CN108492160A - Information recommendation method and device - Google Patents
Information recommendation method and device Download PDFInfo
- Publication number
- CN108492160A CN108492160A CN201810196396.XA CN201810196396A CN108492160A CN 108492160 A CN108492160 A CN 108492160A CN 201810196396 A CN201810196396 A CN 201810196396A CN 108492160 A CN108492160 A CN 108492160A
- Authority
- CN
- China
- Prior art keywords
- characteristic information
- commodity
- end article
- information
- order data
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0623—Item investigation
- G06Q30/0625—Directed, with specific intent or strategy
- G06Q30/0629—Directed, with specific intent or strategy for generating comparisons
Abstract
A kind of information recommendation method of offer of the embodiment of the present invention and device, are related to computer application technology.Technical solution provided in an embodiment of the present invention is when user is unfamiliar with the commodity in Commdity advertisement or picture, the key scanning function that can be carried by application program in user equipment triggers the request that a key scanning obtains merchandise news, and user equipment can trigger camera scanning and obtain commodity picture;Product features information in picture processing extraction picture is carried out to commodity picture;According to the product features information, recommend the merchandise news with the product features information matches.So as to realize the recommendation information for quickly and accurately obtaining strange commodity, the Experience Degree of user is substantially increased.
Description
Technical field
The present embodiments relate to computer application technology more particularly to a kind of information recommendation methods and device.
Background technology
In the case where taking out environment, the prior art is all that user inputs keyword search related content, Huo Zheying by input frame
Businessman's commercial product recommending can be carried out with itself according to commercial circle and user preference.
When user sees some advertisement or sees some commodity picture, especially he is to the uncomprehending situation of the commodity
Under, he needs to read over advertisement or query-related information, then inputs keyword search dependent merchandise letter by input frame
Breath very takes consumption energy, and user experience is poor, and user can cause the dimension of input to have not knowing about for commodity
Limit, it is inaccurate so as to cause recommendation results.
In addition, carry out businessman's vegetable recommendation according to commercial circle and user preference, rely heavily on user the amount of placing an order and
The activity of the user, the content recommended in this way are just very limited.Recommendation is also resulted in if the activity of the user is very low
As a result inaccurate.
Therefore, there is an urgent need to the recommendation informations that a kind of method can quickly and accurately obtain strange commodity.
Invention content
A kind of information recommendation method of offer of the embodiment of the present invention and device, can quickly and accurately obtain pushing away for strange commodity
Recommend information.
In a first aspect, a kind of information recommendation method is provided in the embodiment of the present invention, including:
When detecting key scanning request, scanning obtains commodity picture;
The characteristic information of the end article in picture processing extraction picture is carried out to commodity picture;
According to the characteristic information of the end article, recommend to believe with the matched commodity of the characteristic information of the end article
Breath.
Optionally, according to the characteristic information of the end article, recommend matched with the characteristic information of the end article
Merchandise news, including:
The characteristic information of the end article is compared with the characteristic information of the sample commodity in merchandising database;
Recommended if comparing and unanimously obtaining the corresponding merchandise news of characteristic information for comparing consistent sample commodity;
If comparison is inconsistent, according to the characteristic information of the end article, in merchandising database, determine and the mesh
Mark the associated sample merchandise news of characteristic information of commodity.
Optionally, it according to the characteristic information of the end article, in merchandising database, determines and the end article
The associated sample merchandise news of characteristic information, including:
According to the characteristic information of the end article, in merchandising database, determines and believe with the feature of the end article
The associated product features information that similarity is more than first threshold is ceased, is determined according to the associated product features information corresponding
Associated articles information.
Optionally, the characteristic information of the sample commodity in the characteristic information and merchandising database of the end article is carried out
After comparison, further include:
If comparing inconsistent, it is determined that the similar users of active user, according to the History Order data and phase of active user
Like the History Order data of user, determined in the associated sample merchandise news with the History Order data of active user and
The maximum merchandise news of History Order data similarity of similar users is recommended.
Optionally it is determined that the similar users of active user, including:
According to the History Order data of active user, determine that the History Order data similarity with active user is more than the
The corresponding user of History Order data of the History Order data of two threshold values, the second threshold that similarity is more than uses as current
The similar users at family.
Second aspect, an embodiment of the present invention provides a kind of information recommending apparatus, including
Acquisition module, when for detecting key scanning request, scanning obtains commodity picture;
Characteristic extracting module, the feature for carrying out the end article in picture processing extraction picture to commodity picture are believed
Breath;
Recommending module recommends the characteristic information with the end article for the characteristic information according to the end article
Matched merchandise news.
Optionally, the recommending module specifically includes:
Comparing unit, for believing the feature of the sample commodity in the characteristic information and merchandising database of the end article
Breath is compared;
First recommendation unit, the characteristic information for when comparing consistent, obtaining the consistent sample commodity of comparison are corresponding
Merchandise news is recommended;
Second recommendation unit is used for when comparing inconsistent, according to the characteristic information of the end article, in commodity data
In library, the associated sample merchandise news of characteristic information with the end article is determined.
Optionally, first recommendation unit is used for:
According to the characteristic information of the end article, in merchandising database, determines and believe with the feature of the end article
The associated product features information that similarity is more than first threshold is ceased, is determined according to the associated product features information corresponding
Associated articles information.
