CN111359203B - Personalized railway VR scene interaction method - Google Patents
Personalized railway VR scene interaction method Download PDFInfo
- Publication number
- CN111359203B CN111359203B CN202010156431.2A CN202010156431A CN111359203B CN 111359203 B CN111359203 B CN 111359203B CN 202010156431 A CN202010156431 A CN 202010156431A CN 111359203 B CN111359203 B CN 111359203B
- Authority
- CN
- China
- Prior art keywords
- scene
- user
- time
- browsing
- old user
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63F—CARD, BOARD, OR ROULETTE GAMES; INDOOR GAMES USING SMALL MOVING PLAYING BODIES; VIDEO GAMES; GAMES NOT OTHERWISE PROVIDED FOR
- A63F13/00—Video games, i.e. games using an electronically generated display having two or more dimensions
- A63F13/25—Output arrangements for video game devices
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63F—CARD, BOARD, OR ROULETTE GAMES; INDOOR GAMES USING SMALL MOVING PLAYING BODIES; VIDEO GAMES; GAMES NOT OTHERWISE PROVIDED FOR
- A63F13/00—Video games, i.e. games using an electronically generated display having two or more dimensions
- A63F13/70—Game security or game management aspects
- A63F13/79—Game security or game management aspects involving player-related data, e.g. identities, accounts, preferences or play histories
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Business, Economics & Management (AREA)
- Computer Security & Cryptography (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses an individualized railway VR scene interaction method, which comprises the following steps: s1, judging whether the user is a new user, if so, entering a step S2, otherwise, entering a step S3; s2, recommending hot browsing scenes to the user, and entering the step S5; s3, acquiring personal interests and preferences of the user according to the historical browsing records of the user; s4, recommending browsing scenes according to the personal interests and preferences of the user, and entering the step S5; s5, displaying the corresponding optimal route according to the browsing scene selected by the user; and S6, the user browses scenes according to the optimal route or the custom route. According to the invention, the interest characteristics of the user are abstracted from a large amount of interactive information, the preference of the user is deeply mined, and the multi-level perception and exploration requirements of the user on the railway are fully analyzed, so that an individualized and humanized exploration scheme is provided for the user, and the experience and the interaction efficiency of the user in a railway VR scene are comprehensively improved.
Description
Technical Field
The invention relates to the field of virtual reality, in particular to a personalized railway VR scene interaction method.
Background
Intellectualization is an important direction for future development of railways in the world. With the rapid development of railway technology in China, the development and application range and depth of intelligent new technology in the railway field are continuously expanded, and particularly, a digital twin is an important sign of railway informatization and is also a new way for building intelligent railways.
One core problem that needs to be solved by VR scene exploration analysis is interaction, which is essential in a virtual scene to help users capture target information more quickly, naturally, and efficiently in a virtual environment. The immersion feeling and the experience feeling are more emphasized in the existing game VR scene interaction, and the interaction mode is single. However, the railway line is long, the number of scene objects is large, the spatial relationship is complex, and in the railway construction process, users in different fields and different professional backgrounds are involved, the concerned scene information is different, and a single interaction mode can cause different users to be interfered by additional factors in the interaction process, and the information of interest of the users cannot be accurately acquired, so that the interaction efficiency is low.
Disclosure of Invention
Aiming at the defects in the prior art, the personalized railway VR scene interaction method provided by the invention solves the problem of low interaction efficiency in the existing railway VR scene interaction process.
In order to achieve the purpose of the invention, the invention adopts the technical scheme that:
a personalized railway VR scene interaction method is provided, which comprises the following steps:
s1, reading the login information of the user, judging whether the user is a new user, if so, entering a step S2, otherwise, entering a step S3;
s2, recommending hot browsing scenes to the VR visual interface for the user, and entering the step S5;
s3, acquiring personal interests and preferences of the user according to the historical browsing records of the user;
s4, recommending browsing scenes to a VR visual interface according to the personal interests and preferences of the user, and entering the step S5;
s5, displaying the corresponding optimal route through the VR visual display according to the browsing scene selected by the user;
and S6, browsing scenes by the user according to the optimal route or the user-defined route, and completing interaction of the personalized railway VR scenes.
Further, the specific method for determining whether the user is a new user in step S1 is as follows:
and judging whether the user has a historical browsing record according to the login information of the user, wherein if the user has the historical browsing record, the user is an old user, and if the user does not have the historical browsing record, the user is a new user.
