CN107590154B - Object similarity determination method and device based on image recognition - Google Patents
Object similarity determination method and device based on image recognition Download PDFInfo
- Publication number
- CN107590154B CN107590154B CN201610534932.3A CN201610534932A CN107590154B CN 107590154 B CN107590154 B CN 107590154B CN 201610534932 A CN201610534932 A CN 201610534932A CN 107590154 B CN107590154 B CN 107590154B
- Authority
- CN
- China
- Prior art keywords
- image
- similarity
- preset threshold
- query
- feature
- 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.)
- Active
Links
Images
Abstract
The application discloses an object similarity determination method and device based on image recognition, wherein the method comprises the following steps: receiving an input query image; retrieving according to the query image to obtain a result image; acquiring a first similarity and a second similarity between a query image and a result image; and judging whether the object in the query image and the object in the result image are the same or similar objects or not according to the first similarity and the second similarity. According to the object similarity determination method and device based on image recognition, whether the object in the query image and the object in the result image are the same or similar or not is comprehensively determined according to the first similarity and the second similarity by obtaining the first similarity and the second similarity between the query image and the result image, the object in the query image and the object in the result image can be rapidly and accurately determined to be the same or similar, recognition efficiency is improved, and recognition accuracy is improved.
Description
Technical Field
The present application relates to the field of computer technologies, and in particular, to a method and an apparatus for determining object similarity based on image recognition.
Background
Image searching is an important means for users to obtain information. The user can directly upload the pictures and search for the pictures with the same or similar contents. Particularly in an e-commerce application scene, the method can effectively help users to find out commodities which are interesting by themselves and cannot be accurately described by characters from massive commodities. Currently, after obtaining search results, a user needs to check and confirm whether the products are the same type of products required by the user and select more interesting products from a plurality of search results one by one. However, the above method consumes a lot of time and cost, and the user experience is poor. Therefore, a method for quickly identifying the same type of goods required by the user is needed.
Content of application
The present application is directed to solving, at least to some extent, one of the technical problems in the related art. Therefore, an object of the present application is to provide an object similarity determination method based on image recognition, which can quickly and accurately determine that an object in a query image and an object in a result image are the same or similar, improve recognition efficiency, and improve recognition accuracy.
A second object of the present application is to provide an object similarity determination device based on image recognition.
A third objective of the present application is to provide an information push system based on image recognition.
In order to achieve the above object, an embodiment of a first aspect of the present application provides an object similarity determination method based on image recognition, including: receiving an input query image; retrieving according to the query image to obtain a result image; acquiring a first similarity and a second similarity between the query image and the result image, wherein the first similarity comprises an image overall feature similarity, and the second similarity comprises an image local feature similarity; and judging whether the object in the query image and the object in the result image are the same or similar objects or not according to the first similarity and the second similarity.
According to the object similarity determination method based on image recognition, whether the object in the query image and the object in the result image are the same or similar or not is comprehensively determined according to the first similarity and the second similarity by obtaining the first similarity and the second similarity between the query image and the result image, the object in the query image and the object in the result image can be rapidly and accurately determined to be the same or similar, recognition efficiency is improved, and recognition accuracy is improved.
An embodiment of a second aspect of the present application provides an object similarity determination apparatus based on image recognition, including: the receiving module is used for receiving an input query image; the retrieval module is used for retrieving and obtaining a result image according to the query image; the acquisition module is used for acquiring a first similarity and a second similarity between the query image and the result image, wherein the first similarity comprises the image overall feature similarity, and the second similarity comprises the image local feature similarity; and the judging module is used for judging whether the object in the query image and the object in the result image are the same or similar objects according to the first similarity and the second similarity.
The object similarity determination device based on image recognition in the embodiment of the application comprehensively determines whether the object in the query image and the object in the result image are the same or similar objects according to the first similarity and the second similarity by acquiring the first similarity and the second similarity between the query image and the result image, so that the object in the query image and the object in the result image can be quickly and accurately determined to be the same or similar objects, the recognition efficiency is improved, and the recognition accuracy is improved.