Optionally, the recommending module further includes:
Third recommendation unit, for when comparing inconsistent, the similar users of active user being determined, according to active user's
The History Order data of History Order data and similar users, the determining and active user in the associated sample merchandise news
History Order data and the maximum merchandise news of History Order data similarity of similar users recommended.
Optionally, the recommending module further includes:
Determination unit determines the History Order data with active user for the History Order data according to active user
The History Order data of the History Order data for the second threshold that similarity is more than, the second threshold that similarity is more than are corresponding
Similar users of the user as active user.
The function can also execute corresponding software realization by hardware realization by hardware.The hardware or
Software includes one or more modules corresponding with above-mentioned function.
The third aspect, the embodiment of the present invention also provide a kind of computer storage media, the computer storage media storage
There are one or more computer instruction, the computer instruction to be performed the method realized as described in relation to the first aspect.
Fourth aspect, the embodiment of the present invention also provide a kind of mobile terminal, including memory and processor;Wherein,
The memory is for storing one or more computer instruction, wherein one or more computer instruction
It calls and executes for the processor;
The processor realizes method as described in relation to the first aspect when executing the computer instruction.
The technological means provided through the embodiment of the present invention, when user is unfamiliar with the commodity in Commdity advertisement or picture,
The key scanning function that can be carried by application program in user equipment triggers the request that a key scanning obtains merchandise news, uses
Family equipment can trigger camera scanning and obtain commodity picture;Product features in picture processing extraction picture are carried out to commodity picture
Information;According to the product features information, recommend the merchandise news with the product features information matches.It is fast so as to realize
Speed accurately obtains the recommendation information of strange commodity, substantially increases the Experience Degree of user.
The aspects of the invention or other aspects can more straightforwards in the following description.
Description of the drawings
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Some bright embodiments for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 shows information recommendation method flow diagram according to an embodiment of the invention;
Fig. 2 shows information recommendation method flow diagrams in accordance with another embodiment of the present invention;
Fig. 3 shows the information recommendation method flow diagram according to another embodiment of the invention;
Fig. 4 shows the structural schematic diagram of information recommending apparatus according to an embodiment of the invention.
Specific implementation mode
In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction in the embodiment of the present invention
Attached drawing, technical scheme in the embodiment of the invention is clearly and completely described.
In some flows of description in description and claims of this specification and above-mentioned attached drawing, contain according to
Multiple operations that particular order occurs, but it should be clearly understood that these operations can not be what appears in this article suitable according to its
Sequence is executed or is executed parallel, and the serial number such as 101,102 etc. of operation is only used for distinguishing each different operation, serial number
It itself does not represent and any executes sequence.In addition, these flows may include more or fewer operations, and these operations can
To execute or execute parallel in order.It should be noted that the descriptions such as " first " herein, " second ", are for distinguishing not
Same message, equipment, module etc., does not represent sequencing, does not also limit " first " and " second " and be different type.
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation describes, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, the every other implementation that those skilled in the art are obtained without creative efforts
Example, shall fall within the protection scope of the present invention.
The present invention application scenarios be, for example,:When user sees a strange Commdity advertisement, or translate into a strange quotient
When product picture (or strange commodity video), in the prior art, it can only carry out keying in related merchandise news by application program
Keyword scans for obtaining the relevant information list of commodity, and this mode very expends the time of user, user experience compared with
Difference, it is crucial that, since user can cause the information dimension of input limited or even inaccurate not knowing about for commodity, to lead
Cause the inaccuracy of search recommendation results.
The present invention is based on above-mentioned technical problem, it is preset with merchandising database first, merchandising database is according to using the application
What the History Order data of all users of program were established, there are the commodity sign and commodity of each lower single commodity in merchandising database
Characteristic information;Next, embedded images scanning recognition technology, obtains the product features information in commodity picture in the application,
To which according to the product features information of acquisition, (matching) be compared with the product features information in merchandising database, if in the presence of
Consistent product features information is compared, illustrates the merchandise news there are the commodity in merchandising database, you can is consistent according to comparing
Product features information carry out merchandise news recommendation;It, can be according to product features information, in commodity if comparison is inconsistent
The highest product features information of product features information similarity with above-mentioned acquisition is determined in database, and then obtains similarity most
High merchandise news is recommended.Therefore, technical scheme of the present invention not only increases the approach that user obtains merchandise news, together
When merchandise news recommendation more accurately enrich, user experience greatly improves.
Fig. 1 shows information recommendation method flow diagram according to an embodiment of the invention, as shown in Figure 1, including:
When 101, detecting key scanning request, scanning obtains commodity picture;
It is illustrated so that application program is taken out by Baidu as an example, the embedded images scanning recognition skill in application program is taken out by Baidu
Art takes out the button that Application Program Interface is provided with triggering one key scanning request in Baidu, and user can click one by touch screen
The button of key scanning, you can start user equipment camera and acquisition commodity picture is scanned to commodity picture.
102, the characteristic information of the end article in picture processing extraction picture is carried out to commodity picture;
For example include that image characteristics extraction, uncorrelated image filtering, Large Scale Graphs are carried out to commodity picture when specific implementation
As processing such as feature training, multi-level image classification, training image acquisition and associated picture selections.