Further, the specific method for recommending a hot browsing scene to the VR visual interface to the new user in step S2 includes the following sub-steps:
s2-1, acquiring a user ID, a browsing scene space position, a browsing scene category, a focus object of attention and interaction time in the historical browsing records of the old user, wherein the interaction time comprises an entering time, a leaving time and a staying time;
s2-2, for the ith old user uiCombining the user ID and the n browsing scene space positions into a vector according to the time sequenceAnd using the vector as the user context information; wherein u isiIs the user ID of the ith old user,the nth browsing scene space position of the ith old user;
s2-3, for the Mth scene, combining the user ID of the ith old user and the spatial position of the browsed scene browsed in the scene into a vector in time sequenceTaking the vector as a track scene vector of the ith old user in the Mth scene, and performing line sequence on the track scene vectors of the ith old user in each scene to obtain track scene information of the ith old user; whereinRepresenting the spatial position of the M-th browsed scene in the M-th scene of the ith old user;
s2-4, for the ith old user uiAnd constructing a position vector according to the coordinates of two positions at the time k and the time k +1, the browsing scene category and the focus object of interestThe ith old user uiAll the position vectors are subjected to rank order to obtain the ith old user uiLocation context information of; whereinAs coordinates of the location of time k, ckBrowsing scene category for time k, okFor the focus object of interest at time k,is the coordinate of the location at time k +1, ck+1Class of browsing scenes at time k +1, ok+1The focus of attention object at the moment k + 1;
s2-5, for the ith old user uiThe row sequence of the user ID, the coordinates of the position, the entering time, the leaving time and the staying time is carried out to obtain the ith old user uiScene information vector at f positionThe ith old user uiSequencing the scene information vectors at each position to obtain the ith old user uiRemoving the data with the stay time less than 5 seconds in the initial time scene information corresponding to the ith old user to obtain the time scene information corresponding to the ith old user; whereinThe coordinates of the ith old user at position f,respectively the entry time, the leaving time and the stay time of the ith old user at the position f;
s2-6, projecting the user scene information, the track scene information, the position scene information and the time scene information of the ith old user to a low-dimensional vector space to respectively obtain a user scene embedding vector, a track scene embedding vector, a position scene embedding vector and a time scene embedding vector, and according to a formula:
obtaining average scene embedding vector of ith old userWhereinA vector is embedded for the user scene of the ith old user,a vector is embedded for the locus scene of the ith old user,a vector is embedded for the location scene of the ith old user,embedding a vector for the time scene of the ith old user;
s2-7, embedding the average scene of the ith old user into a vectorAs the input of a hierarchical sampling softmax function, acquiring the recommendation probability of the ith old user to each browsing scene space position, and further acquiring the recommendation probability of each old user to each browsing scene space position;
s2-8, according to the formula:
obtaining the probability P of recommending Mth scene by all old usersMFurther, the probability that all the old users recommend each scene is obtained; where U represents the set of all old users, R represents the set of trajectories generated by all old users, NRGenerating the length of the track for all the old users, wherein N is the total number of the spatial positions of the browsed scenes in the Mth scene; log (·) is a logarithmic function, and Pr (mi) is the recommendation probability of the ith old user to the spatial position of the mth browsing scene;
s2-9, sequencing each scene according to the recommendation probabilities of all old users from large to small, and pushing the sequencing result to a VR visual interface of a new user as a popular browsing scene list.
Further, the specific method in step S3 includes the following sub-steps:
s3-1, acquiring a user ID, a browsing scene spatial position, a browsing scene category, a focus object of attention and interaction time in the historical browsing record of the user, wherein the interaction time comprises an entering time, a leaving time and a staying time;
s3-2, combining the user ID and the n browsing scene space positions into a vector in chronological orderAnd using the vector as the user context information; wherein u isiIs the user ID of the old user,the nth browsing scene space position of the old user;
s3-3, for the Mth scene, combining the user ID of the old user and the spatial position of the browsing scene browsed in the scene into a vector in time sequenceTaking the vector as the track scene information of the old user in the Mth scene; whereinRepresenting the M-th browsed scene space position of the old user in the M-th scene;
s3-4, constructing a position vector according to the coordinates of the old user at two positions of the k moment and the k +1 moment, browsing scene categories and focus attention objectsThe old user uiAll the position vectors of the old user u are subjected to rank ordering to obtain the old user uiLocation context information of; whereinIs time kCoordinates of the location, ckBrowsing scene category for time k, okFor the focus object of interest at time k,is the coordinate of the location at time k +1, ck+1Class of browsing scenes at time k +1, ok+1The focus of attention object at the moment k + 1;
s3-5, obtaining the old user u by sequencing the user ID, the coordinates of the position, the entering time, the leaving time and the staying time of the old useriScene information vector at f positionThe old user uiThe scene information vector at each position is subjected to rank order to obtain the old user uiRemoving the data with the retention time less than 5 seconds in the initial time scene information corresponding to the old user to obtain the time scene information corresponding to the old user; whereinThe coordinates of the old user at the f position,respectively the entering time, the leaving time and the staying time of the old user at the position f;
s3-6, projecting the user scene information, the track scene information, the position scene information and the time scene information of the ith old user to a low-dimensional vector space to respectively obtain a user scene embedding vector, a track scene embedding vector, a position scene embedding vector and a time scene embedding vector, and according to a formula:
obtaining average scene embedding vector of ith old userWhereinA vector is embedded for the user scene of the ith old user,a vector is embedded for the locus scene of the ith old user,a vector is embedded for the location scene of the ith old user,embedding a vector for the time scene of the ith old user;
s3-7, embedding the average scene of the old user into a vectorAs the input of a hierarchical sampling softmax function, acquiring the recommendation probability of the old user to the spatial position of each browsing scene;
s3-8, according to the formula:
probability P of the old user recommending the Mth sceneMFurther obtaining the probability of each scene recommended by the old user; where U represents the set of all old users, RiRepresents the set of tracks generated by the old user, NRThe length of the track generated for the old user, and N is the total number of the spatial positions of the browsed scenes in the Mth scene; log (-) is a logarithmic function, and Pr (mi) is the recommendation probability of the old user for the spatial position of the mth browsing scene;
s3-9, sorting each scene according to the recommendation probability of the old user from large to small, pushing the sorting result to the VR visual interface of the user as the personal interest and the preference list of the user, and entering the step S5.