An embodiment of a third aspect of the present application provides an information push system based on image recognition, including: the client is used for receiving a query image input by a user and sending the query image to the server; the server side comprises the object similarity determination device based on image recognition in the embodiment of the second aspect; the server is further configured to, after determining that the object in the query image and the object in the result image are the same or similar, push the relevant information of the object in the result image to the client, so that the client displays the relevant information.
The information pushing system based on image recognition can rapidly and accurately determine the result image which is the same as or similar to the object in the query image, so that the related information of the object in the result image is pushed to the user, the pushed information is more accurate, the user requirements are met, and the user use experience is improved.
Drawings
Fig. 1 is a flowchart of an object similarity determination method based on image recognition according to an embodiment of the present application;
FIG. 2 is a flow diagram of obtaining a first similarity and a second similarity between a query image and a result image according to one embodiment of the present application;
FIG. 3 is a flow chart of determining whether the objects are the same or similar objects according to one embodiment of the present application;
fig. 4 is a schematic structural diagram of an object similarity determination device based on image recognition according to an embodiment of the present application.
Detailed Description
Reference will now be made in detail to embodiments of the present application, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are exemplary and intended to be used for explaining the present application and should not be construed as limiting the present application.
With the rapid development of the internet, more and more users begin to use the method of searching images to obtain the information required by themselves. For example: the user sees a favorite commodity, but does not know the name, the manufacturer, the specific style and the like of the commodity, so that the user can upload the image corresponding to the commodity to a search engine, and the search engine can search the same or similar image and provide the information related to the commodity for the user. However, if the product in the image of the search result and the image input by the user is the same, the user needs to manually recognize the product. In order to solve the above problem, the present application provides an object similarity determination method based on image recognition.
An object similarity determination method and device based on image recognition according to an embodiment of the present application will be described below with reference to the drawings.
Fig. 1 is a flowchart of an object similarity determination method based on image recognition according to an embodiment of the present application.
As shown in fig. 1, the object similarity determination method based on image recognition may include:
and S1, receiving the input query image.
For example, a user may upload a query image containing an item to a search engine to cause the search engine to retrieve based on the query image.
And S2, obtaining a result image according to the query image retrieval.
The search engine may perform a search based on the query image to obtain a corresponding result image. The result image contains a large number of identical or similar images, and therefore, the result image needs to be further identified to determine whether the object in the result image is identical or similar to the object in the query image.
S3, acquiring a first similarity and a second similarity between the query image and the result image.
In the related art, whether an object in a result image is the same as or similar to an object in a query image is determined mainly by obtaining a similarity between the query image and the result image. The judgment that the similarity is higher than a certain numerical value is the same or similar, and the judgment that the similarity is lower than the certain numerical value is different or dissimilar. The difference between the present application and the related art is that the present application adopts the similarity of two dimensions for determination, that is, four preset thresholds are respectively set for the first similarity and the second similarity, and the comprehensive determination is performed according to the preset thresholds. The first similarity comprises the image overall feature similarity, and the second similarity comprises the image local feature similarity.
Specifically, as shown in fig. 2, obtaining the first similarity and the second similarity between the query image and the result image may include the following steps:
and S31, extracting the first image characteristic and the second image characteristic of the query image.
And S32, extracting a third image feature and a fourth image feature of the result image.
The first image feature and the third image feature are image overall features extracted based on a CNN (Convolutional Neural Network). The second image feature and the fourth image feature are image local features extracted based on a Scale-invariant feature transform (SIFT) algorithm.
And S33, acquiring a first similarity according to the first image characteristic and the third image characteristic.
In particular, the first image feature and the third image feature may be linearly transformed based on a linear transformation matrix. The linear transformation matrix is obtained according to image sample statistics of the same or similar objects. The linear transformation matrix is an eigenvector matrix of a covariance matrix of image samples of the same or similar objects to feature differences, and can enable the distances between the image features of the same or similar objects to be closer, namely the similarity to be higher, thereby improving the recall rate of the same or similar object judgment. After the first image feature and the third image feature are linearly transformed, a distance between the linearly transformed first image feature and the linearly transformed third image feature may be calculated to obtain a first similarity, i.e., a CNN similarity. In this embodiment, the character C may be used. The distance may be, but is not limited to, euclidean distance or euclidean distance.
