CN116342812A - 3D shopping scene construction method and device based on scene style and scene type - Google Patents

Abstract

The invention discloses a 3D shopping scene construction method based on scene styles and scene types, which comprises the following steps: constructing a single scene sequence according to the scene type and the stylization degree selected by the user and the scene style obtained by predicting based on the historical behavior data of the user; filling a background map and background sound effects on a single scene in a single scene sequence by utilizing a pre-constructed material database; acquiring account data of a user, wherein the account data comprises user basic information, and consumption habits and consumption capacities of the user; inputting account data into a pre-constructed prediction model to predict the commodity combination association logic ordering expected by a user; and screening out commodity combination information meeting the requirements from the material database according to the commodity combination association logic ordering obtained through prediction, and outputting a 3D shopping scene. The invention further provides a 3D shopping scene construction system. The method provided by the invention can effectively improve the cognitive efficiency of the user on the commodity and the shopping immersion of the user in the 3D scene.

Description

3D shopping scene construction method and device based on scene style and scene type

Technical Field

The invention belongs to the field of scene reconstruction, and particularly relates to a 3D shopping scene construction method and device based on scene styles and scene types.

Background

XR shopping experience: XR is a comprehensive application of three technologies, VR, AR and MR, and is a development direction of digital experience design in the future. Currently, XR technology has been employed in many business scenarios in ali. In the aspect of VR, the Taobao VR team developed VR shopping function "buy+" as early as 2016, combines the high efficiency of online shopping and the fun of interaction with commodities in offline shopping through VR technology, and preliminarily realizes scene shopping. The panning then also pushes out a 3D interactive business scenario at the mobile end, supporting consumers to view and interact with the high-reduction 3D commodity model in the scenario. In addition, the iHome platform is also provided for the home decoration field, and a virtual 3D space is created by the traditional plane design drawing, so that a consumer can quickly design own indoor space. In the AR aspect, the Ali provides a 2D makeup series and a 3D color series in the hand washing, and the functions of AR try shoes, AR try cosmetics and the like are included, so that the commodity display design shortens the distance between a consumer and a virtual commodity to a certain extent, and the shopping experience of the consumer is enriched.

Scene shopping: the method is different from the current AR shopping experience in that a 3D commodity model is placed at a specific position of a real scene, and the experience of restoring the real commodity super scene in VR shopping. The scene shopping aims at constructing a virtual scene which is matched with commodity semantics and usage scenes, reconstructing the organization form of multiple commodities in the virtual scene, and creating an XR immersive interactive shopping experience.

3D item detail page: the method breaks through the form of a commodity detail page in the traditional 2D image-text form, builds a 3D commodity model, and enables a user to interact with the 3D commodity model through interaction to present an interaction information presentation interface in the forms of model part disassembly, function animation presentation, hot spot element presentation and the like. Meanwhile, information such as configuration parameters affecting user decision in commodities needs to be designed into a form suitable for display in a 3D environment.

Patent CN106875244a discloses a virtual reality shopping method, device and electronic equipment, the method comprising: selecting a target commodity; determining at least one attribute information of the target commodity; selecting a use scene of the target commodity; matching the target commodity with the attribute information in the use scene, and displaying the use state scene of the target commodity in the use scene; the target commodity is added to the shopping cart and paid. However, the method has insufficient scene information and scene expression, and cannot reasonably provide commodities.

The academic literature Shopping in virtual reality: a study on consumers' shopping experience in a stereoscopic virtual reality discloses a VR shop concept in which the product is placed not on shelves but by space in place in an apartment environment. The article studies how the spatial arrangement of products in a non-retail environment affects users, and how actual shopping tasks are supported in VR. The disadvantage is that the construction logic of the scene is not elucidated, and in the scene created by the paper, the consumer can only shop in a single scene and cannot simulate the mobile phone side Feeds stream to have a complete shopping guide link.

Academic document Virtual world system supporting a consumer experience discloses a set of methods to create virtual, consumer shopping experiences that are supported, and describes the design and construction of 3D virtual stores and how to build a database of stores for users to access. The method has the defects that the virtual store is constructed depending on the scene of the off-line real store, the conventional goods shelf type goods organization form is not broken, and various interaction modes with the virtual articles are also lacking.