Further, the average scene of the i-th old user is embedded into the vector V in step S2-7iThe specific method for acquiring the recommendation probability of the ith old user to each browsing scene space position as the input of the hierarchical sampling softmax function is as follows:
according to a hierarchical sampling softmax function formula:
acquiring the recommendation probability Pr (mi) of the ith old user to the spatial position of the mth browsing scene, and further acquiring the recommendation probability of the ith old user to each spatial position of the browsing scene; wherein N isRLength of the track generated for the ith old user; exp (·) is an exponential function with a natural constant e as the base;embedding vectors for the average scene of the ith old userTransposing a matrix; thetap-1Sampling parameters of a p-1 node of a softmax function in a layered mode; b is a value parameter, and when the mth browsing scene space position takes a left branch at the pth node, b is 0; when the mth browsing scene space location takes the right branch at the pth node, b is 1.
Further, the average scene of the old user is embedded into the vector V in step S3-7iThe specific method for acquiring the recommendation probability of the old user to each browsing scene space position as the input of the hierarchical sampling softmax function is as follows:
according to a hierarchical sampling softmax function formula:
obtaining the recommendation probability Pr (mi) of the ith old user to the spatial position of the mth browsing scene) Further obtaining the recommendation probability of the ith old user to the space position of each browsing scene; wherein N isRLength of the track generated for the ith old user; exp (·) is an exponential function with a natural constant e as the base;embedding vectors for the average scene of the ith old userTransposing a matrix; thetap-1Sampling parameters of a p-1 node of a softmax function in a layered mode; b is a value parameter, and when the mth browsing scene space position takes a left branch at the pth node, b is 0; when the mth browsing scene space location takes the right branch at the pth node, b is 1.
The invention has the beneficial effects that: according to the invention, the interest characteristics of the user are abstracted from a large amount of interactive information, the preference of the user is deeply excavated, and the multi-level perception and exploration requirements of the user on the railway are fully analyzed, so that an individualized and humanized exploration scheme is provided for the user, the experience of the user in a railway VR scene is comprehensively improved, and the interaction efficiency of the railway VR is further improved.
Drawings
FIG. 1 is a schematic flow chart of the present invention.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate the understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and it will be apparent to those skilled in the art that various changes may be made without departing from the spirit and scope of the invention as defined and defined in the appended claims, and all matters produced by the invention using the inventive concept are protected.
As shown in fig. 1, the personalized railway VR scene interaction method includes the following steps:
s1, reading the login information of the user, judging whether the user is a new user, if so, entering a step S2, otherwise, entering a step S3;
s2, recommending hot browsing scenes to the VR visual interface for the user, and entering the step S5;
s3, acquiring personal interests and preferences of the user according to the historical browsing records of the user;
s4, recommending browsing scenes to a VR visual interface according to the personal interests and preferences of the user, and entering the step S5;
s5, displaying the corresponding optimal route through the VR visual display according to the browsing scene selected by the user;
and S6, browsing scenes by the user according to the optimal route or the user-defined route, and completing interaction of the personalized railway VR scenes.
The specific method for determining whether the user is a new user in step S1 is as follows: and judging whether the user has a historical browsing record according to the login information of the user, wherein if the user has the historical browsing record, the user is an old user, and if the user does not have the historical browsing record, the user is a new user.