And S34, acquiring a second similarity according to the second image characteristic and the fourth image characteristic.
And the second similarity is the local similarity of the images based on the SIFT algorithm. In this embodiment, the character L may be used.
And S4, judging whether the object in the query image and the object in the result image are the same or similar objects according to the first similarity and the second similarity.
After the first similarity and the second similarity are obtained, whether the object in the query image and the object in the result image are the same or similar objects can be determined according to the first similarity and the second similarity.
Specifically, the first similarity C may be compared with a first preset threshold C1, a second preset threshold C2, a third preset threshold C3 and a fourth preset threshold C4 respectively to obtain a first comparison result. The first preset threshold is greater than a second preset threshold, the second preset threshold is greater than a third preset threshold, and the third preset threshold is greater than a fourth preset threshold, namely C1> C2> C3> C4.
And comparing the second similarity L with a fifth preset threshold L1, a sixth preset threshold L2, a seventh preset threshold L3 and an eighth preset threshold L4 respectively to obtain a second comparison result. The fifth preset threshold is greater than a sixth preset threshold, the sixth preset threshold is greater than a seventh preset threshold, and the seventh preset threshold is greater than an eighth preset threshold, namely L1> L2> L3> L4.
It should be understood that the eight preset thresholds are empirical thresholds obtained through a plurality of experiments.
After the first comparison result and the second comparison result are obtained, whether the object in the query image and the object in the result image are the same or similar objects may be determined according to the first comparison result and the second comparison result.
In particular, as shown in fig. 3. First, whether the second similarity is greater than a fifth preset threshold is judged. And if the second similarity is larger than a fifth preset threshold value, determining that the object in the query image and the object in the result image are the same or similar objects. I.e., L > L1, at this time, regardless of the value of C, it can be determined that the object in the query image is the same as or similar to the object in the result image.
Otherwise, further judging whether the second similarity is larger than a sixth preset threshold value and whether the first similarity is smaller than the first preset threshold value. And if the second similarity is greater than a sixth preset threshold and the first similarity is less than the first preset threshold, determining that the object in the query image and the object in the result image are the same or similar objects. I.e., L1> L2 and C < C1, it can be determined that the object in the query image is the same as or similar to the object in the result image.
Otherwise, further judging whether the second similarity is larger than a seventh preset threshold value and whether the first similarity is smaller than the second preset threshold value. And if the second similarity is greater than a seventh preset threshold and the first similarity is less than the second preset threshold, determining that the object in the query image and the object in the result image are the same or similar objects. I.e., L2> L3 and C < C2, it can be determined that the object in the query image is the same as or similar to the object in the result image.
Otherwise, further judging whether the second similarity is greater than an eighth preset threshold, and whether the first similarity is smaller than a third preset threshold. And if the second similarity is greater than an eighth preset threshold and the first similarity is less than a third preset threshold, determining that the object in the query image and the object in the result image are the same or similar objects. I.e., L3> L4 and C < C3, it can be determined that the object in the query image is the same as or similar to the object in the result image.
Otherwise, further judging whether the first similarity is smaller than a fourth preset threshold. And if the first similarity is smaller than a fourth preset threshold, determining that the object in the query image and the object in the result image are the same or similar objects. I.e., C < C4, at this time, regardless of the value of L, it can be determined that the object in the query image is the same as or similar to the object in the result image.
Otherwise, determining that the object in the query image and the object in the result image are non-identical or non-similar objects.
For example, in an e-commerce application scenario, the object may be a commodity. Because the user does not know the relevant information of the current commodity, such as a brand, a style and the like, the commodity needs to be photographed to obtain a corresponding query image, so that the same-style commodity is queried in an e-commerce search engine according to the query image, and the relevant information corresponding to the same-style commodity is obtained. The commodities of the same type can be the same in shape and style, and the commodities of different selling merchants, or different colors or different selling prices. After determining that the goods in the query image and the goods in the result image are the same-style goods, the related information such as the selling price and the like determined as the same-style goods can be acquired. And then the same-style commodities are sorted by combining other sorting factors such as sales volume, popularity and the like, so that the user can more quickly acquire the same-style commodities meeting the requirements of the user.