Disclosure of Invention

In order to solve the problems, the invention provides a 3D shopping scene construction method based on scene styles and scene types, which can effectively improve the cognitive efficiency of a user on commodities and the shopping immersion of the user in a 3D scene.

A3D shopping scene construction method based on scene styles and scene types comprises the following steps:

step 1, constructing a single scene sequence according to the scene type and the stylization degree selected by a user and the scene style obtained by predicting based on the historical behavior data of the user.

The stylization degree includes commonality, decoration, idealization and fiction.

The historical behavior data includes merchandise style tags that a user historically purchased and browsed.

And 2, filling the background mapping and the background sound effect on the single scene in the single scene sequence constructed in the step 1 by utilizing the pre-constructed material database.

And step 3, acquiring account data of the user, including basic information of the user, and consumption habits and consumption capabilities of the user.

And 4, inputting the account data obtained in the step 3 into a pre-constructed prediction model to predict the commodity combination association logic ordering expected by the user.

The commodity combination association logic ordering comprises scene logic-based association, competition logic-based association, complementary logic-based association and style logic-based association.

And 5, screening out commodity combination information meeting the requirements from the material database according to the commodity combination association logic ordering obtained through prediction in the step 4, filling the commodity combination information into the single scene sequence filled with the background map and the background sound effect in the step 2, and outputting a 3D shopping scene meeting the shopping expectations of the user.

According to the invention, the information is reasonably distributed by considering the content and the hierarchy in the information presentation process based on the stylization and the scene types when the 3D shopping scene is constructed, so that the effective touch of the information is optimized, and meanwhile, various combination association logics are adopted, so that the positions of all commodities in the scene are more reasonable.

Specifically, in step 1, the scene styles include a very simple style common to indoor designs and planar designs, a super-realistic style, and a stereoscopic style.

Specifically, in step 1, the single scene sequence includes a bedroom, a bathroom, a study, a conference room, a street, and a park in order.

Specifically, in step 2, the pre-constructed material database includes commodity model data, basic object data, scene background map data, building space data, scene type data, and perceptual information expression model data.

Specifically, the single scene comprises a background layer, a building space layer, a scene layer, a commodity combination layer and a micro scene layer.

The filling process of the single scene is specifically as follows:

and 2-1, calling scene background map data and perceptual information expression model data in the material data, and filling a background layer.

And 2-2, calling building space data to fill obstacles into a building space layer according to the theoretical movable range of the single scene.

And 2-3, calling the objects with the same semantic label in the basic object data according to the style label corresponding to the scene style, and filling the scene layer.

Specifically, in step 4, the prediction model includes a CNN model for extracting different association logic influence values and an LSTM model for performing time-series prediction based on the association logic influence values.

Specifically, the user basic information comprises an identification ID, gender, age and access preference of the user, wherein the access preference comprises identification codes of commodities, commodity flow characteristics, commodity behavior characteristics and commodity decision cost.

The invention provides a 3D shopping scene construction system, which is realized based on the 3D shopping scene construction method based on scene styles and scene types, and comprises the following steps:

the data acquisition unit is used for acquiring account data and historical behavior data of the user;

the material storage unit is used for storing commodity model data, basic object data, scene background map data, building space data, scene type data and perceptual information expression model data;

the prediction unit predicts the commodity combination association logic ordering and scene style expected by the user according to the account data and the historical behavior data;

and the scene construction unit is used for constructing a corresponding 3D shopping scene by utilizing the data in the material storage unit based on the prediction result of the prediction unit.

Compared with the prior art, the invention has the beneficial effects that:

according to the invention, the stylization and scene types are introduced, and the commodity filling is performed by utilizing various commodity combination association logics, so that the cognitive efficiency of the user on the commodity and the immersion sense of the shopping scene in the 3D scene are improved.

Drawings

Fig. 1 is a flowchart of a 3D shopping scenario construction method provided in the present embodiment;

FIG. 2 is a type diagram of the commodity combination association logic according to the present embodiment;

fig. 3 is a frame diagram of a 3D shopping scene construction system according to the present embodiment.

Detailed Description

Technical solutions in the embodiments of the present application will be clearly described below with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present application are within the scope of the protection of the present application.

The terminology used in the present application is for the purpose of describing particular embodiments only and is not intended to be limiting of the present application.

