a plurality of reference images comprises:
and extracting the last layer of convolution characteristics corresponding to each reference image by adopting the VGG16 as the characteristic vector of the corresponding reference image.
3. The method of claim 1, wherein the determining formula of cosine similarity between two reference images comprises:
Q1＝[x1，x2，x3，...，x512]，
Q2＝[y1，y2，y3，...，y512]，
wherein cos (θ) represents cosine similarity, Q1Feature vector, Q, representing a reference image2Representing the feature vector of another reference image, and n has a value of 512.
4. An apparatus for expediting image retrieval, comprising:
the first acquisition module is used for acquiring the feature vectors of a plurality of reference images and taking each feature vector as an initial class; the initial class is a tree with only one node;
the determining module is used for determining two initial classes with the largest cosine similarity of the feature vectors as an intermediate class, determining the average vector of the two feature vectors corresponding to the intermediate class as the feature vector of the intermediate class root node, and taking the two initial classes corresponding to the intermediate class as two subtrees of the intermediate class root node;
the execution module is used for entering the determination module to repeatedly execute the corresponding process until only one initial class is left, and determining a clustering binary tree comprising the feature vectors of all the reference images according to each intermediate class and the remaining initial class;
and the second acquisition module is used for acquiring the feature vector of the target image, acquiring the target feature vector, searching an intermediate root-like node with the minimum cosine similarity difference between the intermediate root-like node and the target feature vector in the clustering binary tree to acquire a target root node, searching a sub-node with the minimum cosine similarity difference between the sub-node and the target feature vector in a sub-tree of the target root node, and determining the image of the searched sub-node as the image most similar to the target image.
5. The apparatus for expediting image retrieval of claim 4, wherein in one embodiment, the first obtaining module is further configured to:
and extracting the last layer of convolution characteristics corresponding to each reference image by adopting the VGG16 as the characteristic vector of the corresponding reference image.
6. The apparatus for speeding up image retrieval according to claim 4, wherein in one embodiment, the formula for determining cosine similarity between two reference images comprises:
Q1＝[x1，x2，x3，...，x512]，
Q2＝[y1，y2，y3，...，y512]，
wherein cos (θ) represents cosine similarity, Q1Feature vector, Q, representing a reference image2Representing the feature vector of another reference image, and n has a value of 512.
7. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the method of expediting image retrieval of any one of claims 1 to 6 when executing the computer program.
8. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the augmented reality interaction method of any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010064425.4A CN111382299A (en) | 2020-01-20 | 2020-01-20 | Method, device, computer equipment and storage medium for accelerating image retrieval |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010064425.4A CN111382299A (en) | 2020-01-20 | 2020-01-20 | Method, device, computer equipment and storage medium for accelerating image retrieval |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111382299A true CN111382299A (en) | 2020-07-07 |
Family
ID=71218608
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010064425.4A Withdrawn CN111382299A (en) | 2020-01-20 | 2020-01-20 | Method, device, computer equipment and storage medium for accelerating image retrieval |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111382299A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114637873A (en) * | 2022-03-30 | 2022-06-17 | 徐州大工电子科技有限公司 | Intelligent door and window recommendation method and system based on image similarity |
-
2020
- 2020-01-20 CN CN202010064425.4A patent/CN111382299A/en not_active Withdrawn
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114637873A (en) * | 2022-03-30 | 2022-06-17 | 徐州大工电子科技有限公司 | Intelligent door and window recommendation method and system based on image similarity |
| CN114637873B (en) * | 2022-03-30 | 2022-12-23 | 徐州大工电子科技有限公司 | Intelligent door and window recommendation method and system based on image similarity |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Face anti-spoofing via disentangled representation learning | |
| Moreira et al. | Image provenance analysis at scale | |
| CN103336957B (en) | A kind of network homology video detecting method based on space-time characteristic | |
| Zhi et al. | Two-stage pooling of deep convolutional features for image retrieval | |
| WO2013028302A1 (en) | Fast matching of image features using multi-dimensional tree data structures | |
| Miele et al. | Revisiting animal photo‐identification using deep metric learning and network analysis | |
| Park et al. | Fast and robust copy-move forgery detection based on scale-space representation | |
| Kiranyaz et al. | Automatic object extraction over multiscale edge field for multimedia retrieval | |
| CN114694185B (en) | Cross-modal target re-identification method, device, equipment and medium | |
| Wary et al. | A review on robust video copy detection | |
| Dhiman et al. | A novel video shot boundary detection framework employing DCT and pattern matching | |
| Sekhar et al. | A study on segmentation-based copy-move forgery detection using DAISY descriptor | |
| CN109697240A (en) | A kind of image search method and device based on feature | |
| Cuffaro et al. | Segmentation free object discovery in video | |
| CN111382299A (en) | Method, device, computer equipment and storage medium for accelerating image retrieval | |
| Zhang et al. | Rc-cnn: Reverse connected convolutional neural network for accurate player detection | |
| Maigrot et al. | Tampering detection and localization in images from social networks: A CBIR approach | |
| Lei et al. | Fast near-duplicate image detection using uniform randomized trees | |
| Chan et al. | F-sort: An alternative for faster geometric verification | |
| Yuan et al. | A novel figure panel classification and extraction method for document image understanding | |
| Syam et al. | An effective similarity measure via genetic algorithm for Content-Based Image Retrieval with extensive features | |
| Kim et al. | Multi‐camera‐based security log management scheme for smart surveillance | |
| CN111143626A (en) | Group partner identification method, device, equipment and computer readable storage medium | |
| Goshen et al. | Balanced exploration and exploitation model search for efficient epipolar geometry estimation | |
| JP4768358B2 (en) | Image search method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20200707 |