CN117520662A

CN117520662A - Intelligent scenic spot guiding method and system based on positioning

Info

Publication number: CN117520662A
Application number: CN202410012803.2A
Authority: CN
Inventors: 张卓; 金岩; 宋炜; 刘飞
Original assignee: Shaanxi Yunchuang Network Technology Co ltd
Current assignee: Shaanxi Yunchuang Network Technology Co ltd
Priority date: 2024-01-04
Filing date: 2024-01-04
Publication date: 2024-02-06

Abstract

The application discloses a scenic spot intelligent navigation method and system based on positioning, which are used for solving the problems that the utilization of traffic information in scenic spots is insufficient, the limitations exist in scenic spot information summarization analysis and information pushing, and personalized navigation service cannot be provided in the prior art. The intelligent scenic spot guiding method comprises the following steps: acquiring expected preference data of tourists and real-time distribution data of scenic spot personnel; generating a plurality of recommended routes by utilizing expected preference data and real-time distribution data; acquiring position information of tourists in a recommended route, and pushing nearby scenic spot information according to the position information; the method comprises the steps of obtaining actual preference data of tourists through the stay time of each scenic spot and interaction of the tourists and the information of the pushed scenic spots in the tourist browsing process, and utilizing the obtained actual preference data to push the customized scenic spot information of the tourists.

Description

Intelligent scenic spot guiding method and system based on positioning

Technical Field

The application relates to the technical field of scenic spot navigation methods and systems, in particular to a scenic spot intelligent navigation method and system based on positioning.

Background

Existing tour guide systems often rely on preloaded sight and facility information that may provide outdated or erroneous information if not updated in time; the use of a location-based tour guide system may require some familiar procedures such as how to operate the device, how to view the map, etc. If the user interface is not designed to be friendly enough, it may be difficult for the user to operate; at present, manual explanation or recording equipment is mostly used for guiding, information of the equipment is fixed, personalized adjustment is inconvenient, and user experience has a large limitation.

Wherein, application publication number is: CN 103994771A, application name: the invention discloses an intelligent navigation application system for scenic spots and a use method thereof. The application method of the intelligent navigation application system for the scenic spot comprises the following steps: the personal mobile communication equipment sends a service request to a big data center; the big data platform pushes scenic spot information to the personal mobile communication equipment; the user sends a route planning request to the big data platform center through the personal mobile communication equipment; the big data platform selects and filters out the optimal tour route and pushes the optimal tour route to the personal mobile communication device of the tourist. The problems with this patent are: the tourist route can not be intelligently planned according to the pedestrian flow of the scenic spot, and the pushed scenic spot information is fixed.

The application publication number is: CN106485217 a, application name: a method and system for identifying the saturation of a stream of people in a tourist attraction, the method comprising the steps of: video data acquisition: collecting video data by using a video collecting device arranged at each entrance/exit gateway of the scenic spot; extracting target features: carrying out image analysis processing on the collected video data at regular time, and extracting target features; and (3) judging the saturation: analyzing whether each target feature matches stored feature data in memory: if the target features cannot be matched with the feature data in the memory, modifying the population according to the number of the unmatched target features, and storing the target features which are corresponding to the unmatched target features in the memory into the memory. The problems with this patent are: the people stream information data of the scenic spot cannot be effectively utilized, and the collection analysis of various information of the scenic spot and the pushing of the scenic spot information cannot be realized.

Disclosure of Invention

Therefore, the application provides a location-based intelligent scenic spot guiding method and system, which are used for solving the problems that the utilization of traffic information in scenic spots is insufficient, the limitations exist in scenic spot information summarization analysis and information pushing, and personalized guiding service cannot be provided in the prior art.

In order to achieve the above object, the present application provides the following technical solutions:

first aspect: a location-based scenic spot intelligent navigation method comprises the following steps:

acquiring expected preference data of tourists and real-time distribution data of scenic spot personnel;

generating a plurality of recommended routes by utilizing expected preference data and real-time distribution data;

acquiring position information of tourists in a recommended route, and pushing nearby scenic spot information according to the position information;

the method comprises the steps of obtaining actual preference data of tourists through the stay time of each scenic spot and interaction of the tourists and the information of the pushed scenic spots in the tourist browsing process, and utilizing the obtained actual preference data to push the customized scenic spot information of the tourists.

Optionally, the method further comprises:

collecting real-time video data of all places of a scenic spot, and processing frame images in the collected video data to obtain a video image sequence;

extracting a target connected domain of each tourist of each frame of image in a video image sequence, and carrying out continuous and uninterrupted track tracking on the target connected domain in the video image sequence to acquire the real-time distribution data of scenic spot personnel;

and processing the real-time distribution data to obtain the people stream information of each area in the scenic spot at the next moment, and updating the recommended route for the tourist according to the people stream information of each area at the next moment.

