A riding route planning method and system based on label matching and a reference route material library

Abstract

The application relates to the technical field of route planning, and discloses a cycling route planning method and system based on label matching and a reference route material library. The method comprises the following steps: analyzing user-input cycling area, road preference, landscape type and travel time length preference labels, extracting keywords through natural language processing and mapping the keywords to a preset label library to generate scene matching conditions; based on a preset reference route material library, splitting, reorganizing and topologically splicing standardized mature cycling route segments in the library to generate an adaptive cycling path; embedding travel and tourism check-in points, rest stations and public service facilities along the path, limiting the distance threshold of the points and the proportion of the stay time length of a single point; generating multiple sets of differentiated travel schemes, and preferentially outputting the matching results of the material library; and combining real-time point passenger flow to fine-tune the route, supporting manual editing and real-time navigation along the route. The application can improve the adaptability of cycling route planning and user demand.

Description

Technical Field

[0001] This application relates to the technical field of route planning, and in particular to a method and system for cycling route planning based on tag matching and a baseline route database. Background Technology

[0002] As cycling continues to gain popularity, the public's demand for cycling routes is no longer limited to basic transportation functions, but increasingly values ​​the overall experience, including scenery and rest facilities along the way. Current cycling navigation and route planning technologies mostly rely on real-time road networks for computation and deduction. Some solutions focus on calculating physical parameters such as cycling stamina and duration, while general AI planning solutions primarily rely on model computation, generally depending on real-time calculations of the global road network to generate routes. This results in a high overall computational load and significant shortcomings in lightweight operation.

[0003] While current cycling planning tools can provide basic route recommendations, they struggle to deeply integrate cultural and tourism attractions, rest facilities, and route planning, resulting in significant route homogenization and an inability to quickly create suitable combinations based on established route resources.

[0004] As can be seen from the above, the existing cycling route planning methods have a large computational load and insufficient route adaptability. How to improve the adaptability of cycling route planning to user needs still needs to be addressed. Summary of the Invention

[0005] To improve the adaptability of cycling route planning to user needs, this application provides a cycling route planning method and system based on tag matching and a baseline route material library.

[0006] Firstly, this application provides a cycling route planning method based on tag matching and a baseline route database, employing the following technical solution: A cycling route planning method based on tag matching and a baseline route database includes: The system parses the user's input preference tags, which include cycling area, road preference, landscape type and travel duration. Based on the preference tags, keywords are extracted through natural language processing and mapped to a preset tag library to generate corresponding scene matching conditions. Based on a pre-set baseline route material library, and combined with an improved path planning algorithm, an adapted cycling route is calculated. The standardized route segments in the baseline route material library are split, recombined, and topologically stitched together. The baseline route material library stores mature cycling route segments that have been standardized. After the standardized route segments are split, recombined, and topologically stitched together, cultural and tourism check-in points, rest stops, and public service facilities along the route are embedded. When embedding, points that are no more than a preset threshold away from the route are selected, and the time spent at a single point does not exceed a preset proportion of the total travel time. Multiple differentiated itinerary plans are automatically generated based on the user's available cycling time and user needs, focusing on different dimensions such as route maturity, landscape richness, and cycling time, and prioritizing the output of material library matching recommendation results; The system obtains users' real-time locations and makes fine-tuning adjustments based on the real-time passenger flow status of these locations. When the passenger flow at a location exceeds the preset value, the system adjusts the order of stops or replaces alternative locations. It also supports users manually adjusting the route, adding or removing stops along the way, modifying the cycling time, and providing real-time navigation along the route.

[0007] Optionally, the steps of using a pre-built baseline route library, combining an improved route planning algorithm to calculate an adapted cycling route, and splitting, recombining, and topologically stitching standardized route segments within the library include: The collected mature cycling routes are segmented and cut, with cutting nodes set at route intersections, landscape boundaries, and road type changes to generate standardized route segments with appropriate lengths. Each standardized route segment is labeled with a corresponding cycling area label, road type label, landscape type label, segment length, and average cycling time. Establish a topological index for each standardized route segment, recording the starting coordinates, ending coordinates, and numbers of directly connectable adjacent segments for each segment; The standardized route segments with annotations and their corresponding topology indexes are stored together in the baseline route material library.

[0008] Optionally, the steps of splitting, reorganizing, and topologically stitching standardized route segments within the library include: Based on the generated scene matching conditions, a subset of standardized route segments with matching tags is selected from the baseline route material library; Extract the segment whose starting point coincides with the user-specified starting point from the standardized route segment subset as the starting segment, and extract the segment whose ending point coincides with the user-specified ending point as the ending segment; Between the starting segment and the ending segment, intermediate segments that can be sequentially connected are retrieved and combined to form multiple candidate routes that meet the user's available riding time requirements; The connection points of each candidate route are smoothed out, duplicate and turnaround sections are removed, and a suitable cycling path is generated.

[0009] Optionally, fine-tuning can be performed based on real-time visitor flow. The step of adjusting the stop order or replacing alternative locations when visitor flow exceeds a preset value includes: Real-time acquisition of passenger flow density data for various cultural and tourism check-in points and rest stations embedded in the cycling route; When the passenger flow density at a location exceeds the preset passenger flow threshold, calculate the estimated waiting time for that location; If the expected waiting time exceeds the maximum allowed stay time for a single location, search for alternative locations of the same type within a preset range from the baseline route material library; Replace the original points with alternative points, recalculate the total cycling time and route direction, and generate a fine-tuned cycling route.

[0010] Optionally, after smoothing the junctions of each candidate route, removing duplicate and turnaround sections, and generating a suitable cycling path, the process includes: For each candidate route, a route maturity score, landscape richness score, and slope gentleness score are calculated. The route maturity score is based on the historical cycling count of the corresponding standardized route segment, the landscape richness score is based on the number and level of landscape type labels marked on the corresponding segment, and the slope gentleness score is based on the average slope parameter of the corresponding segment. Based on the different weights of the three scores, multiple differentiated itinerary plans are generated. Among them, the first plan has the highest weight for route maturity score, the second plan has the highest weight for landscape richness score, and the third plan has the highest weight for slope gentleness score. The total duration of the generated multiple differentiated trip plans is verified. Plans that exceed the user's available riding time are eliminated, and the plans that meet the requirements are retained and sorted and output according to the preset priority.

[0011] Optionally, the step of generating the fine-tuned cycling route includes: It connects to the user's smart wearable device via Bluetooth interface to obtain the user's heart rate data in real time during cycling. A preset heart rate safety threshold is retrieved. When the user's heart rate is detected to exceed the heart rate safety threshold, the user's current location is located, and a low-intensity standardized route segment with an average slope lower than the preset value within a preset range around the current location is retrieved from the baseline route material library. Replace the high-intensity sections of the current route that have not been ridden with the retrieved low-intensity standardized route segments, and recalculate the total riding time and route connection points; The recommended dwell time at rest stops along the route is adjusted in sync, extending the time users spend at rest stops near low-intensity route segments, and generating a finely tuned cycling route.