In order to solve the problems that the current business scene lacks of a bottom design for three-dimensional scene of multiple commodities, no obvious association logic exists among multiple commodities provided for a user in the scene, so that a cracking sense is generated between the constructed scene and the commodity use scene or commodity semantics, and the cognitive cost of consumers is increased and the shopping experience is reduced.

As shown in fig. 1, a 3D shopping scene construction method based on scene style and scene type is provided, which includes the following steps:

and constructing a single scene sequence according to the scene type selected by the user and the degree of stylization, wherein the degree of stylization comprises commonality, decoration, idealization, perceptibility and fiction, and the scene style is obtained based on the historical behavior data of the user, and the historical behavior data comprises commodity style labels purchased and browsed by the user in a historical manner.

Furthermore, basic information such as the region, age, sex and the like of the user, and behavior characteristics such as consumption habits, consumption capacity and the like during shopping are obtained, a user interest model is constructed by utilizing a machine learning technology, and the user behavior is tracked and updated in real time.

Keywords of styles or artistic expression commonly used in the art design fields such as indoor or architectural design, planar design and the like are collected and stored in a style database, such as a very simple style, a super-realistic style, a stereoscopic style and the like.

And the style of a scene is determined by the specific style simultaneously with its degree of stylization. The style determines the initial position by the system guessing the style characteristics of the user's preference, such as conciseness, luxury, high-grade feel, etc. The stylized degree is subdivided according to the following 5 stylized degrees:

normal (common), styles in the actual world common user activities or behavior tracks, the combination of the constituent elements of the scene and the elements, and the like all conform to common general knowledge of life of a common person.

The style seen in the real world is Decorative, but the more Decorative and artistic style is occasionally selected for the scene component.

Ideal (idealized), slightly superrealistic styles, have vivid and consistent unnatural styles in terms of selection of scene constituent elements.

The combination of scene components is no longer in line with real life, and the combination method is only used for transmitting the perceptual experience.

Fictive, the scene components themselves are no longer in line with real life, and the objects themselves are for delivering a perceptual experience.

The method comprises the steps of collecting scene keywords accessed by the highest frequency in the daily life of a large number of users through big data, treating the real world as continuous space for traversing and splitting, and taking the following two principles as the basis of scene subdivision:

there is a significant difference in user demand and activity between scene units.

The needs and activities within a scene unit have a close relationship. Through traversal of real world space, the types of scene that can be constructed should be limited. The scene is not limited to the home, but is expanded into a social space (such as an office, a library and the like) and added into a public space which is worth constructing social relations and has the capacity of accommodating a plurality of commodities; expanding to open outdoor and real large-scale environment; and expands to a imagination space beyond reality under certain conditions. For example, building a scene sequence from bedroom in a room to outdoor scene: bedroom- > bathroom- > study- > conference room- > street- > park.

And filling the background mapping and the background sound effect of the single scene in the constructed single scene sequence by utilizing the pre-constructed material database.

Further, the material database includes commodity model data, base object data, scene background map data, building space data, scene type data, and perceptual information expression model data.

The single scene comprises a background layer, a building space layer, a scene layer, a commodity combination layer and a micro scene layer.

The filling process is as follows:

after the style and the type of the scene are determined, a distant view map or model and corresponding atmosphere sound effect are called from a material database to form a background layer, and the background layer can use pure material maps or simulate a real/super-real scene on visual expression.

The building space layer is generally composed of building space shapes and dimensions defining a movable range of a user, such as walls or columns of entities are generally used in indoor scenes, and outdoor scenes can be defined by objects which cannot be spanned in the cognition of a person in real life by using fences, clusters and the like. And calling the model in the building material library and increasing the expression of the scene style by giving different material maps to the model.

And calling model decoration of the basic object data and enriching building space. The object in the basic object data is the same as the commodity object in type, and has the same style or semantic label, but has no attributes such as information, price and the like. The scene layer is laid in the building space with minimal components (scene furnishings or decorations) to constitute an efficient knowledge of the scene for the user.

For example, a living room scene, and the required basic components for forming the correct scene cognition comprise a sofa, a tea table and a television; secondary components including sound, floor lights, etc.; in addition to forming a base member for scene recognition, the expression of style is increased by adding decorative members such as christmas trees, carpets, and the like. A compact but cognitively indistinguishable but incomplete living room can be formed.