Optionally, processing the real-time distribution data to obtain people stream information of each area at the next moment in the scenic spot, and updating a recommended route for the tourist according to the people stream information of each area at the next moment, including:

processing the real-time distribution data by using an autoregressive integral moving average algorithm to obtain estimated position information of the target connected domain at the next moment;

and acquiring people flow information of each area in the scenic spot at the next moment according to the estimated position information of all the target connected areas in the scenic spot, screening and sorting a plurality of recommended routes according to the people flow information of each area at the next moment, and updating the recommended routes according to the screening and sorting results.

Optionally, the continuous track tracing is performed on the target connected domain, including:

performing similarity matching on the target connected domain obtained in each frame of image in the video image sequence and the target connected domain in other frame of images to obtain a successfully matched target connected domain in each frame of image;

and marking the successfully matched target connected domain according to the sequence of the frame images, and completing continuous and uninterrupted track tracking of the target connected domain.

Optionally, the generating a plurality of recommended routes by using the expected preference data and the real-time distribution data includes:

collecting expected preference data of tourists in a manner of interactive questionnaires, historical access data or keyword label classification, recording each type of expected preference data as an element, correspondingly distributing a weight value to each element, and respectively summing the weight values and the corresponding elements to obtain weighted overall preference scores;

and calculating the overall preference score and the real-time distribution data through a mean value clustering algorithm to obtain a plurality of clusters, wherein each cluster represents a recommended route to generate a plurality of recommended routes by taking the current position of the tourist as a starting point.

Optionally, the obtaining the position information of the tourist in the recommended route and pushing the nearby scenic spot information according to the position information includes:

acquiring real-time position information of tourists in the recommended route by using a GPS, wi-Fi positioning, bluetooth beacons or Beidou positioning system;

and acquiring the distance between the current position of the tourist and the coordinates of the scenic spot by using a neighbor analysis algorithm, and pushing the scenic spot information to the tourist when the distance is smaller than a preset first threshold value.

Optionally, the scenic spot information pushing method for customizing the tourists comprises the following steps:

acquiring real-time position information of tourists, stay time of each scenic spot, position information of the scenic spot and interaction data of the scenic spot information, wherein the interaction data comprises click rate and interaction time of the scenic spot information;

analyzing the real-time position information of the tourists, the stay time of each scenic spot, the position information of the scenic spot and the interaction data of the scenic spot information to obtain the actual preference data, and generating customized scenic spot information according to the actual preference data by applying a collaborative filtering algorithm or a content-based recommendation algorithm and pushing the customized scenic spot information to the tourists.

Optionally, the method for obtaining the expected preference data of the tourist comprises obtaining the expected preference data of the tourist through an online questionnaire, an online interactive tour game or historical access data.

Optionally, the method further comprises: scenic spot staff updates the scenic spot information, including scenic spot profile, surrounding food, personal features, historical background, and scenic spot activity.

Second aspect: a scenic spot intelligent navigation system based on positioning comprises a video acquisition module, a cloud processing module and a tourist personal communication end;

The video acquisition module is used for acquiring real-time distribution data of scenic spot personnel;

the cloud processing module generates a plurality of recommended routes by utilizing expected preference data and real-time distribution data, acquires position information of tourists in the recommended routes, pushes nearby scenic spot information according to the position information, acquires actual preference data of the tourists through the stay time of each scenic spot of the tourists in the tour process and the interaction of the tourists and the pushed scenic spot information, and pushes customized scenic spot information of the tourists by utilizing the obtained actual preference data;

the tourist personal communication end is used for acquiring expected preference data of the tourist and interacting with the customized scenic spot information and the recommended route pushed by the cloud processing module.

Compared with the prior art, the application has the following beneficial effects:

1. personalized navigation service: and generating a plurality of recommended routes by combining expected preferences of tourists and real-time position information, and updating the recommended routes on a two-dimensional or three-dimensional navigation map in real time. Not only improves the accuracy of navigation, but also increases personalized navigation service and enhances the user experience.

2. Pushing real-time information: the cloud server pushes relevant scenic spot information in real time according to the movement track of the tourist, so that the relevance and timeliness of the information are improved. Meanwhile, by analyzing interaction and stay time of the tourist and the push information, the system can more accurately know and respond to actual preference of the tourist.

3. Improve guide efficiency and experience: through intelligent tour guide scheme, tourist can plan the tour route more high-efficient, reduces waiting and crowded time to enjoy more smooth, pleasant tour experience.

Drawings

For a more visual description of the prior art and the present application, exemplary drawings are presented below.

Fig. 1 is a flowchart of a scenic spot intelligent navigation method according to embodiment 1 of the present application;

FIG. 2 is a flow chart of a scenic spot intelligent navigation method according to embodiment 2 of the present application;

FIG. 3 is a flow chart of a background variance method of embodiment 3 of the present application;

fig. 4 is a flow chart of the connected region marking in embodiment 3 of the present application;

fig. 5 is a flowchart of a tracking algorithm in embodiment 3 of the present application.

Detailed Description

The present application is further described in detail below with reference to the accompanying drawings.