Wherein, image characteristics extraction:In internet, most of images be stored in a manner of bitmap jpeg, png,
In the picture formats such as gif.The features such as this image preserved with dot matrix way is had statement simple, facilitates compression.But
When digital picture being handled and analyzed using the method for computer vision, the image of this expression method tends not to directly
It uses, and needs to convert image to other method progress restatements closer to people to image cognition.This restatement
Process be exactly characteristics of image extraction.During feature extraction, can according to it is different need it is right from different angles
Image is stated, these statements can be light and shade, color, texture, point of interest of image etc..
During being applied to follow-up uncorrelated image filtering and image classification for the characteristics of image that will be extracted, image is special
Sign extraction will not only define the feature of image, while be also required to define the correlation between different images on some special characteristic
Property.The definition of this feature representation correlation, similarity calculation that can be between image on feature space lay the foundation.
Wherein, uncorrelated image filtering:Commodity image on internet is all uploaded by trade company and is marked, this by society
The mark that user's upload can be changed is constantly present the problem of not fully meeting with real image.The generation of this problem exists more
The reason of aspect, as net purchase platform merchandise classification is unsound, between uploader and viewer semantic gap and uploader to searching
Index the excessive optimization etc. held up.If there are incorrect matching between a large amount of labels and image in training data, what training generated divides
Class model will answer noise excessive without meaning.Therefore, in the commodity image and correlation that will be directly crawled from net purchase platform
Mark needs to do a cleaning work to incoherent label in commodity image as before training data.There to be bigger related
Property data and its mark, remained as training data.This work from another perspective, that is, is filtered under same label
With the incoherent image of label.
Wherein, characteristics of image is trained:According to BOW disaggregated models currently popular, image finally needs to be expressed as word packet
Form.The frequency that word packet itself is then occurred by each visual word in image is formed.And visual word is then derived from visual dictionary,
Be trained by training sample caused by.In the application of net purchase platform commodity image class prediction, due in every piece image
Hundreds of points of interest unrelated with scale, size, rotation can be extracted, therefore, compare the quantity of image, visual interest point
Quantity is more surprising.And these visual interest points are trained to visual dictionary, then need the cluster for supporting large-scale data
Algorithm is realized.Specifically, in the present invention, having chosen compared to other cluster higher K-means algorithms of operational efficiency as base
Plinth, and advanced optimized on K-means algorithms, to realize the training of large-scale image feature, final realization image
Visual word packet is expressed.
Wherein, multi-level image is classified:Feature of the commodity image in net purchase platform is other than enormous amount, classification
It is especially more.Common sorting technique is often absorbed in the classification problem for solving two classes or a small amount of classification.And in commodity image class
Not in prediction task, directly using these disaggregated models often will produce classifying quality drastically decline and time complexity increase rapidly
Long problem.For example, the preferable method of some of opposite classifying qualities, can make disaggregated model with the growth of categorical measure
Training time and using disaggregated model predict new samples time at square grade increase.Not only amount of images is huge, classification
Quantity is also not applicable in huge commodity image class prediction.In net purchase platform, the classification of commodity is always with level knot
Structure is presented, and using this artificially defined hierarchical structure, can hierarchically be carried out the assorting process of commodity image.So not
The speed that training can only be accelerated and predicted, if for the different model of different classes of commodity training, additionally it is possible to promote quotient
The accuracy rate of product prediction.Meanwhile the disaggregated model training method of this stratification, be also easier to keep train classification models when just
The balance of negative sample.
Wherein, training image obtains and associated picture selects:Since method used in the present invention needs on net purchase platform
Commodity image and its markup information data supported, so needing to crawl the training image of magnanimity to net purchase platform.However, being
Effectively utilize the commodity image data on net purchase platform, using science method on net purchase platform commodity image and its
Mark sample most important.This is the groundwork that training image obtains.On the other hand, pre- by commodity image classification
After examining system is to the class prediction of commodity image, relevant commodity image, which is directly returned to user, can greatly promote use
The experience that family uses platform.
Product features information is extracted to commodity picture processing by above-mentioned, for example, when scanning gets blue fortune
When the image of dynamic shoes, it is, for example, the information such as plate shoes, canvas shoe, sport footwear, playshoes or sneakers to extract product features information.
When scanning obtains the image of a width white bicycle, extract product features information be, for example, mountain bike, it is common voluntarily
The information such as vehicle, road cycling, touring bicycle or bicycle equipment.When scanning obtains the image of a width pink jacket,
It is, for example, the information such as chiffon shirt, sweater, one-piece dress, loose T and knitting cardigan to extract product features information.
103, according to the characteristic information of the end article, recommend the matched commodity of characteristic information with the end article
Information.
In the embodiment of the present invention, need to build merchandising database, merchandising database includes all commodity of restocking, waits for
Picture, mark and the characteristic information of frame commodity.
Details are as follows for the construction step of merchandising database:
According to the profile diagram structural configuration feature in the three of commodity threedimensional model principal directions, first, principal component point is used
(Principal components analysis, the PCA) method of analysis and scale transformation standardize commodity threedimensional model to unit
In cube, then by commodity threedimensional model parallel projection to three principal planes, three profiles are obtained.Each profile is carried out
Sampling, i.e., from the center of profile diagram to the angular distances divergent-ray such as profile, using the distance at center to profile as sampled value, and to adopting
Sample value carries out SIFT transformation, obtains commodity contour feature.Wherein, the PCA image pre-processing methods of use have translation, rotation and
The advantages of scaling invariance, is the common technological means of image processing field, scale transformation method can be nearest field interpolation,
Bicubic interpolation and bilinear interpolation etc., are specifically no longer described in detail.