[0012] Optionally, after the step of storing the labeled standardized route segments and their corresponding topology indexes in the baseline route material library, the following steps are included: Collect actual cycling trajectory data after the user completes the ride, and extract parameters such as the start point, end point, route segments, and average cycling time of the trajectory; The actual cycling tracks collected are compared with the standardized route segments already existing in the baseline route material library to identify new route segments that have not been included. The system automatically segments and labels the newly identified route segments to generate new standardized route segments that meet the format requirements of the baseline route material library. Establish a topological index for the new standardized route segments, update the segment set and topological connection relationship of the baseline route material library, and complete the dynamic update of the material library.

[0013] Secondly, this application provides a cycling route planning system based on tag matching and a baseline route database, employing the following technical solution: A cycling route planning system based on tag matching and a baseline route database includes: The tag parsing module parses the user's input preference tags, which include cycling area, road preference, landscape type and travel duration. Based on the preference tags, keywords are extracted through natural language processing and mapped to a preset tag library to generate corresponding scene matching conditions. The path stitching module, based on a pre-set baseline route material library, calculates an adapted cycling path using an improved path planning algorithm, and splits, reassembles, and stitches standardized route segments in the baseline route material library into topological segments. The baseline route material library stores standardized and mature cycling route segments. The point embedding module embeds cultural and tourism check-in points, rest stations and public service facilities along the route after the standardized route segments are split, recombined and topologically stitched. When embedding, points that are no more than a preset threshold away from the route are selected, and the time spent at a single point does not exceed a preset proportion of the total travel time. The plan generation module automatically generates multiple differentiated itinerary plans based on the user's available cycling time and user needs, focusing on different dimensions such as route maturity, landscape richness, and cycling time, and prioritizing the output of material library matching and recommendation results; The dynamic fine-tuning module obtains the user's real-time location and makes fine-tuning based on the real-time passenger flow status of the location. When the passenger flow at a location exceeds the preset passenger flow value, it adjusts the order of stops or replaces alternative locations. It also supports users to manually adjust the route, add or remove stops along the way, modify the cycling time, and provide real-time navigation along the route.

[0014] Thirdly, this application provides an electronic device that adopts the following technical solution: An electronic device includes a processor running a program for a cycling route planning method based on tag matching and a baseline route library as described in any one of the preceding claims.

[0015] Fourthly, this application provides a storage medium, which adopts the following technical solution: A storage medium storing a program for a cycling route planning method based on tag matching and a baseline route library as described in any one of the above.

[0016] In summary, this application includes at least one of the following beneficial technical effects: By using multi-dimensional preference tags input by users to complete semantic parsing and matching, and combining them with a baseline route material library composed of pre-set standardized and mature cycling route fragments, basic cycling routes are quickly generated directly by splitting and recombining fragments, instead of relying on real-time iterative calculations of the global road network. At the same time, the access distance and stay duration of cultural and tourism points are limited, which can reduce the deviation between demand and route from the source. Multi-dimensional differentiated itinerary plans are output simultaneously, which can greatly improve the matching degree between cycling routes and users' personalized needs, reduce the computing load of devices, and achieve lightweight route planning.

[0017] This solution can replace supporting cultural and tourism and rest points based on real-time passenger flow dynamics, and can also combine wearable device heart rate data to replace low-intensity road sections. It can also continuously expand and update route materials based on user cycling tracks, continuously optimize the route library's adaptation and coverage, and take into account cycling experience, physical safety and route timeliness, comprehensively solving the problems of single routes, rigid planning and inability to meet users' real travel needs. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a cycling route planning method based on tag matching and a baseline route library, according to an exemplary embodiment.

[0019] Figure 2 This is a structural block diagram of a cycling route planning system based on tag matching and a baseline route material library, according to an exemplary embodiment. Detailed Implementation

[0020] The embodiments of this application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.

[0021] In the description of this specification, the references to "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples" refer to specific features, structures, materials, or characteristics described in connection with the described embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0022] This application discloses a cycling route planning method based on tag matching and a baseline route database, primarily applicable to scenarios such as urban leisure cycling planning, cultural tourism cycling route recommendation, outdoor cycling navigation services, and public fitness cycling guidance. This application, through an architecture combining tag matching and a baseline route database, integrates user preference analysis, route fragment topology stitching, route point embedding, multi-dimensional scheme generation, and dynamic real-time adjustment into a complete planning loop. It relies on pre-built standardized and mature route fragments to replace real-time computation across the entire road network, reducing the computational load on planning from the ground up; it achieves accurate translation of user needs through a multi-dimensional tag system, improving the matching degree between routes and needs from the source; it embeds cultural tourism points and service facilities into the core planning process, achieving a deep integration of cycling accessibility and leisure experience; and, combined with multi-dimensional differentiated schemes and a real-time dynamic adjustment mechanism, it balances route richness and adaptability, providing high-quality cycling route services tailored to the needs of different types of users.

[0023] This application discloses a cycling route planning method based on tag matching and a baseline route database, referring to... Figure 1 ,include: S100: Parse the user's input preference tags, which include cycling area, road preference, landscape type and travel duration. Based on the preference tags, extract keywords through natural language processing and map them to a preset tag library to generate corresponding scene matching conditions.

[0024] In this embodiment of the application, the specific execution process of S100 is as follows: The S101 receives user-inputted preference tags, covering four core dimensions: cycling area, road preference, landscape type, and trip duration. It supports multiple input methods to adapt to different user habits. Input methods include visual option selection, natural language text input, voice input transcription, and map interactive selection. Under the cycling area dimension, users can input specific administrative region names, business districts, and scenic area names, or directly define the cycling area on the electronic map using a selection tool. This covers cycling needs at different spatial scales, including urban areas, suburban areas, areas surrounding specific scenic spots, and riverside / lakeside zones. The road preference dimension includes various common cycling surface types such as asphalt greenways, ordinary municipal roads, suburban dirt roads, and riverside trails. It also provides sub-options such as the presence or absence of motor vehicle interference, whether it includes a dedicated cycling lane, and the gentleness of the slope. Users can select one or multiple options to fully express their requirements for cycling surface conditions. The landscape type dimension includes multiple primary categories such as natural landscapes, historical sites, urban landscapes, and rural scenery. Each primary category has secondary subcategories, such as natural landscapes including river views, lake views, mountain views, and forests, and historical sites including historical districts, cultural relics sites, and ancient villages, catering to different user sightseeing preferences. The travel duration dimension allows users to input specific minutes or select preset duration levels, including commonly used options such as half an hour for short trips, one hour for medium trips, half a day for medium-long trips, and full day for long trips. Users can also indicate whether rest stops are included, providing a benchmark for subsequent duration calculations.