In the description of the present application: unless otherwise indicated, the meaning of "a plurality" is two or more. The terms "first," "second," "third," and the like in this application are intended to distinguish between the referenced objects without a particular meaning in terms of technical connotation (e.g., should not be construed as emphasis on the degree of importance or order, etc.). The expressions "comprising", "including", "having", etc. also mean "not limited to" (certain units, components, materials, steps, etc.).

First aspect:

example 1

As shown in fig. 1, the intelligent scenic spot guiding method based on positioning comprises the following steps:

Example 2

As shown in fig. 2, the intelligent scenic spot guiding method based on positioning comprises the following steps:

the method comprises the steps of obtaining expected preference data of tourists and real-time distribution data of scenic spot personnel, storing and counting the real-time distribution data in a cloud server, carrying out composite calculation on the expected preference data and the real-time distribution data after weighting, and generating a plurality of recommended routes by taking the current position of the tourists as a starting point;

feeding back a plurality of recommended routes to tourists by calling a remote service and a long-link interface and generating a two-dimensional/three-dimensional navigation map;

The method comprises the steps that the position information of a tourist on a guiding map is updated in real time by a cloud server, when the tourist approaches a certain scenic spot on the guiding map, the scenic spot information is automatically pushed to the tourist, the cloud server records and marks the passing point positions of the tourist, the actual preference data of the tourist is judged according to the stay time of the tourist on the point positions and the interaction of the tourist and the pushed scenic spot information, and the obtained actual preference data is utilized to customize the scenic spot information and the pushing of a recommended route of the tourist.

And judging the saturation degree of the local people stream according to the real-time distribution data, and sending early warning information to scenic spot staff for manual dispersion when the local aggregate people stream is larger than a preset third threshold value.

Example 3

The method for acquiring the real-time distribution data of scenic spot personnel comprises the following steps:

the method comprises the steps of collecting real-time video data of all places of a scenic spot, wherein in a scenic spot video monitoring system, a dynamic detection technology is a basic link, and provides support for subsequent target analysis, and the core of the dynamic detection technology is to identify a moving human body from video frames. The method and the device mainly carry out people flow statistics in scenic spots, detect and track human bodies in static scenes and finish counting. As shown in fig. 3, the main steps are: image preprocessing, background modeling, background updating, moving human body detection, tracking and counting; and detecting the motion area by using a background subtraction method, performing target identification to obtain a human body target (target connected domain), and tracking the human body target.

Performing image analysis processing on the acquired video data to obtain a video image sequence;

and extracting a motion foreground object of each tourist of each frame of image from the video image sequence by a background subtraction method, performing motion detection on the motion foreground object to obtain a target connected domain, and continuously and uninterruptedly tracking the target connected domain.

And after the current image and the background are differentiated, the method for extracting the moving object by thresholding the obtained differential image is a background subtraction method. If the background image at the time t is f _b (t) the current frame image is fc (t), then the difference formula of the current image and the background is:

wherein T is a threshold value, and if the gray level difference value of a certain pixel in the current image and the background is greater than the threshold value T, the value in the binarized image is 1, and the pixel belongs to a foreground area. Otherwise, the background is obtained, and the corresponding pixel value is 0.

The background difference method consists in creating an initial background (i.e. background model) by using successive N frames of images of motionless foreground objects, the background model requiring periodic updating and maintenance. The background difference method comprises the steps of constructing, updating, differentiating and subsequent processing of a background model. The background model is revised and updated by using the current image, the latest change of the environment is reflected, the background differentiation is that the current image is compared with the background model to generate a differentiated image containing a moving target, the subsequent processing involves denoising the differentiated image and adjusting the target so as to identify the human body target.

The initialization of the background model is accomplished using statistical analysis, which involves counting the most frequently occurring color values of each pixel to form corresponding background model pixel values. In view of environmental changes, the periodic update of the background is divided into three cases: if the illumination change in the video is slower, a statistical and average method is combined to adjust the background; when an object in a scene suddenly moves or stops, only the change area is updated by using a local updating method, so that the updating speed is improved; if the background changes suddenly, the background is directly reinitialized to ensure the instantaneity and effectiveness of the background, the image sequence is subjected to differential binarization by a maximum inter-class variance segmentation method to generate a binarized foreground image, and finally, the required human body target (target connected domain) is extracted by a connected region marking and merging technology.

The acquired video images are all noisy images, the influence of noise on later processing is large, and the images need to be preprocessed to remove noise interference before high-level image processing is performed.

The median filtering is a nonlinear signal processing method, and after pixels in adjacent areas are ordered according to gray level, the intermediate value is taken as an output value. When the group of values has an odd number of pixels, the value in the middle is the middle value; and when there are an even number of pixels, the average value of the gray scale of the middle two pixels is used as the intermediate value.

Wherein f (x, y) represents an original image, g (x, y) is a processed image, and W is a two-dimensional template.