Characteristic information training is carried out to commodity contour feature, obtains multiple characteristic informations of each commodity.Merchandising database
In can according to the characteristic of commodity add or reduce characteristic information, such as certain famous brand skirt characteristic information can be certain brand,
Clothes, skirt, shortage of money and black and white etc..
Step 103 is specifically to believe the feature of the sample commodity in the characteristic information and merchandising database of the end article
Breath is compared, if comparing consistent, the consistent corresponding sample of sample product features information of acquisition comparison in merchandising database
The mark of this commodity recommends merchandise news corresponding with the mark of sample commodity.
In the embodiment of the present invention, embedded images scanning recognition technology, obtains the target in commodity picture in the application
Product features information, to according to end article characteristic information, be compared with the sample product features information in merchandising database
To (matching), consistent sample product features information is compared if existing, illustrates to exist in merchandising database identical as end article
Sample commodity, sample merchandise news is recommended.When concrete application, user only needs to utilize user equipment (such as hand
Machine) image scanning is carried out to end article picture, you can the information of the sample commodity similar with end article is received, is had splendid
User experience.
Fig. 2 shows information recommendation method flow diagrams in accordance with another embodiment of the present invention, as shown in Fig. 2, working as
When the sample product features information consistent with end article characteristic information being not present in merchandising database, the method includes:
When 201, detecting key scanning request, scanning obtains commodity picture;
It is illustrated so that application program is taken out by Baidu as an example, the embedded images scanning recognition skill in application program is taken out by Baidu
Art takes out the button that Application Program Interface is provided with triggering one key scanning request in Baidu, and user can click one by touch screen
The button of key scanning, you can start user equipment camera and acquisition commodity picture is scanned to commodity picture.
202, the characteristic information of the end article in picture processing extraction picture is carried out to commodity picture;
For example include that image characteristics extraction, uncorrelated image filtering, Large Scale Graphs are carried out to commodity picture when specific implementation
As processing such as feature training, multi-level image classification, training image acquisition and associated picture selections.
203, the characteristic information of end article is compared with the product features information in merchandising database.
204, it when comparing inconsistent, according to the characteristic information of end article, determines and recommends associated articles information.
Specifically, according to the characteristic information of end article, in merchandising database, the characteristic information with end article is determined
Similarity is more than the characteristic information (i.e. associated product features information) of the sample commodity of first threshold, or determines and target quotient
The characteristic information (i.e. associated product features information) of the maximum sample commodity of characteristic information similarity of product, according to associated quotient
Product characteristic information determines corresponding associated articles mark, determines and associated articles corresponding with associated articles mark is recommended to believe
Breath.
In addition, in the present embodiment when the similarity of determining end article and sample commodity, it can be in merchandising database
Each commodity contour feature carries out dictionary training, the corresponding training matrix of each commodity is obtained, by the corresponding training of each commodity
Matrix is stored in merchandising database.It is transversely arranged to the training matrix progress of all commodity in merchandising database to obtain owing set matrix;
End article picture is reconstructed according to preset reconstructed error value and deficient set matrix, it is corresponding heavy to obtain end article picture
Structure matrix;Each sample commodity are calculated according to the corresponding restructuring matrix of end article picture and the corresponding training matrix of sample commodity
Corresponding reconstructed residual value;According to reconstructed residual value, similar to end article or associated sample is determined in merchandising database
Commodity.Specifically, the reconstructed residual value of each sample commodity obtained is arranged according to ascending order, and the smaller expression of value is got over end article
It is similar.
In the embodiment of the present invention, embedded images scanning recognition technology, obtains the target in commodity picture in the application
Product features information, to according to end article characteristic information, be compared with the sample product features information in merchandising database
To (matching), if compare inconsistent, it can determine according to the characteristic information of end article and recommend associated articles information.Tool
Body in application, user only need using user equipment (such as mobile phone) to end article picture carry out image scanning, you can receive with
End article is similar or the information of associated sample commodity, has splendid user experience.
Fig. 3 shows the information recommendation method flow diagram according to another embodiment of the invention, as shown in figure 3, working as
When the sample product features information consistent with end article characteristic information being not present in merchandising database, the method further includes:
When 301, detecting key scanning request, scanning obtains commodity picture;
It is illustrated so that application program is taken out by Baidu as an example, the embedded images scanning recognition skill in application program is taken out by Baidu
Art takes out the button that Application Program Interface is provided with triggering one key scanning request in Baidu, and user can click one by touch screen
The button of key scanning, you can start user equipment camera and acquisition commodity picture is scanned to commodity picture.
302, the characteristic information of the end article in picture processing extraction picture is carried out to commodity picture;
For example include that image characteristics extraction, uncorrelated image filtering, Large Scale Graphs are carried out to commodity picture when specific implementation
As processing such as feature training, multi-level image classification, training image acquisition and associated picture selections.
303, the characteristic information of end article is compared with the product features information in merchandising database.
304, when comparing inconsistent, according to the characteristic information of end article, associated articles information is determined.
Specifically, according to the characteristic information of end article, in merchandising database, the characteristic information with end article is determined
Similarity is more than the characteristic information (i.e. associated product features information) of the sample commodity of first threshold, or determines and target quotient
The characteristic information (i.e. associated product features information) of the maximum sample commodity of characteristic information similarity of product, according to associated quotient
Product characteristic information determines corresponding associated articles mark, determines associated articles information corresponding with associated articles mark.