According to the object similarity determination method based on image recognition, whether the object in the query image and the object in the result image are the same or similar or not is comprehensively determined according to the first similarity and the second similarity by obtaining the first similarity and the second similarity between the query image and the result image, the object in the query image and the object in the result image can be rapidly and accurately determined to be the same or similar, recognition efficiency is improved, and recognition accuracy is improved.
In order to achieve the above object, the present application also provides an object similarity determination device based on image recognition.
Fig. 4 is a schematic structural diagram of an object similarity determination device based on image recognition according to an embodiment of the present application.
As shown in fig. 4, the object similarity determination device based on image recognition may include: a receiving module 110, a retrieving module 120, an obtaining module 130, and a determining module 140. The obtaining module 130 may include a first extracting unit 131, a second extracting unit 132, a first obtaining unit 133, and a second obtaining unit 134. The determination module 140 may include a third acquisition unit 141, a fourth acquisition unit 142, and a determination unit 143.
The receiving module 110 is used for receiving an input query image. For example, a user may upload a query image containing an item to a search engine to cause the search engine to retrieve based on the query image.
The retrieval module 120 is configured to retrieve a result image according to the query image. The retrieval module 120 may perform a retrieval based on the query image to obtain a corresponding result image. The result image contains a large number of identical or similar images, and therefore, the result image needs to be further identified to determine whether the object in the result image is identical or similar to the object in the query image.
The obtaining module 130 is configured to obtain a first similarity and a second similarity between the query image and the result image. The first similarity comprises the image overall feature similarity, and the second similarity comprises the image local feature similarity. The obtaining module 130 may include a first extracting unit 131, a second extracting unit 132, a first obtaining unit 133, and a second obtaining unit 134.
Specifically, the first extraction unit 131 may extract a first image feature and a second image feature of the query image. The second extraction unit 132 may extract a third image feature and a fourth image feature of the result image. The first image feature and the third image feature are image overall features extracted based on a CNN (Convolutional Neural Network). The second image feature and the fourth image feature are image local features extracted based on a Scale-invariant feature transform (SIFT) algorithm.
The first acquiring unit 133 may acquire the first similarity from the first image feature and the third image feature. In particular, the first image feature and the third image feature may be linearly transformed based on a linear transformation matrix. The linear transformation matrix is obtained according to image sample statistics of the same or similar objects. The linear transformation matrix is an eigenvector matrix of a covariance matrix of image samples of the same or similar objects to feature differences, and can enable the distances between the image features of the same or similar objects to be closer, namely the similarity to be higher, thereby improving the recall rate of the same or similar object judgment. After the first image feature and the third image feature are linearly transformed, a distance between the linearly transformed first image feature and the linearly transformed third image feature may be calculated to obtain a first similarity, i.e., a CNN similarity. In this embodiment, the character C may be used. The distance may be, but is not limited to, euclidean distance or euclidean distance.
The second obtaining unit 134 may obtain the second similarity degree according to the second image feature and the fourth image feature. And the second similarity is the local similarity of the images based on the SIFT algorithm. In this embodiment, the character L may be used.
The determining module 140 is configured to determine whether the object in the query image and the object in the result image are the same or similar according to the first similarity and the second similarity. The determination module 140 may include a third obtaining unit 141, a fourth obtaining unit 142, and a determining unit 143.
Specifically, the third obtaining unit 141 may compare the first similarity degree C with a first preset threshold C1, a second preset threshold C2, a third preset threshold C3, and a fourth preset threshold C4, respectively, to obtain a first comparison result. The first preset threshold is greater than a second preset threshold, the second preset threshold is greater than a third preset threshold, and the third preset threshold is greater than a fourth preset threshold, namely C1> C2> C3> C4.
The fourth obtaining unit 142 may compare the second similarity degree L with a fifth preset threshold L1, a sixth preset threshold L2, a seventh preset threshold L3, and an eighth preset threshold L4, respectively, to obtain a second comparison result. The fifth preset threshold is greater than a sixth preset threshold, the sixth preset threshold is greater than a seventh preset threshold, and the seventh preset threshold is greater than an eighth preset threshold, namely L1> L2> L3> L4.