As shown in fig. 4, the connected region labels divide foreground points into some equivalent classes according to the connected relationship, respectively identify pixel points of different equivalent classes, and classify regions with areas smaller than a preset value as backgrounds, where the reserved foreground points are required targets. The method comprises the steps of scanning a binary image from the upper left to the right and from top to bottom by using a pixel marking method as a mark of a connected region, marking whether a current pixel is connected with a neighboring pixel scanned previously or not by the current pixel, scanning a next pixel if the value of the current pixel is 0, marking the current pixel as a mark value of a pixel connected with the current pixel if the value of the current pixel is 1, and if the current pixel is connected with two or more targets, the two or more targets can be regarded as one and combining the targets. When it is checked that a transition from a pixel with a value of 0 to a separate pixel with a value of 1 is made, it is given a new marking value.

The background model establishment and the differential threshold selection cannot be completely accurate, the human body target obtained by the background subtraction method may be broken, and one human body target is divided into a plurality of communicating bodies through the communicating body marks, so that the communicating bodies are combined to obtain the complete human body target. And storing the marked communicating bodies in a linked list.

The steps of merging the communicating bodies are as follows:

A. and ordering the coordinates of the communicating bodies stored in the linked list from left to right and from top to bottom, so that the originally adjacent communicating bodies are adjacent in the linked list.

B. Defining a new linked list, storing the first connected body in the old linked list into the new linked list, and correspondingly deleting the first connected body in the old linked list.

C. And taking out a connected body from the old linked list, and checking whether the connected body and the element stored in the new linked list have overlapping or adjacent. If there is overlap or neighbor, then merge the two communicating bodies and store them in the new linked list; if not, it is stored directly in the new linked list and this link is deleted in the old linked list.

D. And C, judging whether the old linked list is empty, if not, repeating the step C, and if not, ending.

By the above processing, the broken human body targets after background subtraction are combined, and the smaller block-shaped human body targets are not filtered out as noise in the subsequent filtering processing.

The non-target connected domain is removed, and after the labeling and the combination of the connected bodies, a binarized foreground image is obtained, but the foreground region may not only have the target connected domain, but also include other moving objects, such as animals in the scenic region, other sundries, or large-area noise caused by leaf swing and the like, and the connected bodies generated by the noise in the foreground are removed by using a scale filter, so that the size of the connected bodies is much smaller than that of a human body target. The size of the human body has a definite range. The size of the human body target is obtained by counting the number of pixels contained in the connected region, and the range of the human body size is obtained by using priori knowledge. And filtering out larger or smaller communicating bodies in the background through scale comparison, and keeping a communicating region with the size similar to that of a human body, and then, continuing analysis and treatment. Since the scenic spot personnel are in an upright state, the obvious characteristic of the human body at this time, which is different from other moving objects, is the ratio of the height to the width. And obtaining the height-width ratio of the human body by using the vertical and horizontal projection histograms of the binary images. In the practical application process, the ratio of the height to the width of the upright human body is 1.4-5, and the area with the calculated height-width ratio within the range is the human body, otherwise, the area is used as noise filtering. In practical scenic spot applications, the height to width ratio is properly relaxed so as not to miss the human body.

The method for continuously and uninterruptedly tracking the track of the target connected domain comprises the following steps:

dividing the video image sequence into a current frame image and a plurality of candidate frame images, acquiring the positioning of a target connected domain in the current frame image, and matching the target connected domain in the plurality of candidate frame images through a tracking algorithm.

As shown in FIG. 5, the method and the device select the Mean Shift algorithm to track the target connected domain, and solve the problems caused by the fact that the blocking and the target move too fast by adopting the Mean Shift tracking algorithm combined with Kalman prediction aiming at the situation that the moving target is easy to lose when the blocking and the moving target speed are too fast.

Mean Shift tracking algorithm incorporating Kalman filter: in the tracking process, the tracking target is tracked and refreshed at fixed time, and tracking is not required to be performed every frame. In the implementation process, the effect of every 5 frames is good, the movement speed of some tourists is high, and a Kalman filter is introduced to predict a moving target so as to avoid target loss. The Kalman filter has smaller operation amount and accurate prediction, the Kalman filter is introduced to improve the performance of a tracking algorithm, the Kalman filter is an optimal filter for estimating the minimum mean square error of a sequence in a dynamic state, and the current and future values of the signal are estimated according to the time-varying statistical characteristics of the signal.

The Kalman filter consists of two parts: a time update equation and a measurement update equation, the time update equation is regarded as an estimation equation by estimating a motion state with feedback control, and the measurement update equation is a correction equation. The whole Kalman filtering algorithm is an estimation correction algorithm, and the time update equation constructs the next time state, and timely calculates the current state variable and the value of the estimation error covariance. The measurement update equation utilizes the new variable and the prior estimation to construct an improved posterior estimation, which plays a feedback role. The method is applied to the actual scenic spot, mainly predicts the central point (x, y) of the target connected domain, firstly decomposes the central coordinate in the x and y directions to obtain two corresponding components with an included angle of 90 degrees, predicts the components respectively, and finally obtains the tourist target motion state which is the result of the synthesis of the x and y components.