[0025] S102 uses natural language processing (NLP) technology to parse non-standardized text input from users and extract keywords corresponding to four core dimensions. For standardized options directly selected by the user, the corresponding tags are directly extracted and included in the keyword set. For natural language content obtained from text input or speech transcription, word segmentation and part-of-speech tagging are performed first. A pre-defined stop word list is loaded to remove meaningless words such as interjections, auxiliary words, conjunctions, and exclamations. Then, semantic matching is performed based on a cycling-specific thesaurus, and each word is broken down and categorized into the semantic category of the corresponding dimension. For example, if a user inputs "I want to ride on a flat road along the river in the city for about two hours, preferably with good scenery," word segmentation is performed to extract the cycling area dimension for "city," the landscape type dimension for "riverside," the road preference dimension for "flat road," and the travel time dimension for "two hours." "Good scenery" is then mapped to the quality level sub-dimension of the landscape type, completing the multi-dimensional extraction of core keywords. For input content containing ambiguous expressions, semantic inference is performed by combining contextual information and domain common sense to cover the user's expressive intent as much as possible; when there are polysemous expressions, all possible dimension mapping results are retained for further verification in the subsequent tag matching stage.

[0026] S103. The extracted keywords are compared with the preset geographic region tag library, road type tag library, landscape classification tag library, and time interval tag library to calculate the semantic similarity between the keywords and each tag. The preset tag library is constructed using a hierarchical structure, with several second-level tags under each first-level tag. All tags are associated with standardized attribute parameters to ensure the consistency of subsequent filtering logic. Each tag in the geographic region tag library is associated with a corresponding latitude and longitude boundary range, which can be directly converted into spatial filtering conditions; each tag in the road type tag library is associated with parameters such as road surface material, slope range, traffic difficulty, and degree of motor vehicle interference; each tag in the landscape classification tag library is associated with the corresponding type of point retrieval range, priority weight, and landscape level coefficient; the time interval tag library is divided into multiple fixed time intervals, each interval corresponding to different cycling distance reference ranges and route splicing strategies. Semantic similarity calculation is achieved using a pre-trained word vector model. For explicit place names, road names, and proper nouns, the corresponding tags are matched precisely by string, and the similarity score is recorded as the maximum. For descriptive and modifying words, cosine similarity is calculated using word vectors. Tags with a score greater than 0.7 are included in the candidate set, and the tag with the highest similarity is finally selected as the matching result, ensuring that words with the same meaning in different expressions can be accurately mapped to the corresponding standard tags.

[0027] S104 selects the tag with the highest semantic similarity as the matching result for each dimension, generating geographical area constraints, road type constraints, landscape type constraints, and duration constraints respectively. These constraints are then integrated into a unified scene matching condition, serving as the filtering basis for subsequent route retrieval and splicing. When the semantic similarity of a keyword in a certain dimension is close to that of multiple tags, and the difference between the highest and second-highest scores is less than 0.1, multiple tags are simultaneously included in the matching condition. An "OR" logic is used to expand the scope of subsequent route filtering, avoiding the omission of suitable route segments due to single tag matching bias. The integrated scene matching condition includes four core constraint items: spatial range constraints, road surface attribute constraints, landscape tendency constraints, and duration upper limit constraints. The matching confidence of each constraint item is also recorded, providing a basis for the tiered strategy of subsequent route segment filtering. After the scene matching condition is generated, it undergoes format standardization processing to ensure full compatibility with the retrieval interface of the baseline route material library, allowing it to be directly used as retrieval parameters.

[0028] The above steps enable the parsing of user preferences and the generation of scenario matching conditions. This transforms diverse user expressions and interactions into standardized filtering constraints, eliminating the need for users to manually set numerous complex path parameters and effectively lowering the operational threshold for ordinary users. A unified tag mapping system ensures the consistency and accuracy of requirement parsing, providing a precise matching benchmark for subsequent retrieval and filtering of the baseline route resource library. This improves the efficiency of matching routes with user needs from the source of requirements, reduces the computational burden of invalid routes, and lays the foundation for lightweight and efficient route planning.

[0029] S200, based on a pre-set baseline route material library, calculates an adapted cycling route using an improved path planning algorithm, and splits, reassembles, and topologically stitches standardized route segments in the baseline route material library. The baseline route material library stores standardized mature cycling route segments.

[0030] In this embodiment of the application, the specific execution process of S200 is as follows: S201 involves constructing a baseline route resource library, completing the segmentation, attribute labeling, and topological indexing of mature cycling routes, and storing them uniformly. Data from various mature cycling routes is collected, including officially recommended routes, high-traffic user cycling tracks accumulated on the platform, verified mature leisure cycling paths, and classic routes shared by professional cycling organizations. Data from different sources corresponds to different confidence levels, with routes released by official sources and professional organizations having the highest confidence levels and being prioritized as core materials. Each collected mature cycling route is segmented, with segmentation nodes uniformly set at route intersections, landscape boundaries, and road type changes to ensure that the segmented route segments maintain consistent attributes in terms of road surface type, landscape attributes, and road grade, generating standardized route segments of appropriate length. The length of each segment is controlled within a reasonable range based on the characteristics of the area. In urban built-up areas with dense road nodes and rapid landscape changes, the length of each segment is controlled between 1 and 3 kilometers to ensure flexibility in splicing; in suburban areas with sparse road nodes and strong landscape continuity, the length of each segment is controlled between 3 and 8 kilometers to reduce splicing and improve planning efficiency. During the cutting process, short, fragmented road segments are avoided. Fragmented road segments with a length of less than 500 meters after cutting are merged into adjacent segments of the same attribute to avoid an excessive number of segments that would increase the burden of retrieval and splicing.

[0031] Each standardized route segment is labeled with corresponding attribute parameters, including cycling area tags, road type tags, landscape type tags, segment length, and average cycling time. All labeled parameters are consistent with the standard format of the corresponding tag library. The cycling area tag matches the corresponding entry in the geographic area tag library, clearly indicating the geographic region and administrative scope to which the segment belongs. The road type tag labels information such as road surface material, average gradient range, difficulty level, and degree of motor vehicle interference, fully reflecting the cycling conditions of the road segment. The difficulty level is divided into four levels according to gradient and road conditions, corresponding to suitable groups of people with different cycling abilities. The landscape type tag labels the category and level of the landscape along the route. The landscape level is divided into different levels according to the number of points, viewing value, and user reviews. The higher the level of the segment, the higher the weight in the route recommendation. The segment length is accurately recorded in kilometers, including the actual cycling distance and cumulative elevation gain of the segment. The average cycling time is calculated based on a large amount of historical cycling data of similar road segments, calculated according to the average cruising speed of ordinary cyclists, while distinguishing different speed coefficients for flat roads, uphill, and downhill, as the basis for route duration estimation.