In addition, in the present embodiment when the similarity of determining end article and sample commodity, it can be in merchandising database
Each commodity contour feature carries out dictionary training, the corresponding training matrix of each commodity is obtained, by the corresponding training of each commodity
Matrix is stored in merchandising database.It is transversely arranged to the training matrix progress of all commodity in merchandising database to obtain owing set matrix;
End article picture is reconstructed according to preset reconstructed error value and deficient set matrix, it is corresponding heavy to obtain end article picture
Structure matrix;Each sample commodity are calculated according to the corresponding restructuring matrix of end article picture and the corresponding training matrix of sample commodity
Corresponding reconstructed residual value;According to reconstructed residual value, similar to end article or associated sample is determined in merchandising database
Commodity.Specifically, the reconstructed residual value of each sample commodity obtained is arranged according to ascending order, and the smaller expression of value is got over end article
It is similar.
305, the similar users for determining active user, according to the History Order data of active user and the history of similar users
Order data determines and the History Order data of active user and the History Order of similar users in the associated articles information
The maximum merchandise news of data similarity is recommended.
Specifically, the determination of the similar users in relation to active user is:According to the History Order data of active user, determine
Similarity is more than the second threshold by the History Order data for the second threshold being more than with the History Order data similarity of active user
Similar users of the corresponding user of History Order data of value as active user.
It should be noted that the determination in relation to second threshold not only considers matching dimensionality number, it is also contemplated that the weight of dimension point
Match, the frequency of History Order commodity dimension of the weight distribution from active user, the high dimension weight of the frequency is big.
It should be noted that associated articles itself have the information of multiple dimensions, and the commodity in active user's History Order
Also there is the information of multiple dimensions, similarly, the commodity also in the History Order of the similar users of active user also there are multiple dimensions
Information.
Accordingly, it in the determination of associated articles, such as by two parts of History Order merchandise newss progress comparing calculations, calculates
Coupling number of the proportion according to similar dimension, it is associated articles to first elect the most commodity of matching dimensionality;Secondly, it is ordered according to two parts of history
Single commodity dimensional information, it may be determined that the frequency that the dimensional information of some type or certain several type occurs is higher, this implementation
In example, it is believed that the dimension weight of the weight bigger of high frequency time dimension, low frequency time is relatively low.Therefore, when appearance matching dimension
Identical commodity are spent, it is associated articles to first elect the high commodity of weight dimension.
The embodiment of the present invention determines similar associated articles characteristic information according to the characteristic information of end article, certainly by user
The History Order data of body and the History Order data of similar users are combined, and are determined and user itself in associated articles information
History Order data and the maximum merchandise news of History Order data similarity of similar users (meet going through for user itself
The most commodity of the History Order data of history order data and similar users) and carry out top set recommendation.By combining associated articles
The History Order data of information and similar users can increase the calculating dimension for recommending platform, and then can also improve recommendation results
Precision, and the richness of recommendation results can be expanded by the associated articles information of similar users, to stimulate the viscosity of user
With the selection to place an order.
Fig. 4 shows the structural schematic diagram of information recommending apparatus according to an embodiment of the invention, as shown in figure 4, packet
It includes:
Acquisition module 41, when for detecting key scanning request, scanning obtains commodity picture;
Characteristic extracting module 42, the feature for carrying out the end article in picture processing extraction picture to commodity picture are believed
Breath;
Recommending module 43 recommends to believe with the feature of the end article for the characteristic information according to the end article
Cease matched merchandise news.
Optionally, the recommending module 43 specifically includes:
Comparing unit 431 is used for the spy of the sample commodity in the characteristic information and merchandising database of the end article
Reference breath is compared;
First recommendation unit 432, for when comparing consistent, obtaining the characteristic information correspondence for comparing consistent sample commodity
Merchandise news recommended;
Second recommendation unit 433 is used for when comparing inconsistent, according to the characteristic information of the end article, in commodity
In database, the associated sample merchandise news of characteristic information with the end article is determined.
Optionally, first recommendation unit 432 is used for:
According to the characteristic information of the end article, in merchandising database, determines and believe with the feature of the end article
The associated product features information that similarity is more than first threshold is ceased, is determined according to the associated product features information corresponding
Associated articles information.
Optionally, the recommending module 43 further includes:
Third recommendation unit 434 is used for when comparing inconsistent, determining the similar users of active user according to current
The History Order data at family and the History Order data of similar users determine and current in the associated sample merchandise news
The History Order data of user and the maximum merchandise news of History Order data similarity of similar users are recommended.
Optionally, the recommending module 43 further includes:
Determination unit 435 determines the History Order number with active user for the History Order data according to active user
According to the History Order data for the second threshold that similarity is more than, the History Order data for the second threshold that similarity is more than correspond to
Similar users of the user as active user.
Fig. 4 shown devices can execute the method in Fig. 1-embodiment illustrated in fig. 3, and implementing principle and technical effect are no longer
It repeats.
In a possible design, the structure of information recommending apparatus shown in Fig. 4 includes processor and memory, institute
Memory is stated for storing the program for supporting that information recommending apparatus executes information recommendation method in above-mentioned first aspect, the processing
Device is configurable for executing the program stored in the memory.
