Community's image recommendation method of the image multi-dimensional degree marking model based on scene modeTechnical field
The present invention relates to information service field, the society of particularly a kind of image multi-dimensional degree marking model based on scene modeArea's image recommendation method.
Background technology
Community users can obtain more comprehensively various information by community information recommendation function, and the information simultaneously obtaining is fullThe individual demand of sufficient user.Community information content recommendation mainly includes two aspects：User may view data interested andThe scene information that image is associated, for example, recommend the view data of parking condition theme, obtains each simultaneously and recommends image to correspond toScene information, including time, place, subject content etc..The purpose of this functions of modules has two：One is to improve community information clothesThe function of business and mode, can meet the individual demand of different user, and method of service is issued based on conversion by passive informationMove and provide the user diversified information, make user obtain more fully community information；Two is the valency improving community's view dataValue and utilization rate, resident strengthens the wish participating in community service in acquisition information, by contribution plot picture of taking pictures, forms community's clothesBusiness good interaction.
Chinese invention patent CN 105303501A discloses a kind of Community Information Service method based on picture recommendation, bagInclude following steps：
(1) obtain time, place and the subject key words of user's inquiry input, determine user's concern location information；
(2) place, the picture of match query correlation and information are paid close attention to according to object time and user；
(3) location information is paid close attention to according to user and determine related social region, inquiry obtains the picture in this community areaInformation, is pushed to user as recommending picture；
(4) pass through User Interface, inquiry and recommendation results are showed user.
Content of the invention
The technical problem to be solved in the invention is to provide a kind of simple and reliable, practical figure based on scene modeCommunity's image recommendation method as various dimensions marking model.
For solving above-mentioned technical problem, community's figure of the image multi-dimensional degree marking model based on scene mode of the present inventionAs recommending method, comprise the following steps,
S101：Rely on internet crawler technology and mobile phone A PP to upload sharing function, obtain community's view data, be saved inIn local server；
S102：The data obtaining is pre-processed, carries out semantic analysis, by image, semantic displaying simultaneously；
S103：On the basis of step S102, form community's image multi-dimensional degree marking model；
S104：Based on community's image multi-dimensional degree marking model, using VSM method, Similarity Measure is carried out to image,Obtain similar image list；
S105：In conjunction with actual scene, carry out the sort recommendations of image using image similarity list.
Further, in described step S101 for the view data getting and corresponding text data respectively with imageFile format and excel form (data-base recording form) are saved in local server database.
Further, in described step S102, view data and corresponding text data are carried out, and in combination withActual conditions carry out semantic analysis, by image, semantic displaying.
Further, in described step S103 on the basis of image, semantic displaying, extract the specific master selecting imageTopic label, forms image multi-dimensional degree markup information.
Further, described step S104 is based on image multi-dimensional degree markup information, special to image labeling information structuringLevy space, form characteristic vector, the final similarity calculating between image, images above Similarity Measure can be image recommendationReliable foundation is provided to support.
Further, image data tag information is carried out classification mapping, and using label number in classification as featureItem word frequency, forms feature space；Calculate the weight of each characteristic item using TF-IDF model, and structural feature vector.
Further, vector similarity is often calculated using vectorial cosine angle.
Further, through the analysis of the scene information to user's search, obtain user's scene interested, according to similarDegree computational methods obtain the degree of correlation that user pays close attention to information and other information, then press relevancy ranking and recommend image.
After said method, the present invention carries out scenario building to the community's image data format obtaining, in scenario buildingOn the basis of, set up the community's view data various dimensions marking model under scene mode, afterwards community's image is entered using VSM methodRow Similarity Measure, is ranked up recommending in conjunction with scenario-related information and similarity.
Brief description
The present invention is further detailed explanation with reference to the accompanying drawings and detailed description.
Fig. 1 is that the flow process of community's image recommendation method based on the image multi-dimensional degree marking model of scene mode for the present invention is shownIt is intended to.
Specific embodiment
As shown in figure 1, community's image recommendation method of the image multi-dimensional degree marking model based on scene mode of the present invention,Comprise the following steps,
S101：Rely on internet crawler technology and mobile phone A PP to upload sharing function, obtain community's view data, be saved inIn local server；
S102：The data obtaining is pre-processed, carries out semantic analysis, by image, semantic displaying simultaneously；
S103：On the basis of step S102, form community's image multi-dimensional degree marking model；
S104：Based on community's image multi-dimensional degree marking model, using VSM method, Similarity Measure is carried out to image,Obtain similar image list；
S105：In conjunction with actual scene, carry out the sort recommendations of image using image similarity list.
Further, in described step S101 for the view data getting and corresponding text data respectively with imageFile format and excel form (data-base recording form) are saved in local server database.
Further, in described step S102, view data and corresponding text data are carried out, and in combination withActual conditions carry out semantic analysis, by image, semantic displaying.
Further, in described step S103 on the basis of image, semantic displaying, extract the specific master selecting imageTopic label, forms image multi-dimensional degree markup information.
Further, described step S104 is based on image multi-dimensional degree markup information, special to image labeling information structuringLevy space, form characteristic vector, the final similarity calculating between image, images above Similarity Measure can be image recommendationReliable foundation is provided to support.
Further, image data tag information is carried out classification mapping, and using label number in classification as featureItem word frequency, forms feature space；Calculate the weight of each characteristic item using TF-IDF model, and structural feature vector.
Further, vector similarity is often calculated using vectorial cosine angle.
Further, through the analysis of the scene information to user's search, obtain user's scene interested, according to similarDegree computational methods obtain the degree of correlation that user pays close attention to information and other information, then press relevancy ranking and recommend image.
Although the foregoing describing the specific embodiment of the present invention, those skilled in the art should be appreciated that thisIt is merely illustrative of, various changes or modifications can be made to present embodiment, the principle without departing from the present invention and essence,Protection scope of the present invention is only limited by the claims that follow.