Measuring the matching result by using a similarity function, and obtaining candidate frame images containing the target connected domain from a plurality of candidate frame images when the measuring result is in a preset second threshold interval;

and marking the successfully matched candidate frame images according to the sequence of the frame images, so as to realize continuous and uninterrupted track tracking of the target connected domain.

Track tracking data of a target connected domain in a scenic spot in a video image sequence is real-time distribution data;

the method for generating and updating the plurality of recommended routes comprises the following steps:

collecting expected preference data of tourists in an interactive questionnaire, historical access data or keyword label classification mode, classifying the expected preference data to form a data set, correspondingly distributing a weight value to each element in the data set, and respectively summing the weight value and the corresponding element to obtain a weighted overall preference score;

the weighting process expects preference data, the weighting method is: assuming that each guest has a series of preferences,

wherein p is _i Indicating the degree of preference for a certain category of attractions,

a weight w is assigned to each preference _i This weight may be based on historical data,A guest survey or database stores information.

The weighted preference score may be expressed as:

if a tourist has a preference of 8 (full score 10) for cultural attractions, a natural landscape of 6 and an entertainment activity of 4, the preference data can be expressed as: p= {8,6,4}

Assuming the corresponding weights are 0.5, 0.3 and 0.2, respectively, the weighted preference scores are

S=0.5×8+0.3×6+0.2×4=6.2

And calculating the overall preference score and the real-time distribution data through a mean value clustering algorithm, wherein each cluster represents a recommended route by taking the current position of the tourist as a starting point, so as to generate a plurality of recommended routes.

and acquiring the people stream information of each area in the scenic spot at the next moment according to the estimated position information of all the target connected areas in the scenic spot, screening and sorting a plurality of recommended routes according to the people stream information of each area at the next moment, and updating the recommended routes according to the screening and sorting results.

The ARIMA model (autoregressive integral moving average algorithm model) is a time series prediction method suitable for predicting non-seasonal time-based data, and ARIMA can be used to predict changes in people stream over a short period of time.

The ARIMA model consists of three main components:

autoregressive (AR): this part of the model represents the relation between the current value and several historical values preceding it.

Difference (I): this section is used to smooth out non-smooth time series data, typically by first or multiple order differencing.

Running average (MA): this part represents the relationship between the errors of the current value and the historical value.

Model parameters, ARIMA model is generally denoted ARIMA (p, d, q), where p represents the order of the autoregressive term, i.e., the number of historic values; where d represents the differential order, i.e., the number of differential times required for sequence stabilization; where q represents the order of the running average term, i.e., the number of running average terms of the prediction error.

Using ADF (Augmented Dickey-Fuller) test to test the stability of time sequence, if the data is not stable, performing first-order or multi-order difference until the data becomes stable; an autocorrelation function (ACF) and a partial autocorrelation function (PACF) map are used to determine optimal values for parameters p and q; determining optimal p, d and q values by a cross-validation or red pool information criterion (AIC) method and the like; fitting an ARIMA model to the historical data using the selected parameters; checking the independence and normal distribution of residual errors, and ensuring the effectiveness of a model; predicting the flow of people in a short period by using the fitted ARIMA model; the prediction accuracy of the model is evaluated using a Mean Square Error (MSE) or other statistical indicator.

And (3) assuming that the historical real-time distribution data show that the people flow of the scenic spot has obvious daily change but has no obvious seasonal change, collecting the track data of the target connected domain, wherein the data comprise the position coordinates of each target at different time points, and converting the track data into a time sequence format, namely, the position coordinates of the targets corresponding to each time point.

The ARIMA model can be used for predicting the next people flow per hour, firstly, the stability of the time sequence is confirmed through ADF test, then, the values of p and q are confirmed through ACF and PACF images, then, the model is fitted and checked, finally, people flow prediction in a short period is carried out, and therefore people flow information of each area in a scenic spot at the next moment is obtained.

The method for pushing the nearby scenic spot information for the tourist comprises the following steps:

acquiring real-time position data of tourists in the recommended route by using a GPS, wi-Fi positioning, bluetooth beacons or Beidou positioning system; the real-time position data are sent to a cloud server, and the cloud server continuously monitors the relative distance between tourists and all scenic spots;

and calculating through a distance between the current position of the tourist and the coordinates of the scenic spot by using a neighbor analysis algorithm, and triggering pushing when the distance is smaller than a preset first threshold value.

The neighbor analysis algorithm includes:

K-NN nearest neighbor algorithm, in the classification task, the algorithm searches the nearest K neighbor points, and decides the category of the unknown point according to the majority voting principle, in the regression task, the algorithm predicts the value of the unknown point by calculating the average value of the K nearest neighbors;

selecting a value of K, determining the number of nearest neighbors for searching, and determining the optimal value of K through cross verification;

for a given unknown point, calculating the distance between the unknown point and all known points, wherein the distance measurement method comprises Euclidean distance, manhattan distance or cosine similarity;

finding the nearest K neighbors, and finding the nearest K neighbors according to the calculated distance;

For classification problems, the classification of the unknown points is predicted based on the most common classification in the nearest neighbor, and for regression problems, the prediction is based on the average value of the nearest neighbor numerical attributes.