It should be understood that the eight preset thresholds are empirical thresholds obtained through a plurality of experiments.
The determination unit 143 may determine whether the object in the query image and the object in the result image are the same or similar objects according to the first comparison result and the second comparison result.
In particular, as shown in fig. 3. First, the determination unit 143 determines whether the second similarity is greater than a fifth preset threshold. And if the second similarity is larger than a fifth preset threshold value, determining that the object in the query image and the object in the result image are the same or similar objects. I.e., L > L1, at this time, regardless of the value of C, it can be determined that the object in the query image is the same as or similar to the object in the result image.
Otherwise, further judging whether the second similarity is larger than a sixth preset threshold value and whether the first similarity is smaller than the first preset threshold value. And if the second similarity is greater than a sixth preset threshold and the first similarity is less than the first preset threshold, determining that the object in the query image and the object in the result image are the same or similar objects. I.e., L1> L2 and C < C1, it can be determined that the object in the query image is the same as or similar to the object in the result image.
Otherwise, further judging whether the second similarity is larger than a seventh preset threshold value and whether the first similarity is smaller than the second preset threshold value. And if the second similarity is greater than a seventh preset threshold and the first similarity is less than the second preset threshold, determining that the object in the query image and the object in the result image are the same or similar objects. I.e., L2> L3 and C < C2, it can be determined that the object in the query image is the same as or similar to the object in the result image.
Otherwise, further judging whether the second similarity is greater than an eighth preset threshold, and whether the first similarity is smaller than a third preset threshold. And if the second similarity is greater than an eighth preset threshold and the first similarity is less than a third preset threshold, determining that the object in the query image and the object in the result image are the same or similar objects. I.e., L3> L4 and C < C3, it can be determined that the object in the query image is the same as or similar to the object in the result image.
Otherwise, further judging whether the first similarity is smaller than a fourth preset threshold. And if the first similarity is smaller than a fourth preset threshold, determining that the object in the query image and the object in the result image are the same or similar objects. I.e., C < C4, at this time, regardless of the value of L, it can be determined that the object in the query image is the same as or similar to the object in the result image.
Otherwise, determining that the object in the query image and the object in the result image are non-identical or non-similar objects.
The object similarity determination device based on image recognition in the embodiment of the application comprehensively determines whether the object in the query image and the object in the result image are the same or similar objects according to the first similarity and the second similarity by acquiring the first similarity and the second similarity between the query image and the result image, so that the object in the query image and the object in the result image can be quickly and accurately determined to be the same or similar objects, the recognition efficiency is improved, and the recognition accuracy is improved.
In order to achieve the above purpose, the present application also provides an information pushing system based on image recognition.
The information push system based on image recognition can comprise a client side and a server side.
The client is used for receiving a query image input by a user and sending the query image to the server.
The server includes the object similarity determination device based on image recognition described in the previous embodiment. That is, the server side can determine whether the object in the query image and the object in the retrieved result image are the same or similar objects according to the received query image. If the objects are the same or similar objects, the server side can push the related information of the objects in the result image to the client side.
The client may present the user after receiving the relevant information.
For example, in an e-commerce application scenario, the object may be a commodity. Because the user does not know the relevant information of the current commodity, such as a brand, a style and the like, the commodity needs to be photographed to obtain a corresponding query image, so that the same-style commodity is queried in an e-commerce search engine according to the query image, and the relevant information corresponding to the same-style commodity is obtained. The relevant information may then be pushed to the client for presentation. The commodities of the same type can be the same in shape and style, and the commodities of different selling merchants, or different colors or different selling prices.
The information pushing system based on image recognition can rapidly and accurately determine the result image which is the same as or similar to the object in the query image, so that the related information of the object in the result image is pushed to the user, the pushed information is more accurate, the user requirements are met, and the user use experience is improved.
In the description herein, reference to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the application. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Although embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present application, and that variations, modifications, substitutions and alterations may be made to the above embodiments by those of ordinary skill in the art within the scope of the present application.