The basic object data is called, whether the style label is matched or similar to the scene style label is also required to be considered, firstly, the display and decoration capable of forming scene cognition are selected from a common object library, and the common object is replaced by an object which is overlapped with the common object in the next commodity association.

And acquiring account data of the user, wherein the account data comprises user basic information, and consumption habits and consumption capabilities of the user.

And inputting the obtained account data into a pre-constructed prediction model to predict the commodity combination association logic ordering expected by the user.

Still further, the commodity combination association logic ordering includes 1) scene logic based association; 2) Association based on competing logic; 3) Association based on complementary logic; 4) Based on associative of style logic.

And adjusting the duty ratio of different association logics in the selection link of scene construction-commodity combination according to the properties (style and stylization degree) of the scene and the behavior preference of the user. Firstly, the system collects historical behavior data of a user a on an e-commerce website, wherein the historical behavior data comprise identification ID of the user, gender, age and access preference of the user.

And meanwhile, the identification codes, commodity flow characteristics, commodity behavior characteristics and commodity decision cost of the historical access commodities are collected. And establishing a user commodity probability prediction feature vector according to the historical behavior data. And then, extracting influence values of different association logics from the commodity probability prediction feature vector of the user by using a CNN model, establishing a time sequence prediction model by using an LSTM, and finally outputting a final association logic duty ratio prediction result through a full connection layer. Taking as an example a simple (style) living room (scene type) in which a executive user a who likes a naive style builds reality (stylization degree), different associative logics are explained.

The association based on scene logic, which extracts items with which the user will interact frequently when performing various actions, is based on the "user's trip chart" in the scene, typically by learning that people are in the scene, where the occupancy of the association is highest when selecting the commodity combination for that scene.

The process of determining the commodity combination through scene logic association in the living room is as follows:

the system determines from the user feature vector that user a will often have a preference to keep working while drinking coffee in the living room late at night. The associated merchandise is then determined based on the behavioral preferences: according to the access data and the purchase history data in the user history behavior data, calculating the association degree of the behavior features and the associated commodities by adopting an association rule or a filtering algorithm, and taking the first b commodities with the highest association degree as an associated commodity set of the behavior commodities. The goods that are in the behavior can be determined: coffee cup, computer, lamps and lanterns.

Since the scene type is Normal, the scene type tends to be the same as a real scene, and a user has an explicit behavior path in the scene. The association logic with the second highest duty cycle is therefore based on complementary logic association, with the aim of supplementing the items of user activity that are typically used simultaneously with the determined primary merchandise. For example, in a living office scenario for user a, a computer is typically required for use with a charger, mouse, keyboard.

The association logic with the third highest duty ratio is association based on style logic, and after a series of articles with the same style labels are selected through scene logic association, in order to completely express the style of a scene, the articles with labels such as 'quality of life' and 'modeling atmosphere' and semantic use are needed to be added, although the interaction frequency with a user is not high.

In the above example, the items associated with the scene based on the style logic may include CD players, humidifiers, carpets, and the like. The appearance, design or product expression of the commodities capable of improving the quality of life/shaping the atmosphere is similar to the database category under the "brief" style label.

Since many similar items rarely occur in a realistic scenario closely linked to user behavior, the association based on competing logic can adjust the duty cycle to 0 when constructing this scenario.

In order to increase the probability of further knowing the commodity, planting grass and impulsive purchasing by the user in the shopping scene, partial commodity information is provided and expression of perceptual information (micro scene) is constructed. According to the principle of expressing the scene information: "rational information perceptibility, abstract information materialization", and inductively communicate basic information and value of goods. The method is characterized by comprising the following information design modes:

the basic information of the commodity such as the name, the brand, the commodity introduction and the price simulates the off-line mall to be presented in a 3D tag/label mode which accords with the general cognition of consumers.

The interest points, such as ranking list, new products on the market and the like touch the information of the user itching points to call the 3D bubbles of the corresponding contents in the model library to float around the commodity;

the information of the guarantee class such as package mail, upper door installation and the like is presented in a solid model and an action form, for example, the package mail is expressed in the action form of a 3D express little brother.