Classification tasks the presence Jing Zhongzhi classifies attractions according to their historical preferences or behaviors to recommend to the guest the most likely type of attraction that the guest would like, assuming there are different types of attractions such as historic buildings, natural parks, museums, etc., if a guest frequently visited historic buildings and museums in the past, the K-NN algorithm may be used to identify other attractions similar to those attractions and recommend to the guest, in which case the algorithm recommends new attractions according to the rules of the attraction type majority of the guest's preferences by looking for attractions similar to the known preferences (nearest K neighbors).

The regression task involves predicting the score or satisfaction of the guest with respect to the attraction, thereby recommending attraction with a high score, and if the history data contains scores of guests with respect to different attractions, the K-NN algorithm may be used to predict the potential score of guests with respect to non-visited attractions, by finding K scored attractions similar to the target attraction, and calculating the average of these attraction scores, the algorithm predicts the potential score of guests with respect to the attraction, based on which the system may recommend those attractions with a high expected score.

The K-NN algorithm is used to recommend points of attraction to the guest that should be visited next, a specific example:

the data set, mobile phone application or applet has a data set containing a plurality of attractions, each attraction having its geographic location (latitude and longitude) and some attributes such as attraction type (cultural, natural, recreational), tourist score.

Tourist preferences, a tourist marks in the application several types of attractions she likes and gives a score of the attraction visited before.

At the current location, the guest is currently located at a sight spot.

The steps are as follows, the K value is selected: setting a value of K, such as 5, means that the system will consider the nearest 5 attractions.

Calculating the distance: the Euclidean distance is used to calculate the distance of the tourist's current location to all other attractions.

Finding the nearest K scenery spots: based on the distance, the nearest 5 attractions are selected.

Consider guest preferences: from the 5 attractions, attractions conforming to the tourist preference are further screened. For example, if a guest prefers cultural attractions, the system will prioritize such attractions.

Recommending scenic spots: according to the steps, the system recommends 1-2 scenic spots meeting the conditions to tourists.

Suppose that the guest is currently at coordinates (30.2672, -97.7431), K is set to 5. The system calculates the following 5 nearest attractions and their types:

Scenic spot A: (30.2700-97.7500), culture

Scenic spot B: (30.2650, -97.7400), natural

Scenic spot C: (30.2600, -97.7350) amusement

Scenic spot D: (30.2550-97.7500), culture

Scenic spot E: (30.2500, -97.7450), natural

If the guest prefers cultural types, the system will recommend attractions A and D.

In practical applications, other factors, such as current weather, consumption price of scenic spots, etc., may also be considered to provide more accurate recommendations.

The scenic spot information pushing method for customizing tourists comprises the following steps:

the method comprises the steps of tracking position information of tourists in real time by using a positioning technology, and automatically recording the position information of a point position, the stay time of the tourists at the point position and interaction data of the tourists and the scenic spot information when the tourists pass through or stay at the scenic spot, wherein the interaction data comprise click rate and interaction time of the scenic spot information;

and analyzing the stay time of the scenic spot and the interaction data of tourists and the scenic spot information by using a cluster analysis algorithm or a decision tree algorithm to obtain the actual preference data, screening and sequencing the scenic spot information by applying a collaborative filtering algorithm or a content-based recommendation algorithm according to the actual preference data, generating customized scenic spot information and pushing the customized scenic spot information.

And the tourists provide feedback for the recommended content, wherein the feedback comprises that the evaluation results of a plurality of dimensions are obtained through liking, disliking or scoring, and the recommendation algorithm is continuously adjusted and optimized according to the feedback results of the tourists.

The clustering algorithm is based on the principle that data points are grouped, so that points in the same group are similar to each other, and points in different groups are dissimilar, and the common clustering algorithm is K-means, and the working principle is as follows: m points are randomly selected as initial cluster centers first, and then assigned to the nearest cluster center according to the distance from each point to these centers. And then, re-calculating the center of each cluster (namely, the average value of all points in the clusters), repeating the process until the cluster center is not changed significantly, taking weighted tourist preference data and real-time distribution data of scenic spots as input, and clustering the scenic spots by using a K-means or other clustering algorithm. In view of the real-time people stream data and tourist preferences, for example, for tourists preferring cultural attractions, the algorithm would tend to select cultural attractions with less people stream, and eventually each cluster represents a recommended route including attractions preferred according to the tourist preferences and the scenic spot people stream conditions.

This can be achieved by various programming languages and data science toolkits, such as the Scikit-learn library of Python. By the method, the personal preference of tourists and the real-time people stream condition of scenic spots can be effectively combined to generate personalized recommended routes.