[0032] A topology index is established for each standardized route segment, recording the start and end coordinates of each segment, as well as the numbers of directly connectable adjacent segments. The criteria for direct connection are that the distance between the endpoint coordinates of two segments is less than a preset connection threshold, and the road direction is continuous without significant backtracking. The connection threshold is dynamically adjusted based on the road density of the area. In urban areas with dense roads and small endpoint spacing, it is set to 50 meters to ensure connection accuracy; in suburban areas with sparse roads and large endpoint deviations, it is set to 200 meters to improve splicing error tolerance. Two segments with endpoint distances within the threshold range and a direction angle of less than 30 degrees are considered directly connectable. The topology index also records the connection attributes of adjacent segments, including whether the connection point needs to cross an intersection, the type of intersection, whether there is a ramp transition, and whether a short detour connecting path is required. This information is used for subsequent route smoothing and time correction during splicing. The topology index of each segment is updated synchronously with the segment attributes. When a new segment is added, it automatically matches surrounding connectable existing segments, updating the corresponding index relationship bidirectionally to ensure the integrity of the topology network.

[0033] Standardized route segments with annotations and their corresponding topological indexes are stored uniformly in the baseline route resource library. The resource library employs a geographically partitioned storage architecture, dividing storage units by city districts to improve retrieval response speed within a region. It supports multi-dimensional searches by tags, coordinates, numbers, etc., and allows for single-condition or multi-condition filtering. The resource library has a regular data verification mechanism, connecting to urban road management and municipal construction data sources to promptly remove data segments corresponding to invalid or closed road sections, and synchronously update information on segments with changed road attributes, ensuring the timeliness and validity of the data within the library.

[0034] S202 filters route segments based on scene matching conditions, generates candidate routes through topological stitching, and performs smoothing to obtain a suitable cycling path. Based on the generated scene matching conditions, a standardized subset of route segments with matching tags is selected from the baseline route material library, and a hierarchical filtering strategy is used to ensure matching quality. Segments that match all three types of tags are included in the core segment subset as the first choice for route stitching; segments that match two types are included in the first-level alternative subset and are used as supplementary materials when the core segments cannot form a complete route; segments that match one type are included in the second-level alternative subset and are only used when there are insufficient alternative materials. When stitching the route, segments within the core segment subset are used firstly for combination. If the core segments cannot form a complete route that meets the time requirement, the first-level and second-level alternative subsets are used in turn to supplement, so as to improve the tag matching degree of the route as much as possible while meeting user needs.

[0035] The system extracts segments from the standardized route segment subset whose starting point coincides with the user-specified starting point as the starting segment, and segments whose ending point coincides with the user-specified ending point as the ending segment. If the user does not specify a clear ending point, segments with high matching degrees are selected as candidate ending segments based on the starting point and the corresponding cycling distance range for the trip duration. This generates basic circular route combinations in both clockwise and counterclockwise directions to meet the user's leisure cycling needs where the starting point is also the ending point. If the user-specified starting point or ending point does not directly fall on a segment endpoint in the database, the nearest segment endpoint is selected as the connection point. A short connecting path from the starting point to the connection point is added along existing slow-traffic roads and included in the overall route mileage and time calculation.

[0036] Between the starting and ending segments, intermediate segments that can be sequentially connected are retrieved based on a topological index, forming multiple candidate routes that meet the user's available riding time requirements. The retrieval employs a heuristic path search algorithm based on a topological index, expanding segment by segment along the connection relationships and accumulating the average riding time of each segment to control the total time within a preset fluctuation range of the user's available riding time. The fluctuation range of the total time is set to ±10% of the user's available time; the shorter the available time, the smaller the fluctuation range. For short-distance rides under half an hour, the fluctuation range is controlled within ±5% to ensure the accuracy of short-distance riding time estimation. Pruning rules are implemented during the retrieval process to avoid invalid paths such as repeated segment usage, extensive route reversals, and severely excessive cumulative mileage. Segments in the same direction are not repeatedly included in the same candidate route. The tag matching degree of each candidate route is calculated simultaneously during the retrieval process; candidate routes with higher matching degrees are ranked higher for easier subsequent solution selection.

[0037] The process involves smoothing the transitions between candidate routes, adjusting the path direction at the junctions of adjacent segments, eliminating duplicate and turnaround sections, and correcting the length and time statistics of the transitions to generate a continuous and complete adaptive cycling path. The smoothing process includes adding short-distance connecting paths, optimizing intersection turning paths, and merging overlapping short-distance segments to ensure the continuity of the entire route. For special sections such as ramps, traffic light intersections, and pedestrian crossings at transitions, the estimated time for those sections is adjusted accordingly. For example, a fixed waiting time is added at traffic light intersections, and cycling speed is adjusted based on the gradient coefficient on uphill sections to improve the accuracy of time calculations. After processing, the start and end points and total time of the entire route are verified to ensure they meet the user's basic needs. Routes that pass verification are included in the candidate route set.

[0038] S203 collects actual user cycling trajectory data, identifies unrecorded route segments, and dynamically updates the baseline route database. It collects actual cycling trajectory data after the user completes the ride, extracting parameters such as the start and end points, route segments, average riding time, and speed changes. During collection, the raw trajectory data undergoes noise reduction processing, employing a sliding window averaging method combined with speed threshold judgment to eliminate abnormal points caused by positioning drift, correct the route direction, and exclude invalid data segments such as user stops or deviations from the route, ensuring the accuracy and reference value of the trajectory data. Trajectories actively marked as "high-quality routes" by the user are given higher weight and prioritized for inclusion in the database update data source; unmarked ordinary routes undergo validity verification, excluding invalid routes that deviate from the normal road network or have abnormal speeds, before being included in subsequent comparison processing.

[0039] The collected actual cycling trajectories are compared with standardized route segments already existing in the baseline route database, and the spatial overlap between each segment of the trajectory and the segments in the database is matched. The overlap is calculated based on the degree of spatial matching between trajectory points and segment paths. A trajectory point falling within 20 meters on either side of a segment path is considered an overlap. A single trajectory segment with an overlap of less than 70% is identified as a new route segment not yet included in the database. During the comparison process, the trajectory is split according to the unified segment cutting nodes in the database, and each segment is matched with existing segments in the database to distinguish between included segments and newly added segments. At the same time, the actual cycling time data of existing segments is recorded to update and correct the average time parameters of the segments.

[0040] Newly identified route segments are automatically segmented and labeled, with segmentation rules consistent with those of existing segments in the resource library. Segmentation nodes are also set at intersections, landscape boundaries, and road type changes. Based on surrounding road network data, POI data, and label information of adjacent existing segments, a label propagation algorithm automatically labels the new segments with corresponding tags such as cycling area, road type, and landscape type. Simultaneously, combining actual cycling time data and referencing historical statistics of similar road segments, the average cycling time for the segment is calculated, generating new standardized route segments that meet the format requirements of the baseline route resource library. After labeling, data verification is performed to ensure the accuracy of labels and parameters. New segments that pass verification are added to the queue for inclusion in the library. When the same road segment is repeatedly verified by different users' cycling trajectories a preset number of times, it is officially included in the baseline route resource library. A topological index is established for the new standardized route segments, matching them with connectable existing segments and updating the topological connections of corresponding segments bidirectionally, completing the dynamic update of the resource library. As user cycling data continues to accumulate, the coverage and adaptability of the resource library continuously expand, providing users with more high-quality route options.

[0041] By using a baseline route library composed of pre-set standardized route fragments, cycling routes are generated through fragment splitting and recombination. This changes the traditional planning model that relies on real-time computation across the entire road network, significantly reducing the computational complexity and load of route planning. It enables rapid response even on low-powered terminals and in high-concurrency scenarios. Furthermore, all fragments are derived from proven, mature cycling routes, ensuring the reliability of the generated routes and the cycling experience, avoiding issues such as unreasonable or unpleasant sections common in purely algorithm-generated routes. The establishment of a topology index further improves the efficiency and accuracy of fragment splicing. Combined with hierarchical filtering and smoothing mechanisms, the route generation process is both efficient and controllable, with output results that are both reasonable and adaptable. The dynamic update mechanism of the resource library forms a data closed loop, continuously absorbing actual user cycling data to optimize and expand the resource library, achieving iterative improvements in the solution's capabilities.

[0042] After the standardized route segments are split, recombined, and topologically stitched together, S300 embeds cultural and tourism check-in points, rest stops, and public service facilities along the route. When embedding, points that are no more than a preset threshold away from the route are selected, and the time spent at a single point does not exceed a preset proportion of the total travel time.

[0043] In this embodiment of the application, the specific execution process of S300 is as follows: S301 uses the completed cycling route as a baseline to retrieve cultural and tourism check-in points, rest stops, and public service facilities within a preset distance range on both sides of the route, forming a candidate point set. Point data is divided into three main categories: cultural and tourism check-in points include natural landscapes, historical sites, unique check-in points, and popular tourist attractions; rest stops include official cycling rest stops, convenient rest areas, convenience stores, and beverage shops; and public service facilities include public restrooms, drinking fountains, shared bicycle parking areas, and first aid stations. Preset distance thresholds are set according to point type: the distance threshold for cultural and tourism check-in points is set to 1 kilometer, allowing users to reach them via short connections, balancing point richness and cycling continuity; the distance threshold for rest stops and public service facilities is set to 500 meters, ensuring users can reach them without significant detours, meeting immediate service needs. The search uses spatial indexing technology to expand segment by segment along the cycling route, ensuring that all points along the route that meet the distance requirements are included in the filtering scope. At the same time, it records the shortest connection distance between each point and the route and the direction of the connection route.

[0044] S302 calculates the recommended dwell time based on the type and scale of each candidate site, combined with historical statistical data. Different types of sites have different baseline dwell times: large cultural and tourism attractions and distinctive historical sites have longer recommended dwell times, typically between 30 and 60 minutes; small viewing platforms and roadside photo spots have shorter recommended dwell times, typically between 5 and 15 minutes; rest stops have standard dwell times set according to basic rest and water replenishment needs, typically between 5 and 10 minutes; and public service facilities have corresponding dwell times set according to basic usage needs, typically between 2 and 5 minutes. Dwell time also references average user dwell time data for similar sites, setting different correction coefficients for weekdays, weekends, and holidays. During peak tourist seasons on holidays, the estimated dwell time is appropriately increased to align with the actual usage habits of most users. The recommended dwell time for each site is updated synchronously with the site's attributes to ensure the reasonableness of the estimated duration.

[0045] S303 sorts candidate points according to the cycling route and calculates the total cycling time after embedding each point. The total cycling time includes both the time spent cycling along the route and the time spent at each point. During point embedding, the length and corresponding time of the connecting routes are calculated simultaneously, with the time spent on connecting routes calculated based on walking speed and included in the total time statistics. During the embedding process, it is verified that the time spent at a single point does not exceed a preset proportion of the total trip time. This proportion is set at 15% of the total time to prevent a single point from occupying too much of the trip time, ensuring the cycling as the core of the trip and preventing the overall trip rhythm from being excessively diluted. Large points exceeding this proportion are automatically marked as optional points for users to manually select whether to include later.

[0046] S304 verifies the total cycling time after embedding all candidate points. If the total time exceeds the user's available time, points are eliminated in order of landscape value priority from low to high. The total time is recalculated after each eliminated point until the total time meets the requirements. Landscape value priority is set comprehensively based on parameters such as point level assessment, user attention, density of similar points in the surrounding area, and scarcity. High-level, high-attention, and scarce core points are retained first, while low-level, highly homogeneous, and low-attention secondary points are eliminated first. If the time still exceeds the limit after eliminating all low-priority points, the suggested stay time for the retained points is further reduced to control the total time while ensuring the inclusion of core points. Finally, a complete cycling route with embedded points is generated, and the location, type, suggested stay time, and connecting route information of each point are simultaneously marked in the route for easy viewing and selection by the user.

[0047] By strategically embedding cultural and tourism attractions and supporting facilities along the route after it has been pieced together, experiential and service support needs are integrated into the core of route planning, rather than being separately marked as supplementary information. This achieves a deep integration of cycling transportation with cultural and tourism experiences and service guarantees. The dual constraints of distance to and duration of stops ensure accessibility while preventing disruption to the itinerary. This maximizes the richness of the cycling experience without exceeding user time constraints, enhancing the overall value and attractiveness of the route and better meeting the core demands of leisure and sightseeing cycling scenarios.

[0048] The S400 automatically generates multiple differentiated itinerary plans based on the user's available riding time and needs, focusing on different dimensions such as route maturity, landscape richness, and riding time, and prioritizing the output of material library matching and recommendation results.

[0049] In this embodiment of the application, the specific execution process of S400 is as follows: S401 calculates route maturity, landscape richness, and gradient gentleness scores for each candidate route. All three scores are standardized to a percentage scale to eliminate dimensional differences and facilitate subsequent weighting calculations. The route maturity score is calculated based on parameters such as the number of historical rides, user favorites, positive review rate, and negative feedback rate of the corresponding standardized route segment. The more times a segment is used by users, the higher its favorites, and the higher its positive review rate, the higher its maturity and reliability, resulting in a higher score. Segments with numerous negative feedbacks, such as poor road conditions or extensive construction, will have points deducted. The landscape richness score is calculated based on the number of landscape type tags, landscape level, and landscape diversity of the corresponding segment. The more diverse the landscape types along the route, the higher the level of the locations, and the more evenly distributed the landscapes, the higher the score; segments with a single landscape type will have a lower score. The gradient gentleness score is calculated based on a weighted average gradient, maximum gradient, and cumulative elevation gain of the corresponding segment. The lower the average gradient, the gentler the undulations, and the lower the cumulative elevation gain, the higher the score; segments with continuous steep slopes will have a lower score.

[0050] S402 generates multiple differentiated itinerary plans based on the different weightings of the three ratings, each targeting different user groups and usage scenarios. The first plan prioritizes route maturity (60%), with the other two ratings each accounting for 20%. This prioritizes route maturity and reliability, ensuring road conditions and accessibility have been extensively validated by users, and that the route has relatively complete supporting facilities. It is suitable for novice riders with limited cycling experience, families, and riders in complex weather conditions such as rain. The second plan prioritizes landscape richness (60%), with the other two ratings each accounting for 20%. This prioritizes the scenic experience along the route, connecting as many high-quality cultural and tourism sites as possible, providing a rich visual experience. It is suitable for users primarily seeking leisure and sightseeing, tourists, and weekend leisure cycling. The third plan prioritizes gradient gentleness (60%), with the other two ratings each accounting for 20%. This prioritizes a gentle ride, reducing steep slopes and undulating sections, and ensuring even exertion. It is suitable for users with average fitness levels, middle-aged and elderly cyclists, rehabilitation cyclists, and those primarily using the route for commuting. Each plan selects the candidate route with the highest score as the recommended option for that dimension. In addition to the three differentiated plans, a balanced plan is generated simultaneously, with each of the three score categories weighted one-third, serving as the default recommended option for the best overall experience.

[0051] S403 verifies the total duration of multiple generated differentiated itinerary plans, eliminating those exceeding the user's available riding time. If the optimal plan for a certain dimension exceeds the time limit, the next best plan for that dimension is selected as a replacement. If the next best plan still does not meet the requirement, some low-value points are eliminated based on priority within that dimension, compressing the stop time to meet the total time requirement, ensuring that all output plans comply with the time constraint. Plans that meet the requirements are retained and output in a preset priority order, with the balanced plan having the highest overall matching degree in the resource library being output first by default. The remaining plans are arranged in order of their respective dimension characteristics. Each plan simultaneously displays core parameter tags, including total mileage, total time, difficulty level, and key highlights. Users can quickly identify the characteristics of different plans in the plan list and select the corresponding itinerary plan according to their immediate needs.

[0052] By generating differentiated itinerary plans through multi-dimensional weight allocation, this approach covers diverse needs across different users and scenarios, avoiding the problem of a single plan failing to meet various demands. Users can obtain route options that match their preferences without manually adjusting numerous parameters. Prioritizing the output of the plan with the highest matching degree from the resource library ensures both the maturity and reliability of the routes and alignment with core user preferences. The parallel output mode of multiple plans effectively improves the adaptability and freedom of choice in route planning, meeting the needs of users with different cycling abilities and travel purposes, and significantly increasing user satisfaction with the planning results.

[0053] The S500 obtains the user's real-time location and makes fine adjustments based on the real-time passenger flow status of the location. When the passenger flow at a location exceeds the preset passenger flow value, it adjusts the order of stops or replaces alternative locations. It also supports users to manually adjust the route, add or remove stops along the way, modify the riding time, and provide real-time navigation along the route.

[0054] In this embodiment of the application, the specific execution process of S500 is as follows: The S501 system acquires real-time visitor density data for various cultural and tourism check-in points and rest stops embedded in the cycling route. Data sources include real-time visitor flow statistics interfaces from site management, visitor flow heatmap data from the city's public service platform, and real-time location aggregation data from platform users. Multi-source data fusion and verification improve data accuracy. The data update frequency is dynamically adjusted based on the cycling scenario: urban sites experience rapid changes in visitor flow, updating every 15 minutes; suburban sites experience slower changes, updating every 30 minutes. This ensures data timeliness while reducing data interaction and lowering terminal power consumption. Visitor density is calculated as the ratio of the number of people instantly present at a site to its carrying capacity; a higher ratio indicates denser visitor flow.

[0055] S502: When the visitor density at a location exceeds a preset threshold, the estimated waiting time for that location is calculated based on the current visitor density and the location's service rate. The preset threshold is set according to the type and size of the location; popular large attractions have higher thresholds, while smaller rest areas have lower thresholds, aligning with the actual carrying capacity of different locations. If the estimated waiting time exceeds the maximum allowed stay time for a single location, it indicates that the current visitor flow at that location is severely impacting the experience. The system then searches the location database associated with the baseline route for alternative locations within a 3-kilometer radius of the current location. Alternative locations with lower visitor density and similar scenic quality are prioritized, while also considering the connection distance between the alternative location and the original route; shorter connection distances result in higher priority. If no suitable alternative location of the same type is found within a 3-kilometer radius, the search range is gradually expanded to 5 kilometers, or the scenic quality requirement is appropriately lowered to select the next best alternative, ensuring the replaceability of the location as much as possible.

[0056] S503 replaces the original stop with an alternative stop, recalculates the total riding time and route, adjusts the stop order based on the stop location, and generates a fine-tuned riding route. If there are no suitable alternative stops within the range, the stop order of that stop during the trip is adjusted, moving it to a relatively low-traffic period based on passenger flow trends; if adjusting the order still does not improve the experience, the stop at that stop is directly cancelled to ensure the overall trip duration is controllable. After the stop adjustment is completed, navigation information is updated synchronously, and a route adjustment prompt is pushed to the user, explaining the reason for the adjustment and the details of the adjusted route. The navigation path is updated after the user confirms.

[0057] The S504 allows users to manually adjust route direction, add or remove points along the way, and modify cycling time. Users can drag route nodes on the map to adjust the route, search for and add points along the way or delete existing points, and modify the total cycling time to trigger the system to regenerate a corresponding plan. After the user completes the operation, the system recalculates various parameters of the route in real time, updates the time, difficulty, points, and other information of the corresponding trip plan, and verifies whether the adjusted route is feasible. If there are impassable sections, the system will promptly notify the user. During the ride, real-time navigation is provided along the route, with real-time broadcasts of turning prompts, point arrival reminders, road condition change prompts, and gradient warnings. It supports both voice broadcasts and interface prompts. If the user deviates from the route, the system automatically replans connecting routes to ensure that the user completes the ride smoothly.

[0058] The S505 connects to the user's smart wearable device via Bluetooth to acquire real-time heart rate data during cycling. The smart wearable device includes a smartwatch, heart rate monitor, and other cycling accessories that can monitor heart rate. It uses the Bluetooth Low Energy protocol for connection, transmitting heart rate data every 5 seconds after connection, maintaining low power consumption during data transmission to minimize impact on the wearable device's battery life. A preset heart rate safety threshold is provided, set based on the user's age, basic fitness level, and other parameters. The threshold can be automatically generated by the user manually entering personal information, or it can be customized by the user. When the user's real-time heart rate exceeds the corresponding heart rate safety threshold for more than 30 seconds, interference from momentary heart rate spikes is excluded, and the user's current exercise intensity is determined to be too high. At this point, the system locates the user's current position and searches the baseline route library for low-intensity standardized route segments within a 5km radius of the current location, along the direction of travel, with an average gradient of less than 2%. During the search, segments that can replace unridden sections are prioritized, keeping the route endpoint as consistent as possible.

[0059] The system replaces unridden high-intensity sections of the current route with retrieved low-intensity standardized route segments, recalculates the total riding time and route connection points, and smooths out these connections to ensure route continuity. If low-intensity segments cause the total time to exceed the user's preset range by more than 10%, the dwell time at subsequent non-core points is appropriately reduced to control the fluctuation range of the total time. Simultaneously, the suggested dwell time at rest stops along the route is adjusted, extending the user's stay at rest stops near low-intensity route segments to allow sufficient time for recovery, generating a fine-tuned riding route. After the adjustment is complete, a notification is sent to the user explaining the reason for the route adjustment and the adjusted route information. Once the user confirms, the navigation route is updated to ensure the user's physical safety and riding experience.

[0060] By dynamically adjusting route points and stop arrangements based on real-time passenger flow, the system effectively avoids the decline in experience caused by peak passenger flow, ensuring a consistent riding experience. The user-manual adjustment function gives users the space to optimize their routes and adapt to changes in temporary needs during the ride. The heart rate-linked adjustment function further expands the scope of real-time fine-tuning, extending from the experience level to the safety level, comprehensively ensuring comfort and safety during the ride, and improving the real-time adaptability and scenario coverage of route planning.

[0061] Based on the solutions in the embodiments of this application, an example is provided in the actual application scenario of a riverside city's cultural tourism cycling service platform. This city boasts hundreds of kilometers of riverside greenways and numerous cultural and tourism resources, with a large number of daily leisure cycling users. The platform has over one million registered users, and the monthly request volume for cycling route planning exceeds one hundred thousand, with weekends and holidays being peak periods. Traditional route planning methods rely on real-time calculations of the global road network. Each planning operation requires traversing hundreds of thousands of road nodes across the entire city, placing a heavy burden on server computing power. During peak hours, the average time for a single route planning operation exceeds 3 seconds. Furthermore, the generated routes often prioritize the shortest distance or shortest time, resulting in insufficient matching between cultural and tourism attributes and user preferences. Consequently, the actual cycling rate after users select a route is low, leading to low route satisfaction.

[0062] Using the solution in this application embodiment, the user's input preference tags are first parsed in step S100 to generate corresponding scene matching conditions, accurately capturing the user's regional, road surface, landscape, and duration requirements. Whether the user selects options or inputs natural language, they can quickly express their needs without manually adjusting multiple complex parameters, significantly lowering the operational threshold and improving the completeness of the expressed needs. Then, relying on the pre-built baseline route material library in step S200, standardized route segments are split and recombined to quickly generate basic cycling routes. This eliminates the need for full road network calculations each time, reducing the average calculation time for a single route planning to less than 0.8 seconds, lowering the server's peak computational pressure by more than 60%, and significantly improving route planning response speed, maintaining a stable and smooth service experience even during peak hours.

[0063] By embedding cultural and tourism attractions and supporting facilities along the route through the S300 steps, the cultural and tourism experience and service guarantees are integrated into the core of route planning, rather than simply being marked beside the route, effectively enhancing the richness of the route experience. Combined with the multiple differentiated options generated by the S400 steps, users can choose routes with high maturity, abundant scenery, or gentle slopes according to their own needs, significantly expanding their choices and greatly improving the match between routes and user needs. Data shows that the platform's users have more than doubled the richness of route choices, and users with different preferences can find solutions that suit their needs.

[0064] During the ride, the S500 system adjusts the route based on real-time passenger flow and user heart rate data, avoiding peak tourist areas to enhance the sightseeing experience and adjusting the route intensity according to the user's physical condition to ensure cycling safety. During peak holiday periods, the system's automatic replacement of high-traffic areas effectively alleviates the pressure on popular attractions and prevents users from wasting time queuing. The heart rate-linked adjustment function provides safety for middle-aged and elderly users and novice riders, significantly reducing the risks associated with cycling.

[0065] After deploying this solution on the platform, user satisfaction with the routes can be improved, the actual cycling rate can be increased, and the synergistic effect of connecting cultural and tourism sites can gradually emerge. At the same time, the number of cycling visits to cultural and tourism sites along the route can be increased. The solution not only optimizes the platform's computing efficiency and service capabilities, but also provides users with cycling route services that better meet their personalized needs, while promoting the revitalization and utilization of cultural and tourism resources along the route, and fully adapting to the needs of urban cultural and tourism cycling scenarios.

[0066] This application discloses a cycling route planning system based on tag matching and a baseline route database, referring to... Figure 2 ,include: The tag parsing module 001 parses the user's input preference tags, which include cycling area, road preference, landscape type and travel duration. Based on the preference tags, keywords are extracted through natural language processing and mapped to a preset tag library to generate corresponding scene matching conditions. The path splicing module 002, based on a pre-set baseline route material library, calculates an adapted cycling path using an improved path planning algorithm, and splits, reassembles, and topologically splices standardized route segments in the baseline route material library. The baseline route material library stores standardized mature cycling route segments. The point embedding module 003 embeds cultural and tourism check-in points, rest stations and public service facilities along the route after the standardized route segments are split, recombined and topologically spliced. When embedding, it selects points that are no more than a preset threshold away from the route, and the time spent at a single point does not exceed a preset proportion of the total travel time. The plan generation module 004 automatically generates multiple differentiated itinerary plans based on the user's available cycling time and user needs, focusing on different dimensions such as route maturity, landscape richness, and cycling time, and prioritizing the output of material library matching recommendation results; The dynamic fine-tuning module 005 obtains the user's real-time location and makes fine-tuning based on the real-time passenger flow status of the location. When the passenger flow at a location exceeds the preset passenger flow value, it adjusts the stopping order or replaces the alternative location. It also supports users to manually adjust the route, add or remove stops along the way, modify the riding time, and provide real-time navigation along the route.

[0067] This application also discloses an electronic device, including a processor, wherein the processor runs a program for the cycling route planning method based on tag matching and a baseline route material library as described in any one of the above embodiments.

[0068] This application also discloses a storage medium storing a program for the cycling route planning method based on tag matching and a baseline route material library as described in any one of the above embodiments.

[0069] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A cycling route planning method based on tag matching and a baseline route database, characterized in that, include: The system parses the user's input preference tags, which include cycling area, road preference, landscape type and travel duration. Based on the preference tags, keywords are extracted through natural language processing and mapped to a preset tag library to generate corresponding scene matching conditions. Based on a pre-set baseline route library, and combined with an improved path planning algorithm, suitable cycling routes are calculated. Standardized route segments within the baseline route library are then split, recombined, and topologically stitched together. The baseline route library stores mature cycling route segments that have undergone standardization. The steps of using the pre-set baseline route library and the improved path planning algorithm to calculate suitable cycling routes and split, recombined, and topologically stitched together the standardized route segments include: segmenting the collected mature cycling routes, with segmentation nodes set at route intersections, landscape boundaries, and road type changes, generating standardized route segments of suitable length; labeling each standardized route segment with corresponding cycling area tags, road type tags, landscape type tags, segment length, and average cycling time parameters; establishing a topological index for each standardized route segment, recording the start-point coordinates, end-point coordinates, and directly connectable adjacent segment numbers for each segment; and storing the labeled standardized route segments and their corresponding topological indexes in the baseline route library. The steps of splitting, recombining, and topologically stitching standardized route segments within the library also include: selecting a subset of standardized route segments with matching tags from the baseline route material library based on generated scene matching conditions; extracting segments whose starting point coincides with the user-specified starting point of the ride as starting segments from the subset of standardized route segments, and extracting segments whose ending point coincides with the user-specified ending point of the ride as ending segments; searching for intermediate segments that can be sequentially connected between the starting and ending segments, and combining them to form multiple candidate routes that meet the user's available riding time requirements; smoothing the connection points of each candidate route, removing duplicate and turnaround segments, and generating a suitable riding path; After the standardized route segments are split, recombined, and topologically stitched together, cultural and tourism check-in points, rest stops, and public service facilities along the route are embedded. When embedding, points that are no more than a preset threshold away from the route are selected, and the time spent at a single point does not exceed a preset proportion of the total travel time. Multiple differentiated itinerary plans are automatically generated based on the user's available cycling time and user needs, focusing on different dimensions such as route maturity, landscape richness, and cycling time, and prioritizing the output of material library matching recommendation results; After smoothing the transitions between each candidate route, eliminating duplicate and turnaround sections, and generating a suitable cycling path, the process includes: calculating route maturity score, landscape richness score, and gradient gentleness score for each candidate route. The route maturity score is based on the historical cycling frequency of the corresponding standardized route segment; the landscape richness score is based on the number and level of landscape type tags labeled on the corresponding segment; and the gradient gentleness score is based on the average gradient parameter of the corresponding segment. Based on the different weights of the three scores, multiple differentiated itinerary plans are generated. The first plan has the highest weight for route maturity score, the second plan has the highest weight for landscape richness score, and the third plan has the highest weight for gradient gentleness score. The total duration of the generated differentiated itinerary plans is verified, plans exceeding the user's available cycling time are eliminated, and the plans that meet the requirements are retained and sorted according to a preset priority for output. The system obtains users' real-time locations and makes fine-tuning adjustments based on the real-time passenger flow status of these locations. When the passenger flow at a location exceeds the preset value, the system adjusts the order of stops or replaces alternative locations. It also supports users manually adjusting the route, adding or removing stops along the way, modifying the cycling time, and providing real-time navigation along the route.

2. The cycling route planning method based on tag matching and a baseline route database according to claim 1, characterized in that, Fine-tuning based on real-time visitor flow at specific locations, including adjusting the order of stops or replacing alternative locations when visitor flow exceeds preset values, includes the following steps: Real-time acquisition of passenger flow density data for various cultural and tourism check-in points and rest stations embedded in the cycling route; When the passenger flow density at a location exceeds the preset passenger flow threshold, calculate the estimated waiting time for that location; If the expected waiting time exceeds the maximum allowed stay time for a single location, search for alternative locations of the same type within a preset range from the baseline route material library; Replace the original points with alternative points, recalculate the total cycling time and route direction, and generate a fine-tuned cycling route.

3. The cycling route planning method based on tag matching and a baseline route database according to claim 1, characterized in that, The steps involved in generating the fine-tuned cycling route include: It connects to the user's smart wearable device via Bluetooth interface to obtain the user's heart rate data in real time during cycling. A preset heart rate safety threshold is retrieved. When the user's heart rate is detected to exceed the heart rate safety threshold, the user's current location is located, and a low-intensity standardized route segment with an average slope lower than the preset value within a preset range around the current location is retrieved from the baseline route material library. Replace the high-intensity sections of the current route that have not been ridden with the retrieved low-intensity standardized route segments, and recalculate the total riding time and route connection points; The recommended dwell time at rest stops along the route is adjusted in sync, extending the time users spend at rest stops near low-intensity route segments, and generating a finely tuned cycling route.

4. The cycling route planning method based on tag matching and a baseline route database according to claim 1, characterized in that, After the step of storing the labeled standardized route segments and their corresponding topology indexes in the baseline route material library, the following steps are included: Collect actual cycling trajectory data after the user completes the ride, and extract parameters such as the start point, end point, route segments, and average cycling time of the trajectory; The actual cycling tracks collected are compared with the standardized route segments already existing in the baseline route material library to identify new route segments that have not been included. The system automatically segments and labels the newly identified route segments to generate new standardized route segments that meet the format requirements of the baseline route material library. Establish a topological index for the new standardized route segments, update the segment set and topological connection relationship of the baseline route material library, and complete the dynamic update of the material library.

5. A system for running a cycling route planning method based on tag matching and a baseline route library as described in any one of claims 1-4, characterized in that, include: The tag parsing module parses the user's input preference tags, which include cycling area, road preference, landscape type and travel duration. Based on the preference tags, keywords are extracted through natural language processing and mapped to a preset tag library to generate corresponding scene matching conditions. The path stitching module, based on a pre-set baseline route material library, calculates an adapted cycling path using an improved path planning algorithm, and splits, reassembles, and stitches standardized route segments in the baseline route material library into topological segments. The baseline route material library stores standardized and mature cycling route segments. The point embedding module embeds cultural and tourism check-in points, rest stations and public service facilities along the route after the standardized route segments are split, recombined and topologically stitched. When embedding, points that are no more than a preset threshold away from the route are selected, and the time spent at a single point does not exceed a preset proportion of the total travel time. The plan generation module automatically generates multiple differentiated itinerary plans based on the user's available cycling time and user needs, focusing on different dimensions such as route maturity, landscape richness, and cycling time, and prioritizing the output of material library matching and recommendation results; The dynamic fine-tuning module obtains the user's real-time location and makes fine-tuning based on the real-time passenger flow status of the location. When the passenger flow at a location exceeds the preset passenger flow value, it adjusts the order of stops or replaces alternative locations. It also supports users to manually adjust the route, add or remove stops along the way, modify the cycling time, and provide real-time navigation along the route.

6. An electronic device, characterized in that, Includes a processor, wherein the processor runs a program of the cycling route planning method based on tag matching and a baseline route library as described in any one of claims 1-4.

7. A storage medium, characterized in that, A program storing a cycling route planning method based on tag matching and a baseline route library as described in any one of claims 1-4.

Images