Claims (13)
1. An object similarity determination method based on image recognition is characterized by comprising the following steps:
receiving an input query image;
retrieving according to the query image to obtain a result image;
acquiring a first similarity and a second similarity between the query image and the result image, wherein the first similarity comprises an image overall feature similarity, and the second similarity comprises an image local feature similarity;
determining whether the object in the query image and the object in the result image are the same or similar objects according to the first similarity and the second similarity, wherein, whether the second similarity is larger than a fifth preset threshold value or not is judged, if the second similarity is larger than the fifth preset threshold value, then the object in the query image is determined to be the same or a similar object as the object in the result image, if the second similarity is smaller than the fifth preset threshold, further judging whether the second similarity is larger than a sixth preset threshold, and whether the first similarity is smaller than a first preset threshold, the fifth preset threshold is larger than the sixth preset threshold, if the second similarity is greater than the sixth preset threshold and the first similarity is less than the first preset threshold, then it is determined that the object in the query image is the same or similar object as the object in the result image.
2. The method of claim 1, wherein obtaining a first similarity and a second similarity between the query image and the result image comprises:
extracting a first image feature and a second image feature of the query image;
extracting a third image feature and a fourth image feature of the result image;
acquiring the first similarity according to the first image characteristic and the third image characteristic, wherein the first image characteristic and the third image characteristic are image overall characteristics extracted based on a Convolutional Neural Network (CNN);
and acquiring the second similarity according to the second image feature and the fourth image feature, wherein the second image feature and the fourth image feature are image local features extracted based on a SIFT algorithm.
3. The method of claim 2, wherein obtaining the first similarity from the first image feature and the third image feature comprises:
performing linear transformation on the first image characteristic and the third image characteristic based on a linear transformation matrix, wherein the linear transformation matrix is obtained according to image sample statistics of the same or similar objects;
and calculating the distance between the first image characteristic after linear transformation and the third image characteristic after linear transformation to obtain the first similarity.
4. The method of claim 1, further comprising:
if the second similarity is smaller than the sixth preset threshold, further judging whether the second similarity is larger than a seventh preset threshold, and whether the first similarity is smaller than a second preset threshold, wherein the sixth preset threshold is larger than the seventh preset threshold, and the first preset threshold is larger than the second preset threshold;
and if the second similarity is greater than the seventh preset threshold and the first similarity is less than the second preset threshold, determining that the object in the query image and the object in the result image are the same or similar objects.
5. The method of claim 4, further comprising:
if the second similarity is smaller than the seventh preset threshold, further judging whether the second similarity is larger than an eighth preset threshold, and whether the first similarity is smaller than a third preset threshold, wherein the seventh preset threshold is larger than the eighth preset threshold, and the second preset threshold is larger than the third preset threshold;
and if the second similarity is greater than the eighth preset threshold and the first similarity is less than the third preset threshold, determining that the object in the query image and the object in the result image are the same or similar objects.
6. The method of claim 5, further comprising:
if the second similarity is smaller than the eighth preset threshold, further judging whether the first similarity is smaller than a fourth preset threshold, wherein the third preset threshold is larger than the fourth preset threshold;
if the first similarity is smaller than the fourth preset threshold, determining that the object in the query image and the object in the result image are the same or similar objects;
otherwise, determining that the object in the query image and the object in the result image are non-identical or non-similar objects.
7. An object similarity determination device based on image recognition, comprising:
the receiving module is used for receiving an input query image;
the retrieval module is used for retrieving and obtaining a result image according to the query image;
the acquisition module is used for acquiring a first similarity and a second similarity between the query image and the result image, wherein the first similarity comprises the image overall feature similarity, and the second similarity comprises the image local feature similarity;
a determination module for determining whether the object in the query image and the object in the result image are the same or similar object according to the first similarity and the second similarity, wherein, whether the second similarity is larger than a fifth preset threshold value or not is judged, if the second similarity is larger than the fifth preset threshold value, then the object in the query image is determined to be the same or a similar object as the object in the result image, if the second similarity is smaller than the fifth preset threshold, further judging whether the second similarity is larger than a sixth preset threshold, and whether the first similarity is smaller than a first preset threshold, the fifth preset threshold is larger than the sixth preset threshold, if the second similarity is greater than the sixth preset threshold and the first similarity is less than the first preset threshold, then it is determined that the object in the query image is the same or similar object as the object in the result image.