Collaborative filtering is an algorithm that makes recommendations based on user behavior data (e.g., scoring, browsing, purchasing, etc.). It is assumed that if two users have similar ratings in the past for certain items (e.g., sights), their preferences may also be similar for non-rated items.

Example 1, user-item interaction matrix: constructing a matrix to record the scoring or interaction of the user with the object (scenic spot); similarity calculation: calculating the similarity between users or between objects, and using cosine similarity or pearson correlation coefficient; the likely score of the target user for the unrated attraction is predicted based on the scores of similar users.

Assuming that the scores of user a and user B are very close over multiple commonly rated attractions, user a gives a high score for a attraction, but user B has not yet visited, the system predicts that user B will also like this attraction and takes it as a recommendation.

Content-based recommendations recommend similar items based on characteristics of items that the user liked in the past, which method relies on descriptive characteristics of the items, such as type of attraction, subject matter, location, etc.

Example 2, item characteristics representation: creating a feature vector for each attraction, representing its attributes (e.g., history, nature, entertainment, etc.); user preference model: establishing a user preference model according to past behaviors of the user (such as scoring scenic spots or stay time); recommendation generation: and calculating recommendation scores according to the user preference model and the scenic spot characteristics, and selecting the scenic spot with the highest score for recommendation.

User C frequently accesses and highly evaluates historical museums, the system analyzes the user C's preference model, finds that it has a high interest in the historical sight-like spots, and recommends to user C some historical museums that have not been accessed but have similar characteristics.

In practice, collaborative filtering and content-based recommendation are often used in combination to remedy the respective deficiencies. For example, collaborative filtering may be affected by cold start problems (making it difficult to generate recommendations for new users or new attractions), while content-based recommendations may result in insufficient diversity of recommendations, which in combination may provide more accurate and diversified recommendations.

Remote services, referring to cloud-based services, such as Application Program Interfaces (APIs) running on cloud computing platforms, that are capable of handling complex data analysis and route generation tasks; compared with the traditional short link, the long link can continuously maintain the connection state, and real-time data transmission is allowed. In the intelligent navigation system, the intelligent navigation system can be used for continuously pushing updated information to tourists, and the server side firstly carries out fusion processing on preference data and real-time people stream data of the tourists to generate a recommended route. The system needs to be able to dynamically adjust the recommended route according to real-time data (such as the people stream condition of the scenic spots); generating a map using Geographic Information System (GIS) software or a specialized mapping tool; and integrating the generated recommended route into a two-dimensional or three-dimensional map, wherein the three-dimensional map provides more visual and immersive experience for tourists.

The map and the route are displayed through mobile applications or applets, the platforms should be provided with a compact interface, so that tourists can easily understand and follow the route, and the user interface should be capable of displaying the position and recommended route of the tourists and scenic spot information in real time.

And integrating the map generation software and the data processing algorithm with a mobile application or applet platform through an API (application program interface), and realizing real-time communication between a server and a client by using WebSockets or other long-link technologies.

The security of data transmission is ensured, the performance of a background algorithm and a server is optimized by adopting an encryption and security authentication mechanism, and the quick response and high-efficiency data processing are ensured.

The intelligent navigation system can efficiently process data, generate personalized recommended routes and transmit the information to tourists in real time through a concise interface.

The method for acquiring the expected preference data of the tourist comprises the step of acquiring the expected preference data of the tourist through an online questionnaire, an online interactive tour game or historical access data.

The method also comprises the following steps: scenic spot staff updates scenic spot information on the cloud server, wherein the scenic spot information comprises scenic spot introduction, surrounding delicacies, personal characteristics, historical backgrounds and scenic spot activities.

Second aspect:

the intelligent scenic spot guiding system based on positioning comprises a video acquisition module, a cloud processing module, a tourist personal communication end and a scenic spot personnel receiving end;

the video acquisition module is used for acquiring real-time distribution data of scenic spot personnel, and comprises face recognition cameras and infrared people stream monitoring cameras which are arranged at all points of a scenic spot.

The cloud processing module stores and counts the real-time distribution data in a cloud server, performs composite calculation on the weighted expected preference data and the real-time distribution data, and generates a plurality of recommended routes by taking the current position of tourists as a starting point; feeding back a plurality of recommended routes to the tourist personal communication end by calling a remote service and a long-link interface and generating a two-dimensional/three-dimensional navigation map; the method comprises the steps that a cloud server updates position information of tourists on a navigation map in real time, when the tourists approach a certain scenic spot on the navigation map, the scenic spot information is automatically pushed to the tourists, the cloud server records and marks the passing point positions of the tourists, actual preference data of the tourists are obtained through stay time of each scenic spot of the tourists in the tour process and interaction of the tourists and the pushed scenic spot information, and customized scenic spot information is pushed to the tourists by utilizing the obtained actual preference data; judging the saturation degree of local people flow by the real-time distribution data, and sending early warning information to scenic spot staff for manual dispersion when the local aggregate people flow is larger than a preset third threshold value;

The tourist personal communication terminal is used for acquiring expected preference data of the tourist and interacting with the customized scenic spot information and the recommended route pushed by the cloud processing module; the scenic spot personnel receiving end is used for receiving the early warning information sent by the cloud processing module.