Described program includes one or more computer instruction, wherein described in one or more computer instruction supplies
Processor, which calls, to be executed.
The processor is used for:When detecting key scanning request, scanning obtains commodity picture;Figure is carried out to commodity picture
The characteristic information of end article in piece processing extraction picture;According to the characteristic information of the end article, recommend and the mesh
Mark the matched merchandise news of characteristic information of commodity.
The processor is additionally operable to:By the spy of the sample commodity in the characteristic information and merchandising database of the end article
Reference breath is compared;If compare it is consistent, obtain compare the corresponding merchandise news of characteristic information of consistent sample commodity into
Row is recommended;If comparison is inconsistent, according to the characteristic information of the end article, in merchandising database, determine and the mesh
Mark the associated sample merchandise news of characteristic information of commodity.
The processor is additionally operable to:According to the characteristic information of the end article, in merchandising database, determine with it is described
The characteristic information similarity of end article is more than the associated product features information of first threshold, special according to the associated commodity
Reference breath determines corresponding associated articles information.
The processor is additionally operable to:If comparing inconsistent, it is determined that the similar users of active user, according to active user's
The History Order data of History Order data and similar users, the determining and active user in the associated sample merchandise news
History Order data and the maximum merchandise news of History Order data similarity of similar users recommended.
The processor is additionally operable to:According to the History Order data of active user, the History Order with active user is determined
The History Order data for the second threshold that data similarity is more than, the History Order data pair for the second threshold that similarity is more than
Similar users of the user answered as active user.
An embodiment of the present invention provides a kind of computer storage medias, for storing the computer used in information recommending apparatus
Software instruction, it includes be the program involved by information recommending apparatus for executing above- mentioned information to recommend method.
The present invention discloses A1, a kind of information recommendation method, including:
When detecting key scanning request, scanning obtains commodity picture;
The characteristic information of the end article in picture processing extraction picture is carried out to commodity picture;
According to the characteristic information of the end article, recommend to believe with the matched commodity of the characteristic information of the end article
Breath.
In A2, the method as described in A1, according to the characteristic information of the end article, recommend the spy with the end article
Reference ceases matched merchandise news, including:
The characteristic information of the end article is compared with the characteristic information of the sample commodity in merchandising database;
Recommended if comparing and unanimously obtaining the corresponding merchandise news of characteristic information for comparing consistent sample commodity;
If comparison is inconsistent, according to the characteristic information of the end article, in merchandising database, determine and the mesh
Mark the associated sample merchandise news of characteristic information of commodity.
In A3, the method as described in A2, according to the characteristic information of the end article, in merchandising database, determine with
The associated sample merchandise news of characteristic information of the end article, including:
According to the characteristic information of the end article, in merchandising database, determines and believe with the feature of the end article
The associated product features information that similarity is more than first threshold is ceased, is determined according to the associated product features information corresponding
Associated articles information.
In A4, the method as described in seeking A2, by the sample quotient in the characteristic information and merchandising database of the end article
After the characteristic information of product is compared, further include:
If comparing inconsistent, it is determined that the similar users of active user, according to the History Order data and phase of active user
Like the History Order data of user, determined in the associated sample merchandise news with the History Order data of active user and
The maximum merchandise news of History Order data similarity of similar users is recommended.
A5, the method as described in A4 determine the similar users of active user, including:
According to the History Order data of active user, determine that the History Order data similarity with active user is more than the
The corresponding user of History Order data of the History Order data of two threshold values, the second threshold that similarity is more than uses as current
The similar users at family.
The invention also discloses B6, a kind of information recommending apparatus, including:
Acquisition module, when for detecting key scanning request, scanning obtains commodity picture;
Characteristic extracting module, the feature for carrying out the end article in picture processing extraction picture to commodity picture are believed
Breath;
Recommending module recommends the characteristic information with the end article for the characteristic information according to the end article
Matched merchandise news.
In B7, the device as described in B6, the recommending module specifically includes:
Comparing unit, for believing the feature of the sample commodity in the characteristic information and merchandising database of the end article
Breath is compared;
First recommendation unit, the characteristic information for when comparing consistent, obtaining the consistent sample commodity of comparison are corresponding
Merchandise news is recommended;
Second recommendation unit is used for when comparing inconsistent, according to the characteristic information of the end article, in commodity data
In library, the associated sample merchandise news of characteristic information with the end article is determined.
B8, the device as described in B7, first recommendation unit are used for:
According to the characteristic information of the end article, in merchandising database, determines and believe with the feature of the end article
The associated product features information that similarity is more than first threshold is ceased, is determined according to the associated product features information corresponding
Associated articles information.
In B9, the device as described in B7, the recommending module further includes:
Third recommendation unit, for when comparing inconsistent, the similar users of active user being determined, according to active user's
The History Order data of History Order data and similar users, the determining and active user in the associated sample merchandise news
History Order data and the maximum merchandise news of History Order data similarity of similar users recommended.
B10, the device as described in B9, the recommending module further include:
Determination unit determines the History Order data with active user for the History Order data according to active user
The History Order data of the History Order data for the second threshold that similarity is more than, the second threshold that similarity is more than are corresponding
Similar users of the user as active user.
The invention also discloses C11, a kind of computer storage media, the computer storage media is stored with one or more
Computer instruction, the computer instruction are performed the method realized as described in any one of claim A1-A5.