The specific method for recommending the hot browsing scene to the VR visual interface to the new user in step S2 includes the following substeps:
s2-1, acquiring a user ID, a browsing scene space position, a browsing scene category, a focus object of attention and interaction time in the historical browsing records of the old user, wherein the interaction time comprises an entering time, a leaving time and a staying time;
s2-2, for the ith old user uiCombining the user ID and the n browsing scene space positions into a vector according to the time sequenceAnd using the vector as the user context information; wherein u isiIs the user ID of the ith old user,the nth browsing scene space position of the ith old user;
s2-3, for the Mth scene, combining the user ID of the ith old user and the spatial position of the browsed scene browsed in the scene into a vector in time sequenceTaking the vector as a track scene vector of the ith old user in the Mth scene, and performing line sequence on the track scene vectors of the ith old user in each scene to obtain track scene information of the ith old user; whereinRepresenting the spatial position of the M-th browsed scene in the M-th scene of the ith old user;
s2-4, for the ith old user uiAnd constructing a position vector according to the coordinates of two positions at the time k and the time k +1, the browsing scene category and the focus object of interestThe ith old user uiAll the position vectors are subjected to rank order to obtain the ith old user uiLocation context information of; whereinAs coordinates of the location of time k, ckBrowsing scene category for time k, okFor the focus object of interest at time k,is the coordinate of the location at time k +1, ck+1Class of browsing scenes at time k +1, ok+1The focus of attention object at the moment k + 1;
s2-5, for the ith old user uiThe row sequence of the user ID, the coordinates of the position, the entering time, the leaving time and the staying time is carried out to obtain the ith old user uiScene information vector at f positionThe ith old user uiSequencing the scene information vectors at each position to obtain the ith old user uiRemoving the initial time corresponding to the ith old userThe data with the stay time of less than 5 seconds in the context information obtain the time context information corresponding to the ith old user; whereinThe coordinates of the ith old user at position f,respectively the entry time, the leaving time and the stay time of the ith old user at the position f;
s2-6, projecting the user scene information, the track scene information, the position scene information and the time scene information of the ith old user to a low-dimensional vector space to respectively obtain a user scene embedding vector, a track scene embedding vector, a position scene embedding vector and a time scene embedding vector, and according to a formula:
obtaining average scene embedding vector of ith old userWhereinA vector is embedded for the user scene of the ith old user,a vector is embedded for the locus scene of the ith old user,a vector is embedded for the location scene of the ith old user,embedding a vector for the time scene of the ith old user;
s2-7, embedding the average scene of the ith old user into a vectorAs the input of a hierarchical sampling softmax function, acquiring the recommendation probability of the ith old user to each browsing scene space position, and further acquiring the recommendation probability of each old user to each browsing scene space position;
s2-8, according to the formula:
obtaining the probability P of recommending Mth scene by all old usersMFurther, the probability that all the old users recommend each scene is obtained; where U represents the set of all old users, R represents the set of trajectories generated by all old users, NRGenerating the length of the track for all the old users, wherein N is the total number of the spatial positions of the browsed scenes in the Mth scene; log (·) is a logarithmic function, and Pr (mi) is the recommendation probability of the ith old user to the spatial position of the mth browsing scene;
s2-9, sequencing each scene according to the recommendation probabilities of all old users from large to small, and pushing the sequencing result to a VR visual interface of a new user as a popular browsing scene list.
The specific method in step S3 includes the following substeps:
s3-1, acquiring a user ID, a browsing scene spatial position, a browsing scene category, a focus object of attention and interaction time in the historical browsing record of the user, wherein the interaction time comprises an entering time, a leaving time and a staying time;
s3-2, combining the user ID and the n browsing scene space positions into a vector in chronological orderAnd using the vector as the user context information; wherein u isiIs the user ID of the old user,for the old to useThe nth browsing scene space position of the user;
s3-3, for the Mth scene, combining the user ID of the old user and the spatial position of the browsing scene browsed in the scene into a vector in time sequenceTaking the vector as the track scene information of the old user in the Mth scene; whereinRepresenting the M-th browsed scene space position of the old user in the M-th scene;
s3-4, constructing a position vector according to the coordinates of the old user at two positions of the k moment and the k +1 moment, browsing scene categories and focus attention objectsThe old user uiAll the position vectors of the old user u are subjected to rank ordering to obtain the old user uiLocation context information of; whereinAs coordinates of the location of time k, ckBrowsing scene category for time k, okFor the focus object of interest at time k,is the coordinate of the location at time k +1, ck+1Class of browsing scenes at time k +1, ok+1The focus of attention object at the moment k + 1;
s3-5, obtaining the old user u by sequencing the user ID, the coordinates of the position, the entering time, the leaving time and the staying time of the old useriScene information vector at f positionThe old user uiThe scene information vector at each position is subjected to rank order to obtain the old user uiRemoving the data with the retention time less than 5 seconds in the initial time scene information corresponding to the old user to obtain the time scene information corresponding to the old user; whereinThe coordinates of the old user at the f position,respectively the entering time, the leaving time and the staying time of the old user at the position f;
s3-6, projecting the user scene information, the track scene information, the position scene information and the time scene information of the ith old user to a low-dimensional vector space to respectively obtain a user scene embedding vector, a track scene embedding vector, a position scene embedding vector and a time scene embedding vector, and according to a formula:
obtaining average scene embedding vector of ith old userWhereinA vector is embedded for the user scene of the ith old user,a vector is embedded for the locus scene of the ith old user,a vector is embedded for the location scene of the ith old user,embedding a vector for the time scene of the ith old user;
S3-7, embedding the average scene of the old user into a vectorAs the input of a hierarchical sampling softmax function, acquiring the recommendation probability of the old user to the spatial position of each browsing scene;
s3-8, according to the formula:
probability P of the old user recommending the Mth sceneMFurther obtaining the probability of each scene recommended by the old user; where U represents the set of all old users, RiRepresents the set of tracks generated by the old user, NRThe length of the track generated for the old user, and N is the total number of the spatial positions of the browsed scenes in the Mth scene; log (-) is a logarithmic function, and Pr (mi) is the recommendation probability of the old user for the spatial position of the mth browsing scene;
s3-9, sorting each scene according to the recommendation probability of the old user from large to small, pushing the sorting result to the VR visual interface of the user as the personal interest and the preference list of the user, and entering the step S5.