8. The apparatus of claim 7, wherein the acquisition module comprises:
a first extraction unit, configured to extract a first image feature and a second image feature of the query image;
a second extraction unit configured to extract a third image feature and a fourth image feature of the result image;
a first obtaining unit, configured to obtain the first similarity according to the first image feature and the third image feature, where the first image feature and the third image feature are image overall features extracted based on a convolutional neural network CNN;
and the second obtaining unit is used for obtaining the second similarity according to the second image feature and the fourth image feature, wherein the second image feature and the fourth image feature are image local features extracted based on an SIFT algorithm.
9. The apparatus of claim 8, wherein the first obtaining unit is to:
performing linear transformation on the first image characteristic and the third image characteristic based on a linear transformation matrix, wherein the linear transformation matrix is obtained according to image sample statistics of the same or similar objects;
and calculating the distance between the first image characteristic after linear transformation and the third image characteristic after linear transformation to obtain the first similarity.
10. The apparatus of claim 7, wherein the determination module is further configured to:
if the second similarity is smaller than the sixth preset threshold, further judging whether the second similarity is larger than a seventh preset threshold, and whether the first similarity is smaller than a second preset threshold, wherein the sixth preset threshold is larger than the seventh preset threshold, and the first preset threshold is larger than the second preset threshold;
and if the second similarity is greater than the seventh preset threshold and the first similarity is less than the second preset threshold, determining that the object in the query image and the object in the result image are the same or similar objects.
11. The apparatus of claim 10, wherein the determination module is further configured to:
if the second similarity is smaller than the seventh preset threshold, further judging whether the second similarity is larger than an eighth preset threshold, and whether the first similarity is smaller than a third preset threshold, wherein the seventh preset threshold is larger than the eighth preset threshold, and the second preset threshold is larger than the third preset threshold;
and if the second similarity is greater than the eighth preset threshold and the first similarity is less than the third preset threshold, determining that the object in the query image and the object in the result image are the same or similar objects.
12. The apparatus of claim 11, wherein the determination module is further configured to:
if the second similarity is smaller than the eighth preset threshold, further judging whether the first similarity is smaller than a fourth preset threshold, wherein the third preset threshold is larger than the fourth preset threshold;
if the first similarity is smaller than the fourth preset threshold, determining that the object in the query image and the object in the result image are the same or similar objects;
otherwise, determining that the object in the query image and the object in the result image are non-identical or non-similar objects.
13. An information push system based on image recognition comprises:
the client is used for receiving a query image input by a user and sending the query image to the server;
the server side comprises the object similarity determination device based on image recognition according to any one of claims 7 to 12;
the server is further configured to, after determining that the object in the query image and the object in the result image are the same or similar, push the relevant information of the object in the result image to the client, so that the client displays the relevant information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610534932.3A CN107590154B (en) | 2016-07-08 | 2016-07-08 | Object similarity determination method and device based on image recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610534932.3A CN107590154B (en) | 2016-07-08 | 2016-07-08 | Object similarity determination method and device based on image recognition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107590154A CN107590154A (en) | 2018-01-16 |
| CN107590154B true CN107590154B (en) | 2021-03-02 |
Family
ID=61046015
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610534932.3A Active CN107590154B (en) | 2016-07-08 | 2016-07-08 | Object similarity determination method and device based on image recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107590154B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108563675B (en) * | 2018-02-28 | 2021-03-30 | 北京图铭视界科技有限公司 | Electronic file automatic generation method and device based on target body characteristics |