The invention also discloses D12, a kind of mobile terminal, including memory and processors;Wherein,
The memory is for storing one or more computer instruction, wherein one or more computer instruction
It calls and executes for the processor;
The processor realizes the method as described in any one of claim A1-A5 when executing the computer instruction.
It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description,
The specific work process of device and unit, can refer to corresponding processes in the foregoing method embodiment, and details are not described herein.
The apparatus embodiments described above are merely exemplary, wherein the unit illustrated as separating component can
It is physically separated with being or may not be, the component shown as unit may or may not be physics list
Member, you can be located at a place, or may be distributed over multiple network units.It can be selected according to the actual needs
In some or all of module achieve the purpose of the solution of this embodiment.Those of ordinary skill in the art are not paying creativeness
Labour in the case of, you can to understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
It is realized by the mode of software plus required general hardware platform, naturally it is also possible to pass through hardware.Based on this understanding, on
Stating technical solution, substantially the part that contributes to existing technology can be expressed in the form of software products in other words, should
Computer software product can store in a computer-readable storage medium, such as ROM/RAM, magnetic disc, CD, including several fingers
It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes each implementation
Method described in certain parts of example or embodiment.
Finally it should be noted that:The above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;Although
Present invention has been described in detail with reference to the aforementioned embodiments, it will be understood by those of ordinary skill in the art that:It still may be used
With technical scheme described in the above embodiments is modified or equivalent replacement of some of the technical features;
And these modifications or replacements, various embodiments of the present invention technical solution that it does not separate the essence of the corresponding technical solution spirit and
Range.
Claims (10)
1. a kind of information recommendation method, which is characterized in that including:
When detecting key scanning request, scanning obtains commodity picture;
The characteristic information of the end article in picture processing extraction picture is carried out to commodity picture;
According to the characteristic information of the end article, recommend the matched merchandise news of characteristic information with the end article.
2. according to the method described in claim 1, it is characterized in that, according to the characteristic information of the end article, recommend and institute
The matched merchandise news of characteristic information of end article is stated, including:
The characteristic information of the end article is compared with the characteristic information of the sample commodity in merchandising database;
Recommended if comparing and unanimously obtaining the corresponding merchandise news of characteristic information for comparing consistent sample commodity;
If comparison is inconsistent, according to the characteristic information of the end article, in merchandising database, determine and the target quotient
The associated sample merchandise news of characteristic information of product.
3. according to the method described in claim 2, it is characterized in that, according to the characteristic information of the end article, in commodity number
According in library, the associated sample merchandise news of characteristic information with the end article is determined, including:
According to the characteristic information of the end article, in merchandising database, the characteristic information phase with the end article is determined
Associated product features information like degree more than first threshold, corresponding association is determined according to the associated product features information
Merchandise news.
4. according to the method described in claim 2, it is characterized in that, by the characteristic information and merchandising database of the end article
In sample commodity characteristic information be compared after, further include:
If comparing inconsistent, it is determined that the similar users of active user, according to the History Order data of active user and similar use
The History Order data at family determine and the History Order data of active user and similar in the associated sample merchandise news
The maximum merchandise news of History Order data similarity of user is recommended.
5. according to the method described in claim 4, it is characterized in that, determine active user similar users, including:
According to the History Order data of active user, the second threshold that the History Order data similarity with active user is more than is determined
The History Order data of value, the corresponding user of History Order data for the second threshold that similarity is more than is as active user's
Similar users.
6. a kind of device of information recommendation, which is characterized in that including:
Acquisition module, when for detecting key scanning request, scanning obtains commodity picture;
Characteristic extracting module, the characteristic information for carrying out the end article in picture processing extraction picture to commodity picture;
Recommending module, for the characteristic information according to the end article, recommendation is matched with the characteristic information of the end article
Merchandise news.
7. device according to claim 6, which is characterized in that the recommending module specifically includes:
Comparing unit, for by the characteristic information of the sample commodity in the characteristic information and merchandising database of the end article into
Row compares;
First recommendation unit, for when comparing consistent, obtaining the corresponding commodity of characteristic information for comparing consistent sample commodity
Information is recommended;
Second recommendation unit is used for when comparing inconsistent, according to the characteristic information of the end article, in merchandising database
In, determine the associated sample merchandise news of characteristic information with the end article.
8. device according to claim 7, which is characterized in that first recommendation unit is used for:
According to the characteristic information of the end article, in merchandising database, the characteristic information phase with the end article is determined
Associated product features information like degree more than first threshold, corresponding association is determined according to the associated product features information
Merchandise news.
9. device according to claim 7, which is characterized in that the recommending module further includes:
Third recommendation unit, for when comparing inconsistent, the similar users of active user being determined, according to the history of active user
The History Order data of order data and similar users are determined in the associated sample merchandise news and are gone through with active user
The maximum merchandise news of History Order data similarity of history order data and similar users is recommended.