In one embodiment of the invention, the average scene of the i-th old user is embedded into a vector ViThe specific method for acquiring the recommendation probability of the ith old user to each browsing scene space position as the input of the hierarchical sampling softmax function is as follows: according to a hierarchical sampling softmax function formula:
acquiring the recommendation probability Pr (mi) of the ith old user to the spatial position of the mth browsing scene, and further acquiring the recommendation probability of the ith old user to each spatial position of the browsing scene; wherein N isRLength of the track generated for the ith old user; exp (·) is an exponential function with a natural constant e as the base;embedding vectors for the average scene of the ith old userTransposing a matrix; thetap-1Sampling parameters of a p-1 node of a softmax function in a layered mode; b is a value parameter, and when the mth browsing scene space position takes a left branch at the pth node, b is 0; when the mth browsing scene space location takes the right branch at the pth node, b is 1.
Each node in the hierarchically sampled binary tree structure is associated with an embedded vector that computes the associated branch probability. In this configuration, each leaf can be reached from a node of the first layer through an appropriate path. For hierarchical sampling the softmax function can train all parameters in the function by random gradient descent (SGD). In parameter training, the function iterates over all trajectory positions and computes the gradient of the parameters through an error back-propagation algorithm, which is then used to update the parameters in the function until convergence.
In a specific implementation process, parameters such as the spatial position, the spatial attitude, the scaling and the like of equipment such as a VR handle, a VR helmet and the like in a virtual scene can be recorded based on technologies such as spatial positioning and gravity sensing of VR hardware equipment; secondly, the method can utilize technologies such as ray focus and the like to obtain the focus information data of the user; finally, the invention can acquire the background data in the VR scene by an information acquisition technology, and process and optimize the acquired interactive information by an information processing technology.
Obtaining the operating parameters of the handle: firstly, a user selects a specific key on a VR handle to operate according to own requirements, the user can realize displacement in a scene by emitting rays to a designated point through the handle in the scene, the system acquires the operation of the user when using the handle by detecting the state of the handle, the system records the information of the position, the spatial attitude, the scaling, the use time and the like of the handle in the virtual scene in the operation of the handle, and the information is stored in the system through a data persistence technology.
For virtual scene user spatial location data acquisition: and taking the spatial position of a head-mounted display in the VR hardware device in the constructed scene as the position of the current user, and recording the position information of the user in the virtual scene at certain time intervals, wherein the position information comprises X, Y, Z three coordinates in the scene and time record.
For user attention focus information acquisition: the center coordinate in the visual field of the head-mounted display is used as a user focus of attention, the specific implementation mode is that the center of the head-mounted display of a current user emits rays, the rays collide with an object in a scene, a first collision point is used as the current user focus of attention, and focus information comprises X, Y, Z coordinates, recording time and information of the object where the focus is located.
Information acquisition and data processing: background data in a VR scene is acquired through the technology, data preprocessing is carried out on the acquired interactive information through the information processing technology, the processed data format comprises user ID, spatial position of a browsed scene, types of the browsed scene, focus objects, entry time, exit time, stay time and other elements, and an interactive information data set is constructed on the basis. For example, after a user enters a VR scene, each time the user accesses one scene by using a VR handle, a piece of interaction information { u, l, c, o, t is generated1,t2,t3Where l can be represented by coordinates (X, Y, Z), i.e., { user ID, browsing scene spatial location, browsing scene category, focus of interest object, time-in, time-out, dwell time }, i.e., user u is at t1Enters the o focus object in the c type scene with the space position l at the moment and reaches t2The moment leaves and stays for t3And second. For example { UID001, (X, Y, Z), station, security equipment, 2019-01-20/3:30pm, 2019-01-20/4:00pm, 1800s }, which indicates that a user with a user ID of 001 enters the station with spatial position (X, Y, Z) at No. 1/20 pm of 2019 and 3:30 in the afternoon, and watches the security equipment, and leaves at No. 1/20 of 2019 and 4:00 in the afternoon, staying for 1800 seconds.