The system utilizes cloud technology, can realize remote storage and processing of data, and improves reliability and expandability of the system. Meanwhile, the cloud service also enables scenic spots to acquire and analyze data in real time, so that the scenic spots are better managed and operated; the stored data mainly comprises scenic spot longitude and latitude points, point position introduction (pictures, voices, characters, videos and the like), tourist real-time longitude and latitude and routes, conventional routes, intelligent generated routes, tourist personal preference and the like; the system supports a voice interaction function, tourists can initiate inquiry to the system through voice, and the system replies an inquiry result in a voice or text form; the system supports social interactions between guests, such as punching cards, comments, etc. at specific attractions, which may also be part of the data analysis to provide more personalized navigation services.

The system supports an electronic payment function, and tourists can directly purchase related goods and services of scenic spots such as tickets, souvenirs and the like in the system; when the system processes tourist data, related laws and regulations of data security and privacy protection are strictly complied with, and all data are encrypted in the processes of collection, storage and processing, so that the security and privacy of the tourist data are ensured.

Technical implementation of the present application, using front-end foundations such as HTML, CSS, javaScript, managing user interfaces and interactions using a Vue front-end framework, storing data using java back-end technology, and databases such as Redis, mongoDB, mySQL, in combination with interactions and functional implementation using various Web APIs, will comply with relevant data security and privacy regulations when designed and implemented.

Compared with the prior art, the system has the following beneficial effects:

1. the intelligent management and operation of scenic spots are facilitated, and the operation efficiency and service quality of the scenic spots are improved without charging, maintaining, sterilizing and the like for explanation equipment.

2. The system is convenient for users to use, and can explain by only opening the applet for paying without queuing leasing equipment.

3. The tourist tour guide system can provide more convenient and personalized tour guide service and improve the tour experience of tourists.

4. All data are encrypted in the processes of collection, storage and processing, so that the safety and privacy of tourist data are ensured.

5. The tool is easy to use, and by means of visual graphical interfaces and guiding designs, non-professionals can quickly use the tool.

6. The continuous updating and optimizing device has good expandability and maintainability, and can continuously update and optimize the tool according to user feedback and market demands, thereby providing more functions and better use experience.

7. The system can provide relevant audio-visual contents according to the specific positions of tourists, enrich the guiding experience of the tourists and improve the service quality of scenic spots.

Any combination of the technical features of the above embodiments may be performed (as long as there is no contradiction between the combination of the technical features), and for brevity of description, all of the possible combinations of the technical features of the above embodiments are not described; these examples, which are not explicitly written, should also be considered as being within the scope of the present description.

The foregoing has outlined and detailed description of the present application in terms of the general description and embodiments. It should be appreciated that numerous conventional modifications and further innovations may be made to these specific embodiments, based on the technical concepts of the present application; but such conventional modifications and further innovations may be made without departing from the technical spirit of the present application, and such conventional modifications and further innovations are also intended to fall within the scope of the claims of the present application.

Claims

1. An intelligent scenic spot guiding method based on positioning is characterized by comprising the following steps:

2. The location-based attraction intelligent navigation method of claim 1, further comprising:

3. The intelligent navigation method of scenic spot based on positioning of claim 2, wherein the processing of the real-time distribution data to obtain the people stream information of each area in the scenic spot at the next moment, and updating the recommended route for the tourist according to the people stream information of each area at the next moment, comprises:

4. The intelligent navigation method of location-based scenic spot according to claim 2, wherein the continuous and uninterrupted track tracking of the target connected domain comprises:

5. The location-based attraction intelligent navigation method of claim 1, wherein the generating a plurality of recommended routes using expected preference data and real-time distribution data comprises:

6. The intelligent navigation method of scenic spot based on positioning according to claim 1, wherein the steps of obtaining the position information of tourists in the recommended route and pushing nearby scenic spot information according to the position information include:

7. The location-based attraction intelligent navigation method of claim 1, wherein the obtaining the actual preference data of the tourist through the stay time of each attraction and the interaction between the tourist and the pushed attraction information during the attraction, and the pushing of the customized attraction information of the tourist by using the obtained actual preference data comprises the following steps:

8. The location-based attraction intelligent navigation method of claim 1, wherein the means for obtaining the expected preference data of the guest comprises obtaining the expected preference data of the guest via an online questionnaire, an online interactive tour game, or historical access data.

9. The location-based attraction intelligent navigation method of claim 1, further comprising: scenic spot staff updates the scenic spot information, including scenic spot profile, surrounding food, personal features, historical background, and scenic spot activity.

10. The intelligent scenic spot guiding system based on positioning is characterized by comprising a video acquisition module, a cloud processing module and a tourist personal communication end;