Summary of the invention
For overcome the deficiencies in the prior art, one of the objects of the present invention is to provide a kind of advertisements based on crowd characteristicMethod for pushing can be analyzed according to user characteristics combination crowd characteristic, provide more accurate product advertising.
An object of the present invention adopts the following technical scheme that realization：
A kind of advertisement sending method based on crowd characteristic, includes the following steps：Product classification binds step：Obtain userCrowd watches the video image that product advertising is greater than set period of time, extracts the user characteristics of user crowd, and the product is wideThe co-user feature of product classification and user crowd belonging to accusing is bound；User characteristics extraction step：Obtain user video figurePicture extracts the user characteristics of user in video image；Advertisement pushing step：Determination is corresponding with the user characteristics of the user totalThe product advertising of the product classification subordinate bound with common user characteristics is pushed to the user and watched by same user characteristics.
Further, in product classification binding step, the user characteristics of each user of user crowd are obtained,Extract co-user feature of several identical features as user crowd.
Further, when the quantity of the user characteristics of the user crowd of acquisition is more than setting value, then to all user spiesSign is grouped, and is determined co-user feature by group, then carry out intersection processing to the co-user feature of each group, is extracted severalCo-user feature of the identical feature as user crowd.
Further, user characteristics include age, gender, hair style, same to administrative staff, wearing garment ornament.Further, it usesFamily feature is by comparing the classification of each feature of determining user characteristics with user feature database.
Further, the user feature database includes age data library, gender data library, hair style database, colleagueDemographic data library, wearing dress ornament database.
Further, user characteristics are obtained by user's face information, user number, user profile.
Further, user characteristics are extracted to the video image of acquisition, if there are corresponding co-user spies for user characteristicsSign, then directly execute the advertisement pushing step；If corresponding co-user feature is not present in user characteristics, the production is executedProduct classification binding step.
The second object of the present invention is to provide a kind of electronic equipment, can be divided according to user characteristics combination crowd characteristicAnalysis, provides more accurate product advertising.
An object of the present invention adopts the following technical scheme that realization：
A kind of electronic equipment can be run on a memory and on a processor including memory, processor and storageComputer program, the processor realize that one kind as described in one of the object of the invention is based on crowd characteristic when executing described programAdvertisement sending method.
The third object of the present invention is to provide a kind of storage medium, can be divided according to user characteristics combination crowd characteristicAnalysis, provides more accurate product advertising.
The third object of the present invention adopts the following technical scheme that realization：
A kind of computer readable storage medium, is stored thereon with computer program, it is characterised in that：The computer programA kind of advertisement sending method based on crowd characteristic as described in one of the object of the invention is realized when being executed by processor.
Compared with prior art, the beneficial effects of the present invention are：
A kind of advertisement sending method based on crowd characteristic, electronic equipment and storage medium of the invention, by by userProduct classification belonging to the product advertising that the co-user feature of crowd is watched with user crowd is bound, common further according to thisUser characteristics can be to there are the product classification subordinates of the user of corresponding user characteristics push and co-user feature bindingProduct advertising, and then realize product advertising more accurately push, effectively improve the specific aim of advertisement pushing.
Embodiment one provides a kind of advertisement sending method based on crowd characteristic, includes the following steps as shown in Figure 1：
S1：The user crowd of acquisition watches the video image that product advertising is greater than set period of time, extracts the use of user crowdFamily feature binds the co-user feature of product classification belonging to the product advertising and user crowd；
S2：User video image is obtained, the user characteristics of user in video image are extracted；
S3：Determine co-user feature corresponding with the user characteristics of the user, the production that will be bound with common user characteristicsThe product advertising of product classification subordinate is pushed to user viewing.
Online in lower exhibition and sale of products area, the user video image that camera is collected is obtained first, passes through user video imageUser's characteristic information is extracted, if there are corresponding co-user features for user characteristics at this time, directly executes advertisement pushing stepSuddenly, corresponding product advertising is shown to user viewing by display screen；If corresponding co-user is not present in user characteristicsFeature then executes product classification binding step.Product classification is bound in step, watches the product that display screen is shown attentively to eachAdvertisement is more than the user of preset period, and camera can all shoot it, and the user by obtaining camera schemesPicture extracts user characteristics to user images, then extracts in several identical features existing for each user characteristics of user crowdHold the co-user feature as user crowd.It should be noted that the user characteristics of the user crowd extracted reach setting numberAfter mesh, then the extraction of co-user feature can be carried out, and by the common use of product classification belonging to product advertising and user crowdThe binding of family feature.The user characteristics extracted later execute if its feature has part identical with common user characteristicsS3 advertisement pushing step.In addition, the establishment of co-user feature, depends in the identical feature of each user characteristics of user crowdHold, meets the same characteristic features content number that co-user feature is established, customized can set, the present embodiment is full in featureIt determines this three at foot three or more and the above same characteristic features content is co-user feature.When the use of the user crowd of acquisitionThe quantity of family feature is more than setting value, then is grouped to all user characteristics, determines co-user feature by group, then to eachThe co-user feature of group carries out intersection processing, and the co-user for extracting several identical features as user crowd is specialSign.In order to obtain more accurate co-user feature, the user characteristics of certain amount are needed, if directly to multiple user characteristicsCommon trait extraction is carried out, it is possible that the case where common trait is not present, then will first be carried out to multiple user characteristicsCo-user feature is extracted in grouping, is carried out intersection processing to multiple co-user features again after co-user feature extraction, is obtainedFinal co-user feature.