In conclusion, the interest characteristics of the user are abstracted from a large amount of interactive information, the preference of the user is deeply mined, and the multi-level perception and exploration requirements of the user on the railway are fully analyzed, so that a personalized and humanized exploration scheme is provided for the user, the experience of the user in a railway VR scene is comprehensively improved, and the interactive efficiency of the railway VR is further improved.
Claims (5)
1. A personalized railway VR scene interaction method is characterized by comprising the following steps:
s1, reading the login information of the user, judging whether the user is a new user, if so, entering a step S2, otherwise, entering a step S3;
s2, recommending hot browsing scenes to the VR visual interface for the user, and entering the step S5;
s3, acquiring personal interests and preferences of the user according to the historical browsing records of the user;
s4, recommending browsing scenes to a VR visual interface according to the personal interests and preferences of the user, and entering the step S5;
s5, displaying the corresponding optimal route through the VR visual display according to the browsing scene selected by the user;
s6, browsing scenes by the user according to the optimal route or the user-defined route, and completing interaction of the personalized railway VR scenes;
the specific method for recommending the hot browsing scene to the VR visual interface to the new user in the step S2 includes the following sub-steps:
s2-1, acquiring a user ID, a browsing scene space position, a browsing scene category, a focus object of attention and interaction time in the historical browsing records of the old user, wherein the interaction time comprises an entering time, a leaving time and a staying time;
s2-2, for the ith old user uiCombining the user ID and the n browsing scene space positions into a vector according to the time sequenceAnd using the vector as the user context information; wherein u isiFor the i-th old userThe user ID is set to the user ID,the nth browsing scene space position of the ith old user;
s2-3, for the Mth scene, combining the user ID of the ith old user and the spatial position of the browsed scene browsed in the scene into a vector in time sequenceTaking the vector as a track scene vector of the ith old user in the Mth scene, and performing line sequence on the track scene vectors of the ith old user in each scene to obtain track scene information of the ith old user; whereinRepresenting the spatial position of the M-th browsed scene in the M-th scene of the ith old user;
s2-4, for the ith old user uiAnd constructing a position vector according to the coordinates of two positions at the time k and the time k +1, the browsing scene category and the focus object of interestThe ith old user uiAll the position vectors are subjected to rank order to obtain the ith old user uiLocation context information of; whereinAs coordinates of the location of time k, ckBrowsing scene category for time k, okFor the focus object of interest at time k,is the coordinate of the location at time k +1, ck+1Class of browsing scenes at time k +1, ok+1The focus of attention object at the moment k + 1;
s2-5, for the ith old ageHuu (household)iThe row sequence of the user ID, the coordinates of the position, the entering time, the leaving time and the staying time is carried out to obtain the ith old user uiScene information vector at f positionThe ith old user uiSequencing the scene information vectors at each position to obtain the ith old user uiRemoving the data with the stay time less than 5 seconds in the initial time scene information corresponding to the ith old user to obtain the time scene information corresponding to the ith old user; whereinThe coordinates of the ith old user at position f,respectively the entry time, the leaving time and the stay time of the ith old user at the position f;
s2-6, projecting the user scene information, the track scene information, the position scene information and the time scene information of the ith old user to a low-dimensional vector space to respectively obtain a user scene embedding vector, a track scene embedding vector, a position scene embedding vector and a time scene embedding vector, and according to a formula:
obtaining average scene embedding vector of ith old userWhereinA vector is embedded for the user scene of the ith old user,a vector is embedded for the locus scene of the ith old user,embedding a vector, V, for the location context of the ith old usert iEmbedding a vector for the time scene of the ith old user;
s2-7, embedding the average scene of the ith old user into a vectorAs the input of a hierarchical sampling softmax function, acquiring the recommendation probability of the ith old user to each browsing scene space position, and further acquiring the recommendation probability of each old user to each browsing scene space position;
s2-8, according to the formula:
obtaining the willingness value P of all the old users to recommend the Mth sceneMFurther obtaining the willingness value of all the old users for recommending each scene; where U represents the set of all old users, R represents the set of trajectories generated by all old users, NRGenerating the length of the track for all the old users, wherein N is the total number of the spatial positions of the browsed scenes in the Mth scene; log (-) is a logarithmic function, and Pr (m | i) is the recommendation probability of the ith old user to the spatial position of the mth browsing scene;
s2-9, sequencing each scene from large to small according to the recommendation willingness values of all old users, and pushing the sequencing result to a VR visual interface of a new user as a popular browsing scene list.
2. The personalized railway VR scene interaction method of claim 1, wherein the specific method for determining whether the user is a new user in step S1 is as follows:
and judging whether the user has a historical browsing record according to the login information of the user, wherein if the user has the historical browsing record, the user is an old user, and if the user does not have the historical browsing record, the user is a new user.