10. device according to claim 9, which is characterized in that the recommending module further includes:
Determination unit, for the History Order data according to active user, determination is similar to the History Order data of active user
Spend the History Order data for the second threshold being more than, the corresponding user of History Order data for the second threshold that similarity is more than
Similar users as active user.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2017103849409 | 2017-05-26 | ||
CN201710384940.9A CN107133854A (en) | 2017-05-26 | 2017-05-26 | Information recommendation method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108492160A true CN108492160A (en) | 2018-09-04 |
Family
ID=59734036
Family Applications (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710384940.9A Pending CN107133854A (en) | 2017-05-26 | 2017-05-26 | Information recommendation method and device |
CN201810196396.XA Pending CN108492160A (en) | 2017-05-26 | 2018-03-09 | Information recommendation method and device |
Family Applications Before (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710384940.9A Pending CN107133854A (en) | 2017-05-26 | 2017-05-26 | Information recommendation method and device |
Country Status (1)
Country | Link |
---|---|
CN (2) | CN107133854A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109885714A (en) * | 2019-01-18 | 2019-06-14 | 上海扩博智能技术有限公司 | Retail commodity information-pushing method, system, equipment and storage medium under line |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109598517B (en) * | 2017-09-29 | 2023-09-12 | 阿里巴巴集团控股有限公司 | Commodity clearance processing, object processing and category prediction method and device thereof |
CN108520058B (en) * | 2018-03-30 | 2021-09-24 | 维沃移动通信有限公司 | Merchant information recommendation method and mobile terminal |
CN110363555B (en) * | 2018-04-10 | 2024-04-09 | 释空(上海)品牌策划有限公司 | Recommendation method and device based on vision tracking visual algorithm |
CN109214453A (en) * | 2018-08-30 | 2019-01-15 | 上海扩博智能技术有限公司 | Automatic marking method, system, equipment and the storage medium of model training image |
CN109522431B (en) * | 2018-09-20 | 2022-11-04 | 上海博泰悦臻网络技术服务有限公司 | Hotel recommendation method and system based on picture recognition |
CN110889337A (en) * | 2019-11-08 | 2020-03-17 | 福建工程学院 | AR-based description method and AR-based description system |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103412938A (en) * | 2013-08-22 | 2013-11-27 | 成都数之联科技有限公司 | Commodity price comparing method based on picture interactive type multiple-target extraction |
CN103412937A (en) * | 2013-08-22 | 2013-11-27 | 成都数之联科技有限公司 | Searching and shopping method based on handheld terminal |
CN104765891A (en) * | 2015-05-06 | 2015-07-08 | 苏州搜客信息技术有限公司 | Searching shopping method based on pictures |
CN106021551A (en) * | 2016-07-01 | 2016-10-12 | 深圳追科技有限公司 | Consumption auxiliary decision making method based on screenshot information recognition |
-
2017
- 2017-05-26 CN CN201710384940.9A patent/CN107133854A/en active Pending
-
2018
- 2018-03-09 CN CN201810196396.XA patent/CN108492160A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103412938A (en) * | 2013-08-22 | 2013-11-27 | 成都数之联科技有限公司 | Commodity price comparing method based on picture interactive type multiple-target extraction |
CN103412937A (en) * | 2013-08-22 | 2013-11-27 | 成都数之联科技有限公司 | Searching and shopping method based on handheld terminal |
CN104765891A (en) * | 2015-05-06 | 2015-07-08 | 苏州搜客信息技术有限公司 | Searching shopping method based on pictures |
CN106021551A (en) * | 2016-07-01 | 2016-10-12 | 深圳追科技有限公司 | Consumption auxiliary decision making method based on screenshot information recognition |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109885714A (en) * | 2019-01-18 | 2019-06-14 | 上海扩博智能技术有限公司 | Retail commodity information-pushing method, system, equipment and storage medium under line |
Also Published As
Publication number | Publication date |
---|---|
CN107133854A (en) | 2017-09-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108492160A (en) | Information recommendation method and device | |
TWI623842B (en) | Image search and method and device for acquiring image text information | |
EP2805262B1 (en) | Image index generation based on similarities of image features | |
CN101937549B (en) | Picture guidance system for network shopping guidance | |
US11941049B2 (en) | Adaptive search results for multimedia search queries | |
CN101950400B (en) | Picture retrieving method of network shopping guiding method | |
CN109993102B (en) | Similar face retrieval method, device and storage medium | |
CN107683469A (en) | A kind of product classification method and device based on deep learning | |
EP3300002A1 (en) | Method for determining the similarity of digital images | |
CN110083729B (en) | Image searching method and system | |
Bui et al. | Scalable sketch-based image retrieval using color gradient features | |
JP2002288219A (en) | Information search system and method | |
Yang et al. | Prototype-based image search reranking | |
CN103631932A (en) | Method for detecting repeated video | |
CN109308324A (en) | A kind of image search method and system based on hand drawing style recommendation | |
CN107622071B (en) | Clothes image retrieval system and method under non-source-retrieval condition through indirect correlation feedback | |
CN110413825B (en) | Street-clapping recommendation system oriented to fashion electronic commerce | |
CN108717436B (en) | Commodity target rapid retrieval method based on significance detection | |
Yousaf et al. | Patch-CNN: Deep learning for logo detection and brand recognition | |
CN113657087A (en) | Information matching method and device | |
WO2024027347A9 (en) | Content recognition method and apparatus, device, storage medium, and computer program product | |
CN110737796B (en) | Image retrieval method, device, equipment and computer readable storage medium | |
CN115712780A (en) | Information pushing method and device based on cloud computing and big data | |
US11748796B2 (en) | Automatic clustering and mapping of user generated content with curated content | |
JP2016014990A (en) | Moving image search method, moving image search device, and program thereof |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180904 |