As shown in fig. 3, this embodiment further provides a 3D shopping scenario construction system, which is implemented based on the 3D shopping scenario construction method provided in the foregoing embodiment, including:

the data acquisition unit is used for acquiring account data and historical behavior data of the user;

the material storage unit is used for storing commodity model data, basic object data, scene background map data, building space data, scene type data and perceptual information expression model data;

the prediction unit predicts the commodity combination association logic ordering and scene style expected by the user according to the account data and the historical behavior data;

and the scene construction unit is used for constructing a corresponding 3D shopping scene by utilizing the data in the material storage unit based on the prediction result of the prediction unit.

Claims (9)

1. The 3D shopping scene construction method based on scene styles and scene types is characterized by comprising the following steps of:

step 1, constructing a single scene sequence according to scene types selected by a user, the degree of stylization and scene styles obtained through prediction based on historical behavior data of the user, wherein the degree of stylization comprises commonality, decoration, idealization, perceptibility and fiction, and the historical behavior data comprises commodity style labels purchased and browsed by the user in a historical manner;

step 2, filling a background map and background sound effects on a single scene in the single scene sequence constructed in the step 1 by utilizing a pre-constructed material database;

step 3, acquiring account data of a user, including basic information of the user, and consumption habit and consumption capability of the user;

step 4, inputting the account data obtained in the step 3 into a pre-constructed prediction model to predict the commodity combination association logic ordering expected by the user, wherein the commodity combination association logic ordering comprises association based on scene logic, association based on competition logic, association based on complementary logic and association based on style logic;

and 5, screening out commodity combination information meeting the requirements from the material database according to the commodity combination association logic ordering obtained through prediction in the step 4, filling the commodity combination information into the single scene sequence filled with the background map and the background sound effect in the step 2, and outputting a 3D shopping scene meeting the shopping expectations of the user.

2. The 3D shopping scene construction method based on scene style and scene type according to claim 1, wherein in step 1, the scene style includes a very simple style common in indoor design and planar design, an super-realistic style and a stereoscopic style.

3. The method for constructing a 3D shopping scene based on scene style and scene type according to claim 1, wherein in step 1, the single scene sequence sequentially comprises a bedroom, a bathroom, a study room, a conference room, a street and a park.

4. The 3D shopping scene construction method based on scene style and scene type according to claim 1, wherein in step 2, the pre-constructed material database includes commodity model data, basic object data, scene background map data, building space data, scene type data, and perceptual information expression model data.

5. The 3D shopping scene construction method based on scene style and scene type according to claim 1, wherein in step 2, the single scene comprises a background layer, a building space layer, a scene layer, a commodity combination layer and a micro scene layer.

6. The 3D shopping scene construction method based on scene style and scene type according to claim 1 or 5, wherein the filling process of the single scene is specifically as follows:

step 2-1, calling scene background map data and perceptual information expression model data in a material database, and filling the background layer line;

step 2-2, calling building space data to fill barriers to a building space layer according to the theoretical movable range of a single scene;

and 2-3, calling the articles with the same semantic labels in the basic article data according to the style labels corresponding to the scene styles, and filling the scene layers.

7. The 3D shopping scene construction method based on scene style and scene type according to claim 1, wherein in step 4, the prediction model includes a CNN model for extracting different associative logic influence values and an LSTM model for performing time-series prediction based on the associative logic influence values.

8. The 3D shopping scene construction method based on scene style and scene type according to claim 1, wherein in step 4, the user basic information includes identification ID, gender, age and access preference of the user, the access preference includes identification code of commodity, commodity flow characteristics, commodity behavior characteristics and commodity decision cost.

9. A 3D shopping scene construction system, characterized in that it is realized based on the 3D shopping scene construction method based on scene style and scene type according to any one of claims 1-8, comprising:

the data acquisition unit is used for acquiring account data and historical behavior data of the user;

the material storage unit is used for storing commodity model data, basic object data, scene background map data, building space data, scene type data and perceptual information expression model data;

the prediction unit predicts the commodity combination association logic ordering and scene style expected by the user according to the account data and the historical behavior data;

and the scene construction unit is used for constructing a corresponding 3D shopping scene by utilizing the data in the material storage unit based on the prediction result of the prediction unit.

Images