| CN108416902B (en) * | 2018-02-28 | 2021-11-26 | 成都好享你网络科技有限公司 | Real-time object identification method and device based on difference identification |
| CN109241316B (en) * | 2018-08-30 | 2020-10-16 | 北京旷视科技有限公司 | Image retrieval method, image retrieval device, electronic equipment and storage medium |
| CN110222662B (en) * | 2019-06-12 | 2020-11-03 | 北京博雅英杰科技股份有限公司 | Object identification method and device and storage medium |
| CN110532413B (en) * | 2019-07-22 | 2023-08-08 | 平安科技（深圳）有限公司 | Information retrieval method and device based on picture matching and computer equipment |
| CN111783808A (en) * | 2019-07-23 | 2020-10-16 | 北京沃东天骏信息技术有限公司 | Method and apparatus for generating information |
| CN111339884B (en) * | 2020-02-19 | 2023-06-06 | 浙江大华技术股份有限公司 | Image recognition method, related device and apparatus |
| CN111932323B (en) * | 2020-09-29 | 2021-01-15 | 北京每日优鲜电子商务有限公司 | Article information interface display method, device, equipment and computer readable medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102054177A (en) * | 2010-12-29 | 2011-05-11 | 北京新媒传信科技有限公司 | Image similarity calculation method and device |
| CN103413120A (en) * | 2013-07-25 | 2013-11-27 | 华南农业大学 | Tracking method based on integral and partial recognition of object |
| CN103413144A (en) * | 2013-07-29 | 2013-11-27 | 西北工业大学 | Airport detection and recognition method based on local global feature joint decision |
| CN104111960A (en) * | 2013-04-22 | 2014-10-22 | 阿里巴巴集团控股有限公司 | Page matching method and device |
| CN104462502A (en) * | 2014-12-19 | 2015-03-25 | 中国科学院深圳先进技术研究院 | Image retrieval method based on feature fusion |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20140035712A (en) * | 2012-09-14 | 2014-03-24 | 한국전자통신연구원 | Method and apparatus of object recognition |
-
2016
- 2016-07-08 CN CN201610534932.3A patent/CN107590154B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102054177A (en) * | 2010-12-29 | 2011-05-11 | 北京新媒传信科技有限公司 | Image similarity calculation method and device |
| CN104111960A (en) * | 2013-04-22 | 2014-10-22 | 阿里巴巴集团控股有限公司 | Page matching method and device |
| CN103413120A (en) * | 2013-07-25 | 2013-11-27 | 华南农业大学 | Tracking method based on integral and partial recognition of object |
| CN103413144A (en) * | 2013-07-29 | 2013-11-27 | 西北工业大学 | Airport detection and recognition method based on local global feature joint decision |
| CN104462502A (en) * | 2014-12-19 | 2015-03-25 | 中国科学院深圳先进技术研究院 | Image retrieval method based on feature fusion |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107590154A (en) | 2018-01-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107590154B (en) | Object similarity determination method and device based on image recognition | |
| US10902055B2 (en) | System and method of identifying visual objects | |
| US10672052B2 (en) | Method, apparatus, system, and computer program for offering and displaying product information | |
| KR101856120B1 (en) | Discovery of merchants from images | |
| Tonioni et al. | A deep learning pipeline for product recognition on store shelves | |
| US8775401B2 (en) | Shape based picture search | |
| KR101826815B1 (en) | Systems and methods for image-feature-based recognition | |
| Girod et al. | Mobile visual search: Architectures, technologies, and the emerging MPEG standard | |
| US20140254942A1 (en) | Systems and methods for obtaining information based on an image | |
| US20110106656A1 (en) | Image-based searching apparatus and method | |
| US20130202213A1 (en) | Method and system for fast and robust identification of specific product images | |
| US20150161147A1 (en) | Associating still images and videos | |
| CN105574089B (en) | Knowledge graph generation method and device, and object comparison method and device | |
| CN105095435A (en) | Similarity comparison method and device for high-dimensional image features | |
| CN104765891A (en) | Searching shopping method based on pictures | |
| CN101930444A (en) | Image search system and method | |
| WO2016033676A1 (en) | System and method for analyzing and searching imagery | |
| WO2012141655A1 (en) | In-video product annotation with web information mining | |
| US20210073890A1 (en) | Catalog-based image recommendations | |
| CN105792010A (en) | Television shopping method and device based on image content analysis and picture index | |
| CN101950400A (en) | Network shopping guiding method | |
| CN108492160A (en) | Information recommendation method and device | |
| WO2010071617A1 (en) | Method and apparatus for performing image processing | |
| CN114821234A (en) | Network training and target detection method, device, equipment and storage medium | |
| Nguyen et al. | Video instance search via spatial fusion of visual words and object proposals |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |