Intelligent scenic spot tourism planning system

Abstract

The invention discloses an intelligent scenic spot tourism planning system, and relates to the technical field of smart tourism, and the system comprises a mobile terminal data collection layer, an edge node data processing layer, a bidirectional feedback planning layer and a scenic spot resource adaptation layer which are in low-power-consumption real-time interaction through a 5G narrowband Internet of Things in sequence. According to the invention, the mobile terminal data acquisition layer accurately divides unexpected stay types to avoid experience interruption, the edge node data processing layer deduces local wind speed and other hidden environment factors and establishes factor-demand association, and the bidirectional feedback planning layer generates a dynamic personalized route of core-transition-core. The scenic spot resource adaptation layer activates hidden resources and realizes hierarchical people flow regulation and control according to tourist preferences, and meanwhile, depends on APP real-time privacy data processing, edge node encryption storage and 5G narrowband Internet of Things low-power-consumption interaction, so that the defects that an existing system misjudges stay value, lacks hidden factor association and the like are overcome, privacy safety and user experience are considered, and the system is suitable for popularization and application. And stable operation of the system is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of smart tourism technology, specifically to an intelligent scenic area tourism planning system. Background Technology

[0002] With the development of smart tourism, intelligent scenic area tourism planning systems have become a core tool for improving the service quality of scenic areas. Their main functions include generating preset tour routes based on tourist profiles, real-time navigation, attraction introductions, and crowd flow warnings. Related technologies have been implemented in most 4A and 5A scenic areas. Currently, the core design logic of existing intelligent scenic area tourism planning systems is efficiency-oriented, aiming to achieve the goal of visiting the most core attractions in the shortest time by optimizing the spatial distance of routes and the order of attraction visits. This only addresses how to enable tourists to complete their tours more efficiently, without addressing the core issue of how to match the tour experience with the implicit environmental needs of tourists.

[0003] However, existing systems have a fundamental flaw in practical applications: they cannot establish a logical connection between unexpected stay behaviors, implicit environmental factors, and personalized planning. This leads to a severe mismatch between efficiency-oriented and visitor experience-oriented needs. This flaw is specifically manifested in three interrelated levels, ultimately constraining the system's ability to adapt to different experiences: 1. Misjudging the value of unintended stays and failing to identify experience needs: The system defines non-urgent stays that deviate from the route as planning deviations and only prompts tourists to return or shorten subsequent routes to make corrections. It fails to recognize that this is a proactive choice by tourists to have a better local environment experience, and instead regards it as an efficiency interference, resulting in a gap in the identification of experience needs.

[0004] 2. Lack of monitoring and correlation of implicit environmental factors makes it difficult to meet the needs: Only the macro-area environment of the scenic area is monitored, but implicit factors such as temperature difference and wind speed in the micro-space are not measured. The occasional data also lacks factor-profile-planning correlation.

[0005] 3. Limited personalization: The plans are based solely on age, physical strength, and interests, without taking into account environmental preferences such as being sensitive to heat or preferring a gentle breeze, resulting in homogenized plans.

[0006] In summary, the shortcomings of existing systems all stem from the inability to establish a logical connection between unexpected stays, implicit environmental factors, and personalized planning. This limits their efficiency-oriented approach, making it difficult to identify tourists' implicit needs and activate the implicit environmental resources of scenic areas. A technical solution is urgently needed to overcome this logical gap and achieve an upgrade from efficiency-first to experience-driven approaches. Therefore, this paper proposes an intelligent scenic area tourism planning system to overcome the aforementioned problems. Summary of the Invention

[0007] The purpose of this invention is to provide an intelligent scenic area tourism planning system to solve the problems mentioned in the background art.

[0008] To solve the above-mentioned technical problems, the present invention provides an intelligent scenic area tourism planning system, which includes a mobile terminal data acquisition layer, an edge node data processing layer, a bidirectional feedback planning layer, and a scenic area resource adaptation layer that sequentially interacts with 5G narrowband Internet of Things low power consumption in real time. The mobile terminal data acquisition layer includes a multi-dimensional behavioral data acquisition module, a biometric auxiliary judgment submodule, and a stay behavior validity judgment submodule, which are used to convert tourists' unexpected stay behaviors into demand signal packets containing experience intensity markers; The edge node data processing layer includes a mobile sensor node multiplexing module and a behavior-environment association mapping module, which are used to deduce implicit environmental factors such as temperature difference, wind speed, and light comfort in the micro-space based on demand signal packets, and generate tourist environmental preference labels. The two-way feedback planning layer includes a personalized route generation module and a real-time feedback iteration module, which are used to dynamically plan tour routes based on preference tags and adjust routes according to real-time tourist behavior data. The scenic area resource adaptation layer includes a hidden resource map generation module and a pedestrian flow-environment collaborative regulation module, which are used to activate the hidden resources of the scenic area and realize hierarchical pedestrian flow regulation based on tourist tags.

[0009] Furthermore, the multi-dimensional behavioral data acquisition module includes a touch screen pressure acquisition unit, a screen interaction frequency determination unit, and a data fusion unit; The touch screen pressure acquisition unit records the touch screen pressure value once every 500ms by calling the TouchEvent API through the mobile APP. After continuous acquisition for 1 minute, a pressure dataset is generated. Combined with the lightweight acquisition piles with infrared distance sensors deployed in the scenic area, the pressure fluctuation value is calculated using a sliding window algorithm with a window size of 10 data points. When the fluctuation value is <0.2N and a stop assistance signal is received, it is marked as a deep experience grip state. When the fluctuation value is >0.5N, it is marked as a shallow observation grip state. The screen interaction frequency determination unit identifies the application package name operated by the user after unlocking by calling the ActivityManager API of the APP, and counts the number of unlocks and operation types within 10 minutes. When unlocking ≤1 time and the operation type is viewing environmental photos, it is marked as a deep experience interaction state. When unlocking >2 times / 5 minutes and the operation type is browsing social software, it is marked as a shallow observation interaction state. The data fusion unit aggregates touch screen pressure data and screen interaction data to provide basic behavioral data for demand signal packets.

[0010] Furthermore, the biometric-assisted judgment submodule includes a facial micro-expression processing unit and a touchscreen operation stability detection unit; The facial micro-expression processing unit uses the APP to call the front camera to capture one frame every 30 seconds. It automatically crops the eye-mouth area corresponding to the preset coordinate box (100,80)-(160,140) based on the 16:9 screen ratio. The cropped area is then processed in grayscale and Canny edge detection with a threshold of 50-150. The deviation of the eye contour center point and the fluctuation of the mouth corner arc are calculated for three consecutive frames. If both conditions are met and the lightweight acquisition pile detects that the user stays for more than 3 minutes, a depth visual experience mark is output. Otherwise, a shallow observation mark is output. The touch screen operation stability detection unit collects touch screen pressure values ​​every 100ms and records the timestamp by calling the MotionEvent API through the APP, calculates the pressure fluctuation range within 2 consecutive minutes, and counts the duration of a single touch. If the pressure is stable and the contact is continuous, it outputs a deep experience stable operation flag; otherwise, it outputs a temporary operation flag.

[0011] Furthermore, the dwell behavior validity determination submodule classifies dwell behavior through four-dimensional logical judgment: Step 1, Location Determination: The APP obtains the phone's GPS coordinates and receives the ID signal of nearby lightweight data collection points. If the coordinates of the two are less than 10 meters and last for more than 3 minutes, the location is determined to be stationary for more than 3 minutes; otherwise, it is considered a temporary stop. Step 2, grip posture determination: Call the pressure fluctuation value output by the touch screen pressure acquisition unit. If it is <0.3N, proceed to the next step; otherwise, pause temporarily. Step 3, Ambient Sound Determination: The APP calls the microphone to collect ambient sound. It uses a lightweight FFT algorithm that consumes less than 10% of the CPU to separate natural sound from 200-5000Hz and artificial sound from 50-200Hz. If the proportion of natural sound is greater than 50%, proceed to the next step; otherwise, pause temporarily. Step 4, Biometrics Judgment: If the biometrics-assisted judgment submodule outputs a high mark for depth visual experience + depth experience stable operation, it determines that the depth experience is valid and generates a high-priority demand signal packet; if it outputs a low mark for shallow observation + temporary operation, it determines that the shallow experience is valid and generates a medium-priority demand signal packet; when the mark is mixed, if the pressure fluctuation value is <0.2N, it is judged as shallow valid stay, otherwise it is temporary stay.

[0012] Furthermore, the mobile sensor node multiplexing module derives implicit environmental factors in the following way: (1) Derivation of local wind speed: One wind speed sensor node is deployed every 50 meters in the scenic area. The edge node receives the dwell position in the demand signal packet and matches the wind speed data of the nearby sensor node. When the demand signal packet is marked as high, the wind speed is retained to one decimal place and the wind adaptability intensity is marked as excellent. When it is marked as low, the wind speed is retained to an integer place and the wind adaptability intensity is marked as average. (2) Derivation of light comfort: The scenic area deploys light sensors to collect real-time light intensity. The edge nodes combine the age range of tourists uploaded by the APP to retrieve the preset age range-light tolerance lower limit table and calculate the comfort score: 4-5 points for light intensity ≤ tolerance lower limit, 2-3 points for light intensity > tolerance lower limit and ≤ 3000 lux, 1 point for light intensity > 3000 lux, and the score is reduced by 1 point when the demand signal packet is marked as low. (3) Derivation of micro temperature difference: SHT30 temperature and humidity sensors are deployed in the scenic area to collect real-time temperature and calculate the temperature difference with the surrounding area. The edge nodes are combined with the dwell time uploaded by the APP. If the dwell time is >5 minutes, the temperature difference data is valid; otherwise, the average value of the surrounding nodes is taken.

[0013] Furthermore, the workflow of the behavior-environment association mapping module is as follows: Step 1: The edge node receives the experience intensity marker of the demand signal packet and the implicit environmental factor data collected by the mobile sensor node multiplexing module. Step 2: Set the weight coefficients for experience intensity labels, with a weight of 1.5 for high labels and a weight of 1.0 for low labels; Step 3: Generate tourist environmental preference labels based on implicit environmental factor data and weighting coefficients; Step 4: The edge node establishes a MySQL database of visitor ID-preference tags. When iterating and updating the tags, the weight of the new tag is 0.7 and the weight of the old tag is 0.3.

[0014] Furthermore, the personalized route generation module plans routes through the following steps: Step 1: Deploy a scenic area node database at the edge nodes, including the locations of core / hidden nodes, real-time environmental data of sensor nodes, and experience intensity types; Step 2: Calculate the experience matching value based on implicit environmental factor data. The calculation logic is consistent with the lighting comfort scoring logic. Step 3: Filter nodes with an experience matching score of ≥70, prioritize core nodes, and sort them in the order of core → transition → core, with a walking time between nodes ≤15 minutes; Step 4: The APP calls the Gaode Map SDK to push the route and simultaneously sends environmental forecast information; The workflow of the real-time feedback iteration module is as follows: Step 1: The app uploads touch screen operation data and micro-expression data every 2 minutes; Step 2: If the edge node detects two consecutive data points that satisfy eye gaze switching > 3 times / minute and touch screen duration < 1 minute, the experience intensity is determined to decrease from high to low. Step 3: Edge nodes search for transitional nodes with a matching experience score of ≥70, push route adjustment suggestions to the APP, and update the route after user confirmation.

[0015] Furthermore, the workflow of the latent resource map generation module is as follows: Step 1: Count the deep dwell frequency and total dwell frequency of hidden nodes at the edge nodes, and calculate the deep dwell percentage. Step 2: Create a heat map using ArcGIS, with the heat map colors being red, yellow, and blue, and then push it to the scenic area management platform; Step 3: Add QR code AR navigation to core nodes and add convenient facilities to transition nodes; The workflow of the pedestrian-environment coordinated control module is as follows: Step 1: Deploy Hikvision DS-2CD3T46WDV3-I3 infrared people counters in key scenic spots, ensuring a people density of >3 people / m². 2 Time-triggered traffic splitting; Step 2: Filter environmental preference tags of surrounding tourists at edge nodes to distinguish between strong and weak preferences; Step 3: The APP pushes diversion routes of core hidden nodes to tourists with strong preferences and diversion routes of transitional nodes to tourists with weak preferences. The scenic area management platform displays the diversion progress in real time.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. By using multi-dimensional behavior collection, biometric-assisted judgment, and four-dimensional validity assessment of stay in the mobile terminal data collection layer, unexpected stays are subdivided into deep / shallow valid stays and temporary stays. This avoids the existing system misjudging them as efficiency interference items and interrupting the experience with pop-up windows, and successfully captures tourists' proactive experience needs for the local environment.

[0017] 2. The edge node data processing layer derives implicit factors such as local wind speed, light comfort, and micro temperature difference by reusing sensor nodes. It establishes factor-demand associations by combining experience intensity labels, and generates tourist environmental preference labels. This solves the shortcomings of existing systems that only monitor the macro environment and do not have factor associations, and allows planning schemes to accurately match tourists' implicit preferences such as being afraid of heat and preferring a light breeze.

[0018] 3. The two-way feedback planning layer uses a scenic area node library containing implicit environmental data to select nodes with high matching values ​​to generate the core → transition → core route. At the same time, the route is adjusted every 2 minutes based on the real-time behavior of tourists, breaking away from the single mode of planning based solely on age and physical strength in the existing system, and realizing a dynamic and personalized route driven by experience.

[0019] 4. The scenic area resource adaptation layer uses in-depth stay data to create heat maps and activate hidden nodes; combined with visitor flow counters and preference tags, it pushes different diversion routes to tourists with strong / weak preferences, solving the problem that the existing system cannot activate hidden resources and divert visitors indiscriminately, thereby improving resource utilization and tour order.

[0020] 5. The APP processes privacy data in real time, encrypts and stores tags at edge nodes and grants hierarchical authorization, and achieves low-power interaction through 5G narrowband IoT, balancing privacy and security with user experience and ensuring stable system operation. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of an intelligent scenic area tourism planning system according to the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figure 1 The present invention provides a technical solution: See Figure 1 As shown, an embodiment of an intelligent scenic area tourism planning system is as follows: I. System Overall Architecture: This system includes a mobile terminal data acquisition layer (APP + scenic area auxiliary acquisition equipment), an edge node data processing layer, a two-way feedback planning layer, and a scenic area resource adaptation layer. The four layers achieve low-power real-time interaction through 5G narrowband Internet of Things (NB-IoT) (the communication protocol adopts NB-IoT R17 version, and the transmission rate is 250kbps-1Mbps).

[0024] The invention aims to achieve the following: By constructing a distributed sensing, behavior detection, and biometric-assisted judgment system using a user's mobile app (calling the system's standard interface) and lightweight data collection points in the scenic area; and by employing a privacy-preserving app data collection, environmental monitoring of scenic area sensor nodes, and dynamic planning of experience intensity, unexpected stay behaviors are refined into shallow / deep experiences. Combined with the results of implicit environmental factor deduction, personalized solutions are generated to support the activation of implicit resources in the scenic area, solving the problems of misjudging stay value, lack of implicit factor correlation, and single personalized dimension in existing systems.

[0025] II. Core Technology Solutions at Each Level: (I) Mobile Terminal Data Acquisition Layer: Signaling Unexpected Dwelling Behaviors and Deepening Experience Processing 1. Multi-dimensional behavioral data collection module: By using a mobile app to call the system's standard interface and in collaboration with scenic area auxiliary data collection points, a three-tiered transformation of behavioral data, demand signals, and experience intensity is achieved. The specific steps are as follows: (1) Optimization of existing data collection dimensions: Pressure sensing for phone holding posture: ①APP level: The user's touch screen pressure is obtained through the TouchEvent API built into the mobile phone system (a standard interface supported by both Android and iOS). The pressure value is recorded once every 500ms, and the pressure dataset is generated after continuous collection for 1 minute. ②Scenic Area Assistance: Deploy lightweight data collection stations (including infrared distance sensors, model: Sharp GP2Y0A21YK) along trails and rest areas. When the sensor detects that the distance between the user and the data collection station is less than 1.5 meters (determining that the user is staying nearby), it will simultaneously send a stay assistance signal to the APP. ③ Data Fusion: The APP combines the touch screen pressure dataset with the collection station signal and calculates the fluctuation value through a sliding window algorithm (window size 10 data points): if the fluctuation value is <0.2N and the collection station signal is received, it marks the deep experience grip state; if the fluctuation value is >0.5N, it marks the shallow observation grip state; when it is between 0.2N and 0.5N, the data is temporarily stored and will be combined with biometrics for judgment.

[0026] Screen interaction frequency type differentiation implementation: The app uses the ActivityManager API (Android standard interface) to identify the application package name used by the user after unlocking: Step 1: Record the type of application used each time you unlock the app (e.g., viewing photos of the surrounding environment in the photo album, browsing social media apps in WeChat / TikTok). Step 2: Count the number of unlocks and operation types within 10 minutes: If unlocking ≤ 1 time and it is to view environmental photos, mark it as a deep experience interaction state; if unlocking > 2 times / 5 minutes and it is to browse social media, mark it as a shallow observation interaction state.

[0027] (2) Biometric-assisted determination submodule: ① Local mapping of facial micro-expressions: The specific steps are as follows: The app accesses the system's standard camera interface (user authorization required, privacy compliance guaranteed), and uses location data collection points within the scenic area. Step 1: After the user authorizes the app, it calls up the front camera and takes one frame every 30 seconds. The image is processed in real time within the app and the original image is not stored. Step 2: Automatically crop the eye and mouth area using the APP's built-in algorithm (based on a 16:9 screen ratio preset coordinate frame (100,80)-(160,140), corresponding to a physical size of 2cm×3cm), and delete other areas of the image; Step 3: Run the micro-expression feature algorithm within the app (lightweight model, memory usage <50MB): S3-1: Grayscale processing (grayscale value = 0.299×R + 0.587×G + 0.114×B), grayscale range 0-255; S3-2: Canny edge detection (threshold 50-150) extracts the contours of the eyes and corners of the mouth; S3-3: Calculate the deviation of the center point of the eye contour in 3 consecutive frames (<2 pixels indicates sustained gaze) and the fluctuation of the corner of the mouth (<5° indicates small fluctuation). S3-4: If both conditions are met, and the scenic area data collection point detects that the user stays for more than 3 minutes, output a deep visual experience marker; otherwise, output a shallow observation marker. Step 4: Delete the feature data immediately after processing, keeping only the labeling results to avoid privacy leaks.

[0028] ② Touchscreen operation stability test: The app can obtain touch screen operation data by calling the MotionEvent API (a standard system interface), without needing to modify the fingerprint sensor. Step 1: Record the pressure-time curve when the user touches the screen (collect pressure values ​​once every 100ms and record the timestamps simultaneously). Step 2: Calculate the pressure fluctuation range within 2 consecutive minutes (<0.2N indicates stable pressure). Step 3: Calculate the duration of a single touch (the time difference from the first detection of pressure >0.1N to <0.1N; a duration >2 minutes is considered a sustained contact). Step 4: If the pressure is stable and the contact is continuous, output a stable operation mark for deep experience, with a high superimposed signal strength; otherwise, output a temporary operation with low superimposed strength.

[0029] (3) Data fusion and demand signal packet generation: The app aggregates touch screen pressure, screen interaction, micro-expression, and touch screen stability data, and combines them with the location of the stop sent by the scenic area's data collection piles (the data collection pile ID corresponds to the preset GPS coordinates) to generate a demand signal packet containing an experience intensity marker (high / low), the location of the stop, and a 10-second ambient sound clip (collected by the app using the microphone's standard interface). This packet is then uploaded to the edge node via NB-IoT (the app calls the system's network interface).

[0030] 2. Submodule for determining the validity of dwelling behavior: Based on APP data and scenic area data collection points, the system uses a four-dimensional logic to classify and determine dwell behavior. The specific process is as follows: Step 1: Location Determination (APP + Data Collection Point Collaboration): The APP obtains the phone's GPS coordinates (system standard interface) and simultaneously receives the ID signal of nearby data collection points. If the two match (coordinate deviation < 10 meters) and this continues for more than 3 minutes, the stationary position is determined to have a duration of more than 3 minutes; otherwise, it is considered a temporary stop.

[0031] Step 2: Grip posture determination (based on touchscreen pressure data): If the pressure fluctuation value output by the APP is less than 0.3N, proceed to the next step; otherwise, pause temporarily.

[0032] Step 3: Ambient sound determination (APP microphone acquisition): The app uses the microphone to collect ambient sound (system standard interface). It separates natural sound (200-5000Hz) from artificial sound (50-200Hz) using a built-in FFT algorithm (lightweight, CPU usage <10%). If the natural sound accounts for more than 50%, proceed to the next step; otherwise, pause temporarily.

[0033] Step 4: Biometric determination (based on micro-expressions + touchscreen stability): If the APP outputs a high marker (depth visual experience + depth experience stable operation), it is determined that the depth experience stay is effective and a high-priority demand signal packet is generated. If the output is low-marked (shallow observation + temporary operation), it is determined that the shallow effective experience stay is generated, and a medium-priority demand signal packet is generated. When mixing markings, take into account the pressure fluctuation value (<0.2N is effective for shallow markings, otherwise temporarily stop markings).

[0034] (II) Edge Node Data Processing Layer: Indirect Derivation of Latent Environmental Factors + Strength Correlation (through Scenic Area Sensor Nodes): 1. Mobile sensor node multiplexing module: The specific steps for deriving latent environmental factors by deploying micro-environmental sensor nodes in scenic areas and using location data from an app are as follows: (1) Derivation of local wind speed (including experience adaptation intensity classification): Step 1: Deploy "LoRa low-power wind speed sensor nodes" (model: FS3000, cost <50 yuan / unit, 1 unit per 50 meters) in the scenic area to collect wind speed data in real time (accuracy 0.1m / s) and upload it to the edge node via NB-IoT; Step 2: The edge node receives the "dwelling location (collection pile ID)" uploaded by the APP and matches it with the wind speed data of the sensor nodes near that location (e.g., collection pile A corresponds to sensor node A1, wind speed 2.1m / s). Step 3: Combine experience intensity classification: If the required signal packet is marked as high, retain one decimal place for the wind speed (2.1 m / s), and mark it as excellent for light wind adaptability; if it is marked as low, retain an integer number of places (2 m / s), and mark it as average.

[0035] (2) Derivation of light comfort (including experience tolerance threshold): Step 1: The scenic area sensor node (model: BH1750, light sensor) collects real-time light intensity (unit: lux) and uploads it to the edge node; Step 2: Combine the age range of tourists uploaded by the APP (filled in by the user when registering for the APP) with the edge node, and retrieve the preset age range - light tolerance lower limit table (1500 lux for those over 60 years old, 2000 lux for those aged 20-30). Step 3: Calculate the comfort score: 4-5 points for light intensity ≤ the lower tolerance limit, 2-3 points for light intensity > the lower tolerance limit and ≤ 3000 lux, and 1 point for light intensity > 3000 lux; the score decreases by 1 point when the mark is low.

[0036] (3) Derivation of microscopic temperature difference: Step 1: The scenic area's sensor node (model: SHT30, temperature and humidity sensor) collects real-time temperature and calculates the temperature difference with the surrounding area (e.g., sensor node B collects a temperature of 25℃, while surrounding node C collects a temperature of 27℃, resulting in a temperature difference of 2℃). Step 2: Combine the dwell time uploaded by the APP to the edge nodes (>5 minutes to determine environmental adaptation) to correct the temperature difference (the temperature difference data is valid if the dwell time is >5 minutes, otherwise the average value of the surrounding nodes is taken).

[0037] 2. Behavior-Environment Association Mapping Module: Generate preference labels based on demand signal packets and scenic area sensor node data. Specific steps: Step 1: The edge node receives the demand signal packet (experience intensity marker) uploaded by the APP and the implicit environmental factor data of the sensor node; Step 2: Set the weighting coefficients (high label 1.5, low label 1.0). Step 3: Generate labels (e.g., wind speed 2.1m / s + high mark → strong preference for light wind 2.0-2.2m / s); Step 4: Establish a visitor ID-tag MySQL database at the edge node and update iteratively (new tag weight 0.7, old tag weight 0.3).

[0038] (III) Two-way feedback planning layer: experience-driven dynamic closed-loop planning + intensity adaptation: 1. Personalized route generation module: Route planning based on scenic area node database and preference tags, specific steps: Step 1: Build the scenic area node database (edge ​​node deployment): Includes the location of core / hidden nodes (coordinates corresponding to the data collection pile ID), real-time environmental data of sensor nodes (updated every 15 minutes), and experience intensity type (e.g., Bamboo Forest Path corresponds to sensor node A2, wind speed 2.0m / s, core experience node). Step 2: Calculate the experience matching value (formula as before, data from sensor nodes); Step 3: Filter nodes (matching score ≥ 70, prioritize core nodes), sort by core → transition → core, walking time ≤ 15 minutes; Step 4: The APP pushes the route (calls the system navigation interface, such as Gaode Map SDK) and sends environmental forecasts at the same time (such as the wind speed of the bamboo forest path ahead is 2.0m / s, and it is recommended to stay for 40 minutes).

[0039] 2. Real-time feedback iteration module: The app uploads behavioral data in real time, and edge nodes adjust routes. Specific steps: Step 1: The app uploads touch screen operation and micro-expression data every 2 minutes; Step 2: Determine the change in experience intensity at the edge nodes (e.g., if two consecutive data points satisfy eye gaze switching > 3 times / minute + touch duration < 1 minute → intensity decreases from high to low). Step 3: Edge node searches for surrounding transition nodes (matching value ≥ 70 points), the APP pushes adjustment suggestions (such as whether to go to Songyin Tea House for 15 minutes), and updates the route after user confirmation.

[0040] (iv) Scenic Area Resource Adaptation Layer: Activation and Feedback of Implicit Resources 1. Latent Resource Map Generation Module: Generating a heat map based on APP dwell time data and scenic area sensor node data: Specific steps: Step 1: Analyze the deep dwell frequency (number of times the user has a deep and effective experience) of hidden nodes at the edge nodes and the total frequency, and calculate the percentage. Step 2: Create a heat map using ArcGIS (red > 60%, yellow 30%-60%, blue < 30%), and push it to the scenic area management platform; Step 3: Add QR code AR guides to core nodes (tourists scan the code to access the APP AR function, using the system's standard interface), and add convenient facilities to transition nodes.

[0041] 2. Pedestrian-Environment Coordinated Control Module: Based on infrared people flow counters and preference tag-based crowd control, the specific steps are as follows: Step 1: Deploy infrared crowd counters (Hikvision DS-2CD3T46WDV3-I3) at key scenic spots, with a density > 3 people / m². 2 Trigger traffic splitting; Step 2: Filter surrounding tourist tags (strong / weak preference) at edge nodes; Step 3: APP push traffic routing (strong preference pushes core hidden nodes, weak preference pushes transitional nodes), and the management platform displays the progress.

[0042] III. Examples of Scenarios: 1. Addressing scenarios where the value of unintended stays is misjudged: Example scenario: 65-year-old Aunt Li stops at an ancient bridge in a scenic area. The mobile app calls the front camera (with authorization) and takes an image of her eyes and mouth every 30 seconds. The image is pointed to the river surface to the west for 3 consecutive frames (with a deviation of 1 pixel), and the corner of her mouth fluctuates by 3°. The app detects a fluctuation of 0.15N in touch screen pressure for 3 minutes. The infrared sensor of a nearby data collection station detects that she is standing next to it and sends a location signal at the same time.

[0043] Expected outcome: The system determined the stay to be a deeply effective experience, and the app did not prompt for a return (unlike the current system), but instead pushed an audio explanation of the century-old bridge's construction techniques (played through the app's speaker). Aunt Li extended her stay by 20 minutes, and reported that she was not interrupted and could listen to the explanation with peace of mind, which was better than last time, resolving the current system's problem of misjudging the value of a stay.

[0044] 2. Addressing the impact of scenarios lacking monitoring and correlation of latent environmental factors: Example scenario: Xiao Zhang (25 years old), who is afraid of the heat, stays on the Xigu platform. The APP uploads his location to the edge node and matches the data of nearby sensor nodes (temperature difference 3.2℃, wind speed 2.1m / s, light comfort 4 points). Combined with the deep experience determined by the APP, a strong preference label of temperature difference 2.7-3.7℃ + breeze 2.0-2.2m / s is generated.

[0045] Expected result: The system recommends the bamboo forest path (sensor nodes show a temperature difference of 3.0℃ and a wind speed of 2.0m / s at 10:00 AM), instead of the sunflower field recommended by the existing system (temperature difference of 1.5℃ and no wind). Xiao Zhang stayed in the bamboo forest for 40 minutes and reported that it was cool there, and he was surprised to find such a place in the scenic area, thus resolving the issue of the existing system lacking latent factor correlation.

[0046] 3. Solve the problem of scene effects with limited personalization: Example scenarios: Grandpa Wang (70 years old) stays at the Songyin Tea House. The APP detects that he frequently switches touch screens (>3 times / minute), which is judged as a shallow experience and generates a weak preference low light 1500-2500 lux label; Xiao Li (22 years old) stays at the mountain top viewing platform. The APP detects that he touches the screen steadily (>2 minutes), which is judged as a deep experience and generates a strong preference high light 2500-3500 lux label.

[0047] Expected results: The system planned a route for Grandpa Wang from the pine-shaded teahouse (transition, 1800 lux of sunlight) to the wisteria trellis (core, 2000 lux of sunlight), and for Xiao Li from the mountaintop (core, 3000 lux of sunlight) to the ginkgo trail (transition, 2800 lux of sunlight). Both reported that the routes suited their preferences; Grandpa Wang was shaded, and Xiao Li had a bright view, thus addressing the current system's lack of personalization and uniformity.

[0048] 4. Added scene effects for resource development guided by experience intensity: Example scenario: The heat map of the scenic area shows that 80% of visitors lingered under the wisteria trellis (red), and 35% lingered on the north side of the artificial hill (yellow). The scenic area has added a QR code AR guide under the trellis (visitors can scan the code to access the APP's AR function and view the virtual wisteria growth process), and added anti-corrosion wooden benches and drinking fountains on the north side of the artificial hill.

[0049] Expected results: After one month, the time spent under the flower stand increased from 30 minutes to 45 minutes. Visitors reported that AR was interesting and wanted to bring their families. The number of visitors staying on the north side of the artificial hill increased from 50 to 80. Visitors reported that they could rest and drink water when they were tired. This solved the problem that the existing system could not activate hidden resources.

[0050] 5. The effect of tiered pedestrian flow control in various scenarios: Example scenario: At 10 AM on Saturday, the number of people at the waterfall viewing platform reaches 4 per square meter. 2 The system filters out 20 tourists with a strong preference for light breezes (such as Xiao Zhang) and 30 tourists with a weak preference (such as Grandpa Wang) in the surrounding area, and pushes the Xigu platform (core, with a flow of 0.3 people / m²) to Xiao Zhang. 2 (Wind speed 2.1 m / s), pushing the pine-shaded teahouse (transition, pedestrian flow 0.8 people / m) towards Grandpa Wang. 2 ).

[0051] Expected results: After 15 minutes, the number of people at the viewing platform dropped to 2.5 people / ㎡. Xiao Zhang could take photos in the valley without queuing, and Grandpa Wang could rest and then go to the viewing platform without having to wait long. Feedback showed that the flow of people was not crowded after the diversion, and that they did not go to the wrong place. This solved the problem of indiscriminate flow diversion in the existing system.

[0052] IV. System Implementation Process: (I) First Phase: Deployment and Preparation at the Scenic Area (1-2 months): 1. Hardware site selection and installation: Based on the scenic area's topographic map (such as in CAD format), a lightweight data collection station (Sharp GP2Y0A21YK infrared sensor) is deployed every 50 meters along the trails, in rest areas, and at hidden nodes (such as under the wisteria trellis or on the north side of the artificial hill). The base of the data collection station is fixed with concrete (1.2 meters high to prevent children from touching it), and the power supply is provided by solar panels (10W power, with a battery life of ≥72 hours in cloudy or rainy weather). Infrared crowd counters (Hikvision DS-2CD3T46WDV3-I3) were deployed at key scenic spots (waterfall viewing platform, ancient bridge), installed at a height of 2.5 meters, with the lens angle covering 90% of the scenic area; One set of environmental sensing nodes (FS3000 wind speed sensor + BH1750 light sensor + SHT30 temperature and humidity sensor) is deployed every 50 meters. The sensors are encapsulated in a waterproof box (IP65 protection level) and connected to the NB-IoT network through a LoRa gateway.

[0053] 2. Edge node setup: The scenic area operation and maintenance center was selected as the deployment location for the edge node. An industrial-grade server (CPU: Intel Xeon E3-1230v6, memory 16GB, hard disk 1TB SSD) was configured, and a MySQL 8.0 database (used to store tourist tags and node environment data) and ArcGIS Server (used to generate experience intensity heat maps) were installed. Deploy NB-IoT gateways (Huawei MG3000) to ensure coverage of 95% of the scenic area and transmission latency ≤100ms.

[0054] (II) Second Phase: User-side App Development and Testing (1 month): 1. App Function Development: Developed using Android Studio (Android) and Xcode (iOS), supporting Android 8.0 and above, and iOS 12.0 and above. Core functional modules: Permission request module (when the app is first opened, a pop-up window requests permissions for "camera, microphone, GPS, and network", with a privacy statement: data is only processed in real time and raw images / audio are not stored). Data acquisition module (calls standard interfaces such as TouchEvent API and MotionEvent API to collect touch screen pressure at 500ms / time and micro-expression images at 30 seconds / time); Route receiving and navigation module (integrates Amap SDK, supports voice navigation and node environment preview pop-up).

[0055] 2. App Testing: Recruit 50 tourists of different ages (20 aged 20-30, 20 aged 40-50, and 10 aged 60 and above) to conduct a one-week test at the scenic area to verify: Data collection accuracy (touchscreen pressure fluctuation error ≤ 0.05N, micro-expression judgment accuracy ≥ 90%). Impact on battery life (after 4 hours of continuous use, the phone's battery consumption is ≤20%).

[0056] (III) Third Phase: System Integration and Formal Operation (1 month): 1. Joint debugging and testing: Simulated scenario: Ten test tourists are arranged to stay in the scenic area. The APP uploads the demand signal packet. After receiving it, the edge node infers the implicit environmental factors, generates preference tags, and then pushes the route. The verification process takes ≤3 seconds. Fault test: Disconnect one environmental sensor node and check whether the edge node automatically switches to the data of the surrounding nodes (switching time ≤ 5 seconds).

[0057] 2. Formal Operation: The scenic area promotes its APP through its official website and WeChat account. Tourists can scan the QR code to download and register (when registering, they need to fill in their age range for matching light tolerance thresholds). The scenic area management system can view heat maps and visitor flow progress in real time, and generate a system operation report every week (including data on hidden node stays and visitor satisfaction feedback).

[0058] V. System Deployment and Maintenance: (a) Hardware maintenance requirements: 1. Regular inspections: Data collection piles and environmental sensing nodes: On-site inspection once a month, focusing on whether the solar panels are dusty (wipe with a dry cloth) and whether the waterproof box is leaking (replace the aging sealing ring). Infrared people counter: Clean the lens once a quarter (wipe with lens paper) and calibrate the counting accuracy (the error should be <5% compared with manual counting).

[0059] 2. Replacement of faulty parts: Spare parts reserve: Reserve 10% of data collection points and environmental sensor nodes (e.g., if 100 data collection points are deployed in a scenic area, reserve 10 as spares). Replacement process: Upon receiving a fault alarm (the edge node detects that the device is offline for more than 10 minutes and automatically sends an SMS to the maintenance personnel), we will arrive at the site within 2 hours to replace the device. After replacement, the device will be activated through the maintenance platform.

[0060] (ii) Software maintenance: 1. App Update: One minor update per quarter (bug fixes, interface optimization), and one major update per year (adding new features, such as AR navigation supporting 3D models). A pop-up window will remind users during the update that the update package size is ≤50MB (to avoid excessive consumption of visitor traffic).

[0061] 2. Edge node algorithm iteration: Every six months, the micro-expression feature algorithm is optimized based on tourist data (including adding a sub-module for recognizing the eye contours of tourists wearing glasses to improve accuracy). The age-tolerance light tolerance limit table is updated monthly (based on visitor feedback, such as visitors over 60 years old who still find 1500 lux too dazzling, so it is lowered to 1200 lux).

[0062] (III) Data security maintenance: 1. Privacy and data protection: On the app: Raw micro-expression images and touch screen pressure data are processed in real time and deleted immediately; only the result data such as experience intensity markers are uploaded. Edge nodes: Visitor IDs and preference tags are stored in encrypted form (using AES-256 encryption algorithm), and the database is automatically backed up to a remote server every 24 hours.

[0063] 2. Data access permissions: The scenic area management system has three levels of access control: Administrator (can view all data and export reports); Maintenance personnel (can only view equipment operation data, without guest tag access); Service personnel (can only view heatmaps and diversion suggestions, but do not have database access).

[0064] VI. System Expansion and Application Directions: (a) Collaborating with scenic area merchants to enhance the consumer experience: 1. Personalized consumption recommendations: Edge nodes combine tourist preference tags to push riverside tea house discounts to tourists who strongly prefer a gentle breeze and in-depth experience (such as 20% off tea drinks with a screenshot of the route on the APP). Push information about cultural and creative stores to tourists who are experiencing a transitional period (such as the cultural and creative store next to Songyin Tea House, which offers AR scanning function as a gift when purchasing a souvenir album).

[0065] 2. Merchant data feedback: Merchants can view visitor preference tags (such as 80% of visitors preferring weak light) through the scenic area management platform and adjust their business strategies accordingly (such as adding sunshades to cultural and creative stores).

[0066] (ii) Integrating cultural and tourism IPs to enhance the enjoyment of the experience: 1. IP-themed route generation: If the scenic area has an ancient bridge legend IP, the APP will push IP-themed routes to tourists who want to experience the ancient bridge in depth (such as the ancient bridge → the former residence of the legendary craftsman → the IP cultural and creative museum), and the navigation voice will be voiced by the IP character.

[0067] 2. AR Interactive Tasks: AR tasks can be set up at core experience nodes (such as scanning wisteria flowers to trigger IP characters to explain the history of the flower stand). Completing the tasks will earn experience points (which can be redeemed for discounts on scenic area tickets).

[0068] (iii) Accessibility and coverage of special tourist groups: 1. Suitable for visually impaired tourists: The app adds a voice-guided navigation function, which calls the phone's screen reader interface to convert route information and node environment (such as a bamboo forest path ahead, wind speed of 2.0 m / s, and no steps) into voice. The data collection station will now vibrate to alert visitors to the accessible pathway.

[0069] 2. Suitable for elderly tourists The APP interface is simplified (the font size is increased to 18 points, and the button size is ≥2cm×2cm), and the route push reduces jumps (one step directly to the navigation page). Edge nodes push slow-paced routes to elderly tourists (walking time between nodes is extended to 20 minutes, and rest nodes are added).

Claims

1. An intelligent scenic area tourism planning system, characterized in that, It includes a mobile terminal data acquisition layer, an edge node data processing layer, a bidirectional feedback planning layer, and a scenic area resource adaptation layer, which sequentially interact with 5G narrowband IoT low-power real-time interaction. The mobile terminal data acquisition layer includes a multi-dimensional behavioral data acquisition module, a biometric auxiliary judgment submodule, and a stay behavior validity judgment submodule, which are used to convert tourists' unexpected stay behaviors into demand signal packets containing experience intensity markers; The edge node data processing layer includes a mobile sensor node multiplexing module and a behavior-environment association mapping module, which are used to deduce implicit environmental factors such as temperature difference, wind speed, and light comfort in the micro-space based on demand signal packets, and generate tourist environmental preference labels. The two-way feedback planning layer includes a personalized route generation module and a real-time feedback iteration module, which are used to dynamically plan tour routes based on preference tags and adjust routes according to real-time tourist behavior data. The scenic area resource adaptation layer includes a hidden resource map generation module and a pedestrian flow-environment collaborative regulation module, which are used to activate the hidden resources of the scenic area and realize hierarchical pedestrian flow regulation based on tourist tags.

2. The intelligent scenic area tourism planning system as described in claim 1, characterized in that: The multi-dimensional behavioral data acquisition module includes a touch screen pressure acquisition unit, a screen interaction frequency determination unit, and a data fusion unit; The touch screen pressure acquisition unit records the touch screen pressure value once every 500ms by calling the TouchEvent API through the mobile APP. After continuous acquisition for 1 minute, a pressure dataset is generated. Combined with the lightweight acquisition piles with infrared distance sensors deployed in the scenic area, the pressure fluctuation value is calculated using a sliding window algorithm with a window size of 10 data points. When the fluctuation value is <0.2N and a stop assistance signal is received, it is marked as a deep experience grip state. When the fluctuation value is >0.5N, it is marked as a shallow observation grip state. The screen interaction frequency determination unit identifies the application package name operated by the user after unlocking by calling the ActivityManager API of the APP, and counts the number of unlocks and operation types within 10 minutes. When unlocking ≤1 time and the operation type is viewing environmental photos, it is marked as a deep experience interaction state. When unlocking >2 times / 5 minutes and the operation type is browsing social software, it is marked as a shallow observation interaction state. The data fusion unit aggregates touch screen pressure data and screen interaction data to provide basic behavioral data for demand signal packets.

3. The intelligent scenic area tourism planning system as described in claim 2, characterized in that: The biometric-assisted judgment submodule includes a facial micro-expression processing unit and a touchscreen operation stability detection unit; The facial micro-expression processing unit uses the APP to call the front camera to capture one frame every 30 seconds. It automatically crops the eye-mouth area corresponding to the preset coordinate box (100,80)-(160,140) based on the 16:9 screen ratio. The cropped area is then processed in grayscale and Canny edge detection with a threshold of 50-150. The deviation of the eye contour center point and the fluctuation of the mouth corner arc are calculated for three consecutive frames. If both conditions are met and the lightweight acquisition pile detects that the user stays for more than 3 minutes, a depth visual experience mark is output. Otherwise, a shallow observation mark is output. The touch screen operation stability detection unit collects touch screen pressure values ​​every 100ms and records the timestamp by calling the MotionEvent API through the APP, calculates the pressure fluctuation range within 2 consecutive minutes, and counts the duration of a single touch. If the pressure is stable and the contact is continuous, it outputs a deep experience stable operation flag; otherwise, it outputs a temporary operation flag.

4. The intelligent scenic area tourism planning system as described in claim 3, characterized in that: The dwell behavior validity determination submodule classifies dwell behaviors based on four-dimensional logical judgments: Step 1, Location Determination: The APP obtains the phone's GPS coordinates and receives the ID signal of nearby lightweight data collection points. If the coordinates of the two are less than 10 meters and last for more than 3 minutes, the location is determined to be stationary for more than 3 minutes; otherwise, it is considered a temporary stop. Step 2, grip posture determination: Call the pressure fluctuation value output by the touch screen pressure acquisition unit. If it is <0.3N, proceed to the next step; otherwise, pause temporarily. Step 3, Ambient Sound Determination: The APP calls the microphone to collect ambient sound. It uses a lightweight FFT algorithm that consumes less than 10% of the CPU to separate natural sound from 200-5000Hz and artificial sound from 50-200Hz. If the proportion of natural sound is greater than 50%, proceed to the next step; otherwise, pause temporarily. Step 4, Biometrics Determination: If the biometrics-assisted determination submodule outputs a high mark for depth visual experience + stable depth experience operation, it determines that the depth experience is valid and generates a high-priority demand signal packet; if it outputs a low mark for shallow observation + temporary operation, it determines that the shallow experience is valid and generates a medium-priority demand signal packet; when there is a mixed mark, if the pressure fluctuation value is <0.2N, it is determined as shallow valid stay, otherwise it is temporary stay.

5. The intelligent scenic area tourism planning system as described in claim 4, characterized in that: The mobile sensor node multiplexing module derives implicit environmental factors in the following way: (1) Derivation of local wind speed: One wind speed sensor node is deployed every 50 meters in the scenic area. The edge node receives the dwell position in the demand signal packet and matches the wind speed data of the nearby sensor node. When the demand signal packet is marked as high, the wind speed is retained to one decimal place and the wind adaptability intensity is marked as excellent. When it is marked as low, the wind speed is retained to an integer place and the wind adaptability intensity is marked as average. (2) Derivation of light comfort: The scenic area deploys light sensors to collect real-time light intensity. The edge nodes combine the age range of tourists uploaded by the APP to retrieve the preset age range-light tolerance lower limit table and calculate the comfort score: 4-5 points for light intensity ≤ tolerance lower limit, 2-3 points for light intensity > tolerance lower limit and ≤ 3000 lux, 1 point for light intensity > 3000 lux, and the score is reduced by 1 point when the demand signal packet is marked as low. (3) Derivation of micro temperature difference: SHT30 temperature and humidity sensors are deployed in the scenic area to collect real-time temperature and calculate the temperature difference with the surrounding area. The edge nodes are combined with the dwell time uploaded by the APP. If the dwell time is >5 minutes, the temperature difference data is valid; otherwise, the average value of the surrounding nodes is taken.

6. The intelligent scenic area tourism planning system as described in claim 5, characterized in that: The workflow of the behavior-environment association mapping module is as follows: Step 1: The edge node receives the experience intensity marker of the demand signal packet and the implicit environmental factor data collected by the mobile sensor node multiplexing module. Step 2: Set the weight coefficients for experience intensity labels, with a weight of 1.5 for high labels and a weight of 1.0 for low labels; Step 3: Generate tourist environmental preference labels based on implicit environmental factor data and weighting coefficients; Step 4: The edge node establishes a MySQL database of visitor ID-preference tags. When iterating and updating the tags, the weight of the new tag is 0.7 and the weight of the old tag is 0.

3.

7. The intelligent scenic area tourism planning system as described in claim 6, characterized in that: The personalized route generation module plans routes through the following steps: Step 1: Deploy a scenic area node database at the edge nodes, including the locations of core / hidden nodes, real-time environmental data of sensor nodes, and experience intensity types; Step 2: Calculate the experience matching value based on implicit environmental factor data. The calculation logic is consistent with the lighting comfort scoring logic. Step 3: Filter nodes with an experience matching score of ≥70, prioritize core nodes, and sort them in the order of core → transition → core, with a walking time between nodes ≤15 minutes; Step 4: The APP calls the Gaode Map SDK to push the route and simultaneously sends environmental forecast information; The workflow of the real-time feedback iteration module is as follows: Step 1: The app uploads touch screen operation data and micro-expression data every 2 minutes; Step 2: If the edge node detects two consecutive data points that satisfy eye gaze switching > 3 times / minute and touch screen duration < 1 minute, the experience intensity is determined to decrease from high to low. Step 3: Edge nodes search for transitional nodes with a matching experience score of ≥70, push route adjustment suggestions to the APP, and update the route after user confirmation.

8. The intelligent scenic area tourism planning system as described in claim 7, characterized in that: The workflow of the latent resource map generation module is as follows: Step 1: Count the deep dwell frequency and total dwell frequency of hidden nodes at the edge nodes, and calculate the deep dwell percentage. Step 2: Create a heat map using ArcGIS, with the heat map colors being red, yellow, and blue, and then push it to the scenic area management platform; Step 3: Add QR code AR navigation to core nodes and add convenient facilities to transition nodes; The workflow of the pedestrian-environment coordinated control module is as follows: Step 1: Deploy Hikvision DS-2CD3T46WDV3-I3 infrared people counters in key scenic spots, ensuring a people density of >3 people / m². 2 Time-triggered traffic splitting; Step 2: Filter environmental preference tags of surrounding tourists at edge nodes to distinguish between strong and weak preferences; Step 3: The APP pushes diversion routes of core hidden nodes to tourists with strong preferences and diversion routes of transitional nodes to tourists with weak preferences. The scenic area management platform displays the diversion progress in real time.

Images