3. The personalized railway VR scene interaction method of claim 1, wherein the specific method in step S3 includes the following sub-steps:
s3-1, acquiring a user ID, a browsing scene spatial position, a browsing scene category, a focus object of attention and interaction time in the historical browsing record of the user, wherein the interaction time comprises an entering time, a leaving time and a staying time;
s3-2, combining the user ID and the n browsing scene space positions into a vector in chronological orderAnd using the vector as the user context information; wherein u isiIs the user ID of the old user,the nth browsing scene space position of the old user;
s3-3, for the Mth scene, combining the user ID of the old user and the spatial position of the browsing scene browsed in the scene into a vector in time sequenceTaking the vector as the track scene information of the old user in the Mth scene; whereinRepresenting the M-th browsed scene space position of the old user in the M-th scene;
s3-4, constructing a position vector according to the coordinates of the old user at two positions of the k moment and the k +1 moment, browsing scene categories and focus attention objectsThe old user uiAll the position vectors of the old user u are subjected to rank ordering to obtain the old user uiLocation context information of; whereinAs coordinates of the location of time k, ckBrowsing scene category for time k, okFor the focus object of interest at time k,is the coordinate of the location at time k +1, ck+1Class of browsing scenes at time k +1, ok+1The focus of attention object at the moment k + 1;
s3-5, obtaining the old user u by sequencing the user ID, the coordinates of the position, the entering time, the leaving time and the staying time of the old useriScene information vector at f positionThe old user uiThe scene information vector at each position is subjected to rank order to obtain the old user uiRemoving the data with the retention time less than 5 seconds in the initial time scene information corresponding to the old user to obtain the time scene information corresponding to the old user; whereinThe coordinates of the old user at the f position,respectively the entering time, the leaving time and the staying time of the old user at the position f;
s3-6, projecting the user scene information, the track scene information, the position scene information and the time scene information of the ith old user to a low-dimensional vector space to respectively obtain a user scene embedding vector, a track scene embedding vector, a position scene embedding vector and a time scene embedding vector, and according to a formula:
obtaining average scene embedding vector of ith old userWhereinA vector is embedded for the user scene of the ith old user,a vector is embedded for the locus scene of the ith old user,embedding a vector, V, for the location context of the ith old usert iEmbedding a vector for the time scene of the ith old user;
s3-7, embedding the average scene of the old user into a vectorAs the input of a hierarchical sampling softmax function, acquiring the recommendation probability of the old user to the spatial position of each browsing scene;
s3-8, according to the formula:
the old user recommends the willingness value P of the Mth sceneMFurther obtaining the willingness value of the old user for recommending each scene; wherein R isiRepresents the set of tracks generated by the old user, NRThe length of the track generated for the old user, and N is the total number of the spatial positions of the browsed scenes in the Mth scene; log (-) is logarithmicFunction, Pr (m | i) is the recommendation probability of the old user to the spatial position of the mth browsing scene;
s3-9, sorting each scene according to the recommendation willingness value of the old user from big to small, pushing the sorting result to the VR visual interface of the user as the personal interest and the preference list of the user, and entering the step S5.
4. The personalized railway VR scene interaction method of claim 1, wherein the average scene of the ith old user is embedded into a vector in step S2-7The specific method for acquiring the recommendation probability of the ith old user to each browsing scene space position as the input of the hierarchical sampling softmax function is as follows:
according to a hierarchical sampling softmax function formula:
acquiring the recommendation probability Pr (m | i) of the ith old user to the spatial position of the mth browsing scene, and further acquiring the recommendation probability of the ith old user to each spatial position of the browsing scene; wherein N isRLength of the track generated for the ith old user; exp (·) is an exponential function with a natural constant e as the base;embedding vectors for the average scene of the ith old userTransposing a matrix; thetap-1Sampling parameters of a p-1 node of a softmax function in a layered mode; b is a value parameter, and when the mth browsing scene space position takes a left branch at the pth node, b is 0; when the mth browsing scene space location takes the right branch at the pth node, b is 1.