The user characteristics of the present embodiment include age, gender, hair style, same to administrative staff, wearing garment ornament.User characteristics are logicalUser's face information, user number, user profile is crossed to obtain.User characteristics are by comparing determining user with user feature databaseThe classification of each feature of feature.The user feature database include age data library, gender data library, hair style database,Colleague demographic data library, wearing dress ornament database.Each Database mode is as follows：
1, the foundation in age data library.
By neural network algorithm, user is divided into " 15-25 years old, 25-35 years old, 35-45 years old, 45 years old or more " four agesSection, corresponds to all age group attribute for user and classifies.I.e. respectively obtain " 15-25 years old, 25-35 years old, 35-45 years old, 45 years old withOn " a large number of users mug shots of four age brackets, the user's face photo of acquisition is pre-processed, feature extraction, repeatedlyThe feature vector of extraction is corresponded to all age group section and stored, completes the user of aforementioned four age bracket by recognition trainingAge data library is established.
2, the foundation in gender data library.
By face recognition technology, user's face information is identified, to complete the foundation of user's gender data library.
3, the foundation of hair style database.
It is pre-processed by the photo of a large amount of each hair styles of acquisition, feature extraction, learning training, by the feature vector of extractionCorresponding different hair style type is stored, to complete user's hair style Database.
4, the foundation in colleague demographic data library.
By face recognition technology, user's face information is identified, to judge user characteristics and its shopping preferences.For example, push away perambulator user its buy preference it is mostly related with infant article.
5, the foundation of dress ornament database is dressed.
It is pre-processed, feature extraction, learning training, will be mentioned by obtaining the largely photo of each brand dress ornaments and handbagThe feature vector taken corresponds to different purchasing power sections and is stored, to complete user's purchasing power Database.
It, can age to user characteristics, gender, hair style, same to administrative staff, wearing clothes after the foundation for completing above-mentioned databaseFive features are adornd to be identified.The following are the identification methods of each feature：
1, the age identifies
The face-image for extracting crowd in sales exhibition sales field video, is pre-processed, feature extraction, by pretreated userFacial information is compared with age data library, to obtain the age bracket of video user.
2, gender identifies
The face-image for extracting crowd in sales exhibition sales field video, is pre-processed, feature extraction, passes through recognition of face skillArt pretreated user's face information is compared with gender data library, to obtain the gender of video user.
3, hair style identifies
The face-image for extracting crowd in sales exhibition sales field video, is pre-processed, feature extraction, by pretreated userFacial information is compared with hair style database, to obtain the hair style of video user.
4, colleague's personal identification
The face-image for extracting crowd in sales exhibition sales field video, is pre-processed, feature extraction, and obtaining for colleague's number is passed throughIt takes, to judge that user is to do shopping or hold in the arms alone baby and to push perambulator, carries old man's trip, lovers, more people go on a journeySituations such as, to show which same administrative staff has, label to user and user group, to show that the shopping of user is inclinedIt is good.
5, garments worn identifies
The wearing dress ornament for extracting crowd in sales exhibition sales field video is compared, to obtain use with wearing dress ornament databaseFamily dress ornament and the corresponding purchasing power section of handbag type, and purchasing power class mark is carried out to the user.
By comparing with user feature database, that is, it can determine classification belonging to each feature of user characteristics.
In the following, being exemplified by Table 1, the establishment of co-user feature is illustrated.As shown in table 1：
Table 1
Four are illustrated in table 1 and has determined that four user characteristics classified belonging to feature, pass through co-user featureThree or three or more same characteristic features contents are extracted, then the co-user feature with the binding of A product is：20-25 years old, female, bob.If the user characteristics obtained after so from camera, then can be corresponding to the user characteristics comprising features described above contentUser's advertisement watches advertising display on the display screen in sales exhibition area for the user.
It should be noted that there can be the case where user loses interest in this after product advertising is pushed to user, lead toThe lasting user video for obtaining and receiving the product advertising of a certain product classification is crossed, user is extracted and watches display screen attentively more than setting timeThe user base number of section (the present embodiment is 1.5 seconds), is considered as to the interested user base number of the product classification, further calculates to obtainInterested user's ratio, when interested user's ratio is lower than preset value, the product classification binding before showing is depositedIn error, need to re-start product classification binding step S1, at this time to ensure more accurately to user's advertisement.
A kind of advertisement sending method based on crowd characteristic of the present embodiment, by by the co-user feature of user crowdProduct classification belonging to product advertising with user crowd's viewing is bound, can be to presence further according to the co-user featureUser's push of corresponding user characteristics and the product advertising of the product classification subordinate of co-user feature binding, and then realizeProduct advertising more accurately pushes, and effectively improves the specific aim of advertisement pushing, while realizing and being pushed away based on crowd characteristic personalizationMarketing model under the line of product is recommended, the enthusiasm of customer consumption is increased, promotes the enthusiasm of purchase.