5. The personalized railway VR scene interaction method of claim 3, wherein the average scene of the old user is embedded into a vector in step S3-7The specific method for acquiring the recommendation probability of the old user to each browsing scene space position as the input of the hierarchical sampling softmax function is as follows:
according to a hierarchical sampling softmax function formula:
acquiring the recommendation probability Pr (m | i) of the ith old user to the spatial position of the mth browsing scene, and further acquiring the recommendation probability of the ith old user to each spatial position of the browsing scene; wherein N isRLength of the track generated for the ith old user; exp (·) is an exponential function with a natural constant e as the base;embedding vectors for the average scene of the ith old userTransposing a matrix; thetap-1Sampling parameters of a p-1 node of a softmax function in a layered mode; b is a value parameter, and when the mth browsing scene space position takes a left branch at the pth node, b is 0; when the mth browsing scene space location takes the right branch at the pth node, b is 1.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010156431.2A CN111359203B (en) | 2020-03-09 | 2020-03-09 | Personalized railway VR scene interaction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010156431.2A CN111359203B (en) | 2020-03-09 | 2020-03-09 | Personalized railway VR scene interaction method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111359203A CN111359203A (en) | 2020-07-03 |
CN111359203B true CN111359203B (en) | 2021-09-28 |
Family
ID=71198381
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010156431.2A Active CN111359203B (en) | 2020-03-09 | 2020-03-09 | Personalized railway VR scene interaction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111359203B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112395323A (en) * | 2020-09-18 | 2021-02-23 | 江苏园上园智能科技有限公司 | Interaction method based on user experience data |
CN113076436B (en) * | 2021-04-09 | 2023-07-25 | 成都天翼空间科技有限公司 | VR equipment theme background recommendation method and system |
CN113704605A (en) * | 2021-08-24 | 2021-11-26 | 山东库睿科技有限公司 | Service information recommendation method and device, electronic equipment and medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8065254B1 (en) * | 2007-02-19 | 2011-11-22 | Google Inc. | Presenting a diversity of recommendations |
CN104794207A (en) * | 2015-04-23 | 2015-07-22 | 山东大学 | Recommendation system based on cooperation and working method of recommendation system |
CN108303108A (en) * | 2017-12-05 | 2018-07-20 | 华南理工大学 | A kind of personalized route recommendation method based on vehicle historical track |
CN108733653A (en) * | 2018-05-18 | 2018-11-02 | 华中科技大学 | A kind of sentiment analysis method of the Skip-gram models based on fusion part of speech and semantic information |
CN110738370A (en) * | 2019-10-15 | 2020-01-31 | 南京航空航天大学 | novel moving object destination prediction algorithm |
CN110807150A (en) * | 2019-10-14 | 2020-02-18 | 腾讯科技(深圳)有限公司 | Information processing method and device, electronic equipment and computer readable storage medium |
-
2020
- 2020-03-09 CN CN202010156431.2A patent/CN111359203B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8065254B1 (en) * | 2007-02-19 | 2011-11-22 | Google Inc. | Presenting a diversity of recommendations |
CN104794207A (en) * | 2015-04-23 | 2015-07-22 | 山东大学 | Recommendation system based on cooperation and working method of recommendation system |
CN108303108A (en) * | 2017-12-05 | 2018-07-20 | 华南理工大学 | A kind of personalized route recommendation method based on vehicle historical track |
CN108733653A (en) * | 2018-05-18 | 2018-11-02 | 华中科技大学 | A kind of sentiment analysis method of the Skip-gram models based on fusion part of speech and semantic information |
CN110807150A (en) * | 2019-10-14 | 2020-02-18 | 腾讯科技(深圳)有限公司 | Information processing method and device, electronic equipment and computer readable storage medium |
CN110738370A (en) * | 2019-10-15 | 2020-01-31 | 南京航空航天大学 | novel moving object destination prediction algorithm |
Also Published As
Publication number | Publication date |
---|---|
CN111359203A (en) | 2020-07-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111359203B (en) | Personalized railway VR scene interaction method | |
CN110309427B (en) | Object recommendation method and device and storage medium | |
CN109086439A (en) | Information recommendation method and device | |
CN111078939A (en) | Method, system and recording medium for extracting and providing highlight image in video content | |
CN104782138A (en) | Identifying a thumbnail image to represent a video | |
CN105493057A (en) | Content selection with precision controls | |
CN111949877B (en) | Personalized interest point recommendation method and system | |
CN112734068B (en) | Conference room reservation method, conference room reservation device, computer equipment and storage medium | |
CN111708876A (en) | Method and device for generating information | |
CN112215129A (en) | Crowd counting method and system based on sequencing loss and double-branch network | |
CN110555428A (en) | Pedestrian re-identification method, device, server and storage medium | |
CN111783895B (en) | Travel plan recommendation method, device, computer equipment and storage medium based on neural network | |
CN108197203A (en) | A kind of shop front head figure selection method, device, server and storage medium | |
CN110465089A (en) | Map heuristic approach, device, medium and electronic equipment based on image recognition | |
CN111859142A (en) | Cross-equipment migration recommendation system based on interconnection and intercommunication home platform and working method thereof | |
CN113158038A (en) | Interest point recommendation method and system based on STA-TCN neural network framework | |
JPWO2010084839A1 (en) | Likelihood estimation apparatus, content distribution system, likelihood estimation method, and likelihood estimation program | |
CN113742590A (en) | Recommendation method and device, storage medium and electronic equipment | |
CN112784177A (en) | Spatial distance adaptive next interest point recommendation method | |
CN116257704A (en) | Point-of-interest recommendation method based on user space-time behaviors and social information | |
CN113158735A (en) | Dense event description method based on graph neural network | |
CN111414538A (en) | Text recommendation method and device based on artificial intelligence and electronic equipment | |
CN116628310B (en) | Content recommendation method, device, equipment, medium and computer program product | |
CN117763492B (en) | Network security tool intelligent recommendation method and device based on time sequence spatial characteristics and preference fluctuation | |
CN116932893B (en) | Sequence recommendation method, system, equipment and medium based on graph convolution network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |