Anchor point triggered physical examination ar propaganda pushing method

Abstract

The present application relates to the field of medical information technology, in particular to a medical examination AR propaganda and education pushing method based on anchor point triggering, which comprises the following steps: obtaining an anchor point template, a propaganda and education content contract, a near field wireless signal clue, a semantic confirmation result, behavior state data and a business event generation evidence record; constructing an evidence interaction graph based on the evidence record and performing consistency determination to obtain an anchor point identifier, an anchor point confidence level and an evidence fingerprint, and determining a window budget and generating a propaganda and education content unit sequence accordingly; when the content contains augmented reality superimposed presentation and the confidence level reaches a threshold, applying a presentation token and executing superimposed presentation, otherwise executing two-dimensional interface presentation and generating a pushing record; generating a closed loop evidence package based on the pushing record, the evidence fingerprint and the business event, which is used to update the evidence reliability parameter and the propaganda and education content contract. The present application reduces the probability of wrong pushing, improves the matching degree of pushing time, and supports continuous self-calibration.

Description

Technical Field

[0001] This invention relates to the field of medical information technology, specifically to a method for pushing AR-based health check-up education based on anchor point triggering. Background Technology

[0002] The high concurrency, strong process constraints, and frequent spatial relocation in health checkup centers mean that if the educational content is not precisely aligned with the examinee's location, current examination stage, and available time, misaligned push notifications, inappropriate timing, redundant information, or omissions can easily occur. This leads to misunderstandings and decreased compliance among examinees, and increases the burden on medical staff with repeated verbal explanations. Common industry practice involves content display driven by a single location or triggering event, lacking multi-source evidence consistency verification and a traceable evidence chain. If environmental interference, business event delays, or personnel movement cause short-term misjudgments, push notifications may occupy attention at inappropriate times, affecting waiting order and examination cooperation. Simultaneously, the lack of a closed-loop feedback mechanism between push notification results and business outcomes makes it difficult to continuously adjust configuration rules according to changes in hospital layout, equipment migration, and changes in patient behavior, resulting in insufficient long-term operational stability. Summary of the Invention

[0003] This invention provides a method for AR-based health check-up education push based on anchor point triggering, which can at least solve the problems of anchor point misjudgment and push timing mismatch caused by inconsistent triggering from multiple sources at the health check site.

[0004] A method for pushing AR-based health checkup information based on anchor points includes the following steps: Acquire anchor point templates, educational content contracts, near-field wireless signal clues, semantic confirmation results, behavioral status data, and business events to generate evidence records; Based on the evidence records, an evidence interaction graph is constructed and consistency is determined to obtain anchor point identifiers, anchor point confidence levels, and evidence fingerprints. Based on the anchor point confidence levels, behavioral state data, and business events, the window budget is determined, and a sequence of publicity and education content units is generated based on the window budget and the publicity and education content contract. When the sequence of propaganda content units includes augmented reality overlay and the anchor confidence level reaches the threshold, a presentation token is requested from the near-field wireless broadcasting device and augmented reality overlay is executed; otherwise, a two-dimensional graphical interface is executed and a push record is generated. A closed-loop evidence package is generated based on push records, evidence fingerprints, and business events, which is used to update evidence reliability parameters and propaganda content contracts.

[0005] In one possible implementation, generating evidence records includes: Time alignment is performed on near-field wireless signal cues, semantic confirmation results, behavioral status data, and service events. Based on field mapping rules, near-field wireless signal clues, semantic confirmation results, behavioral status data, and business events are converted into structured evidence entries; Structured evidence entries include at least a timestamp, evidence type identifier, and anchor point candidate identifier; aggregated structured evidence entries yield evidence records.

[0006] In one possible implementation, constructing the evidence interaction graph includes: An evidence interaction graph is established using structured evidence items as nodes; A consistency relationship is established between the anchor candidate identifiers corresponding to the semantic confirmation results and the anchor candidate identifiers corresponding to the near-field wireless signal cues. Based on the anchor point template, the stage type corresponding to the anchor point candidate identifier is determined, and the process stage identifier is determined based on the business event. Conflict relationships are established, which are used to indicate that the stage type is inconsistent with the process stage identifier.

[0007] In one possible implementation, consistency determination includes: Within a preset time window, anchor point identifiers are determined based on consistency and conflict relationships, and a set of conflict labels is generated; The anchor confidence level is updated hierarchically based on the set of conflicting tags. The anchor confidence level includes at least the arrival level and the push level.

[0008] In one possible implementation, generating evidence fingerprints includes: The set of evidence types is determined based on evidence records. The set of evidence types includes at least near-field wireless signal clue evidence, semantic confirmation result evidence, behavioral state data evidence, and business event evidence. Get the set of conflicting tags; Evidence fingerprints are generated by encoding the evidence type set and conflict label set. The evidence fingerprints are used to indicate the source of evidence and the state of conflict in terms of anchor confidence level.

[0009] In one possible implementation, determining the window budget includes: The stage type corresponding to the anchor point identifier is determined based on the anchor point template; Map phase types to process phase identifiers based on business events; The behavior state category is determined based on the behavior state data, and the behavior state category is either walking state or stationary state. The window type is determined based on the process stage identifier and behavior status category; The budget duration is determined based on the window type, and a budget credibility identifier is generated. The window budget includes the budget duration and the budget credibility identifier.

[0010] In one possible implementation, generating a sequence of mission content units based on a window budget and mission content contracts includes: Based on the mission content contract, the mission content unit is matched and screened. The mission content contract should include at least the set of applicable anchor points, the type of applicable stage, the prohibition conditions, the minimum budget duration, the priority rules and the dependency relationship. A candidate set is obtained when the anchor point identifier belongs to the set of applicable anchor point identifiers, the process stage identifier matches the applicable stage type, the budget duration is not less than the shortest budget duration, and the prohibition condition is not met. Based on the candidate set, a sequence of missionary content units is generated according to priority rules and dependency relationships.

[0011] In one possible implementation, requesting a presentation token from the near-field wireless broadcasting device includes: Micro-area identifiers are determined based on anchor point templates and anchor point identifiers; Send a token request message, which includes at least the micro-area identifier, anchor identifier, and presentation duration parameter; Upon receiving a token authorization message, perform augmented reality overlay rendering and write the token identifier into the push record; Upon receiving a token rejection message, a two-dimensional graphical interface is rendered and the reason for degradation is written to the push record.

[0012] In one possible implementation, generating a closed-loop evidence package includes: By using anchor point identifiers and timestamps as association keys, push records, evidence fingerprints, and business events are linked to obtain a closed-loop evidence package; A closed-loop evidence package should include at least anchor point identifiers, anchor point confidence levels, window budgets, a sequence of educational content units, evidence fingerprints, and business events. The updated evidence reliability parameters include: based on the frequency of occurrence of conflict label sets in the statistical evidence fingerprint of the closed-loop evidence package, the evidence reliability parameters are updated accordingly.

[0013] In one possible implementation, updating the mission content contract includes: The matching rate between the sequence of educational content units and the corresponding process stage identifiers of business events is calculated based on the closed-loop evidence package. In response to a matching rate falling below a preset threshold, update at least one of the prohibition conditions and minimum budget duration of the missionary content contract, and generate a contract version identifier; issue the updated missionary content contract based on the contract version identifier.

[0014] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows: By employing evidence interaction graphs and consistency judgment techniques, verifiable fusion of near-field signals, semantic confirmation, behavioral states, and business events is achieved, reducing erroneous pushes caused by anchor point misjudgments. Through window budget and content contract matching techniques, objective constraints on push duration and content arrangement are implemented, reducing interference in mobile scenarios and improving information capacity during waiting periods. By employing presentation tokens and micro-zone concurrency control techniques, resource controllability and degradation traceability of augmented reality overlay presentations are achieved. Through closed-loop evidence packages and reliability parameter update techniques, continuous calibration of configuration and evidence weights is achieved, enhancing long-term operational stability and maintainability. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the execution flow of the method of the present invention; Figure 2 This is a schematic diagram of the F1 index for anchor point identification in a specific embodiment of the present invention; Figure 3 This is a schematic diagram of the window budget duration distribution in a specific embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the ratio of augmented reality to two-dimensional representation in a specific embodiment of the present invention; Figure 5 This is a schematic diagram illustrating the changing trend of the closed-loop index in a specific embodiment of the present invention. Detailed Implementation

[0016] The embodiments of the present disclosure will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the disclosure. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the present disclosure for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts of the present disclosure.

[0017] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0018] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0019] Anchor point triggering transforms the "location and process stage of the examinee" into calculable and verifiable triggering conditions, thereby driving the presentation of educational content at the appropriate time and on the appropriate medium. In hospital physical examination scenarios, anchor points can be a combination of spatial and semantic anchor points: spatial anchor points reflect the factual basis of the examinee entering a certain functional micro-area, while semantic anchor points reflect the confirmation result of the object or identification information currently faced by the examinee. Anchor point triggering is not simply location hit or QR code triggering, but rather it converges multi-source clues into unified evidence and completes consistency judgment. Then, using anchor point identification and anchor point confidence level as the core, it drives subsequent window budgeting, content arrangement, and presentation control, transforming educational push from "passive display" to "evidence-driven active triggering." Based on this, this invention constructs a complete process around evidence recording, evidence interaction diagrams, consistency judgment, window budgeting, presentation tokens, and closed-loop evidence packages to achieve stable triggering and sustainable calibration of augmented reality educational content in physical examination scenarios.

[0020] like Figure 1 As shown, a method for pushing AR-based health checkup education based on anchor point triggering includes the following steps: Acquire anchor point templates, educational content contracts, near-field wireless signal clues, semantic confirmation results, behavioral status data, and business events to generate evidence records; Anchor point templates are used to define the correspondence between spatial identifiers and business semantics for each physical examination site within the hospital. Educational content contracts specify the permitted educational content units and their constraints for different examination sites and different business stages. Upon entering the examination area, the terminal first retrieves the anchor point templates and educational content contracts from the configuration service and caches them locally. Simultaneously, it continuously collects near-field wireless signal cues via a near-field wireless module. These cues can originate from Bluetooth beacon broadcast identifiers, Wi-Fi access point identifiers, or near-field communication tag identifiers. When the terminal needs to confirm the semantics of a site, it obtains the semantic confirmation result. This confirmation can be obtained by scanning the site's QR code using a camera or by manually selecting the site name on the interface. The smart wearable device continuously outputs behavioral status data, which is used to distinguish between walking and standing still, forming continuous behavioral segments on the terminal side. Business events are pushed through the business system interface or obtained through terminal polling. Business events include at least events such as project check-in, waiting for examination, entering the examination area, and completing the examination, along with the event occurrence time. The terminal aggregates anchor point templates, propaganda content contracts, near-field wireless signal clues, semantic confirmation results, behavioral status data, and business events into evidence records, and writes the evidence records into a circular cache for direct use in subsequent consistency judgment and point convergence processing.

[0021] In one optional progressive implementation, evidence records are generated under a unified time reference to ensure that data from different sources are comparable within the same time window. The terminal performs time alignment on near-field wireless signal cues, semantic confirmation results, behavioral state data, and business events. Time alignment uses the terminal system clock as the master clock, performs time zone unification and deviation correction on timestamps from business system interfaces, compensates for sampling times from wearable devices based on Bluetooth connection latency, and records near-field wireless signal cues by reception time and appends a reception sequence number.

[0022] Field mapping rules are used to transform data from different sources into the same structure. These rules specify at least the set of values ​​for the evidence type identifier, the encoding method for anchor candidate identifiers, the timestamp precision, and the default handling method for missing fields. The terminal converts near-field wireless signal clues into wireless clue entries, semantic confirmation results into semantic confirmation entries, behavioral status data into behavioral entries, and business events into business event entries, forming a structured set of evidence entries. Each structured evidence entry includes at least a timestamp, an evidence type identifier, and an anchor candidate identifier. The timestamp is the unified time when the entry occurs or is received. The evidence type identifier distinguishes between wireless clues, semantic confirmations, behavioral statuses, and business events. The anchor candidate identifier is a point identifier or a set of point identifiers defined in the anchor template.

[0023] When aggregating structured evidence entries, the terminal merges them into the same evidence record according to a preset time window. It also performs deduplication on wireless clue entries that appear repeatedly within the same time window and have the same anchor point candidate identifier, merges fragments of consecutive behavior entries, and rearranges out-of-order business event entries by timestamp. When semantic confirmation entries are missing, the evidence record still retains both wireless clue entries and business event entries for subsequent steps to initially converge point candidates and trigger supplementary confirmation. The aggregated evidence record outputs structured data organized by time window, serving as the input data source for subsequent evidence interaction graph construction.

[0024] Based on the evidence records, an evidence interaction graph is constructed and consistency is determined to obtain anchor point identifiers, anchor point confidence levels, and evidence fingerprints. Based on the anchor point confidence levels, behavioral state data, and business events, the window budget is determined, and a sequence of publicity and education content units is generated based on the window budget and the publicity and education content contract. After the evidence record is formed, the terminal constructs an evidence interaction graph using the structured evidence items in the record as input, and completes a consistency judgment within a preset time window, outputting anchor point identifiers, anchor point confidence levels, and evidence fingerprints. Subsequently, it determines the window budget by combining the anchor point confidence level, behavioral state data, and business events, and uses the window budget to constrain the selection and arrangement of propaganda content contracts, generating a sequence of propaganda content units. The anchor point identifier, as the unique location result of the physical examination point, participates in the selection of subsequent presentation methods; the anchor point confidence level is used to constrain the trigger boundary of augmented reality overlay presentation; the evidence fingerprint is used to solidify the evidence source and conflict state of this judgment; and the propaganda content unit sequence serves as the direct input for push execution.

[0025] In one possible implementation, the evidence interaction graph adds an operational representation of "nodes and relational edges" to the original evidence records, enabling unified calculation of the support relationship between semantic confirmation results and near-field wireless signal clues, and the conflict relationship between stage types and process stage identifiers. The terminal establishes the evidence interaction graph using structured evidence items as nodes. Node fields include at least a timestamp, evidence type identifier, and anchor candidate identifier, and may also include fields such as wireless signal reception strength, service event type, and behavior state fragment identifier. The evidence interaction graph is preferably stored using a combination of a node table and an adjacency table; the node table stores node fields, and the adjacency table stores the relational edges between nodes.

[0026] A consistency relationship is established between the anchor candidate identifiers corresponding to the semantic confirmation results and the anchor candidate identifiers corresponding to the near-field wireless signal cues: when the anchor candidate identifier given by the semantic confirmation results falls into the set of anchor candidate identifiers given by the near-field wireless signal cues, a consistency relationship edge is written between the corresponding nodes, and the number of hits and the most recent hit time of the consistency relationship edge are recorded for accumulation in subsequent time windows.

[0027] Based on anchor point templates, the stage type corresponding to the anchor point candidate identifier is determined, and the process stage identifier is determined based on business events, establishing conflict relationships: when the stage type and the process stage identifier are inconsistent within the same preset time window, a conflict relationship edge is written between the corresponding nodes, and the conflict type and occurrence time are recorded; the conflict type is used to indicate the category boundary of the inconsistency between the stage type and the process stage identifier, such as stage ahead, stage behind, or stage crossing. To adapt to the jitter of wireless signals within the hospital, near-field wireless signal clue nodes are allowed to be written multiple times within the preset time window and undergo deduplication aggregation. The deduplication rule can be based on the joint key of the anchor point candidate identifier and the evidence type identifier, updating only the most recent hit time and the number of hits; if the semantic confirmation result node appears repeatedly within the preset time window, the latest one is used as the standard, and the previous one is retained for tracing. The evidence interaction graph generates a graph snapshot at the end of each preset time window. The graph snapshot serves as the input for consistency determination, avoiding cross-window mixed calculations that lead to unstable anchor point convergence.

[0028] In one possible implementation, the consistency determination adds a constraint point of "conflict label set and confidence level grading update" to the evidence interaction graph, which is used to ensure that the boundary between pushable and non-pushable is grounded in verifiable process conditions. Within a preset time window, the consistency determination determines anchor point identifiers based on consistency and conflict relationships and generates a conflict label set: the terminal accumulates the number of hits of consistency relationship edges for each anchor point candidate identifier and the number of occurrences of conflict relationship edges, while simultaneously writing the conflict type carried by the conflict relationship edge into the conflict label set; the conflict label set can be deduplicated by conflict type, and the first and last occurrence times of each type of conflict are recorded.

[0029] The selection of anchor point identifiers should preferably follow the principle of "more support, fewer conflicts, and closer time": when multiple anchor point candidate identifiers simultaneously satisfy the consistency relationship hit, the anchor point candidate identifier with a higher consistency relationship hit count should be selected first; when the consistency relationship hit counts are the same, the anchor point candidate identifier with a lower number of conflict relationships should be selected first; when they still cannot be distinguished, the anchor point candidate identifier with a semantic confirmation result timestamp closer to the center of the preset time window should be selected first, so as to reduce misjudgments caused by expired confirmations.

[0030] Anchor confidence levels are updated in a tiered manner based on the conflict tag set. Anchor confidence levels include at least arrival and push levels: when semantic confirmation results exist and the conflict tag set is empty, the anchor confidence level is updated to the push level; when semantic confirmation results are missing or the conflict tag set contains conflict types such as stage skipping or stage lag, the anchor confidence level is maintained or reverted to the arrival level; when the conflict tag set contains only minor conflicts and the business event undergoes a self-consistent transfer within a preset time window, the push level is allowed to be maintained, but the budget confidence indicator is downgraded to limit the presentation duration and interaction intensity in subsequent window budgets. Through this tiered approach, subsequent steps can directly obtain the executable trigger boundaries without repeatedly parsing the evidence interaction graph.

[0031] In one possible implementation, the evidence fingerprint adds a "traceable code" constraint point to the evidence record and conflict label set, enabling the evidence source and conflict state of the anchor confidence level to be stably reused without having to read back the original evidence entries. The terminal determines the evidence type set based on the evidence record, which includes at least near-field wireless signal clue evidence, semantic confirmation result evidence, behavioral state data evidence, and business event evidence, and obtains the conflict label set; the terminal encodes the evidence type set and conflict label set to generate the evidence fingerprint.

[0032] The preferred encoding method is bit-based encoding: a fixed bit is assigned to each type of evidence, and a fixed bit is assigned to each type of conflict, generating an evidence type bitmap and a conflict type bitmap. The evidence type bitmap is used to indicate whether the current anchor point determination simultaneously possesses near-field wireless signal clue evidence and semantic confirmation result evidence, while the conflict type bitmap is used to indicate whether the current anchor point determination has an inconsistency between the stage type and the process stage identifier. The evidence fingerprint can be obtained by concatenating the evidence type bitmap and the conflict type bitmap in a fixed order, with a short checksum appended at the end to reduce storage or transmission errors; the short checksum can be implemented using cyclic redundancy check or hash truncation.

[0033] Evidence fingerprints are used to indicate the source of evidence and the state of conflict for anchor confidence levels: when the evidence fingerprint shows a lack of semantic confirmation evidence, subsequent window budgets can directly set the budget confidence level to medium or low; when the evidence fingerprint shows that the conflict type bitmap has stage jumps, subsequent advocacy content contract matching can prioritize more conservative advocacy content units to avoid triggering inappropriate content when the business stage is uncertain. Evidence fingerprints, along with anchor identifiers and anchor confidence levels, are written into the push context for direct use in subsequent closed-loop evidence package statistics and parameter updates.

[0034] In one possible implementation, the window budget adds engineered constraints—"window type, budget duration, and budget credibility identifier"—to the anchor point identifier and business event, ensuring that push decisions fall within an executable timeframe. The terminal determines the stage type corresponding to the anchor point identifier based on the anchor point template. The stage type describes the semantic category of the examination location within the process, such as check-in area, waiting area, examination area, or departure area. The terminal maps the stage type to a process stage identifier based on the business event. The process stage identifier describes the specific stage the examinee is currently in within the process, with the business event timestamp serving as the stage's effective time.

[0035] The terminal determines the behavior state category based on behavior state data, which is limited to walking or stationary states: walking is determined by the number of steps per unit time exceeding a threshold, and stationary states are determined by the number of steps per consecutive time period falling below a threshold; when behavior state data is missing within a preset time window, the mobile terminal's acceleration stability is allowed as a supplementary source of determination. The terminal determines the window type based on the process stage identifier and the behavior state category. The window type includes at least a short window type, a long window type, and a no-push window type: walking corresponds to the short window type, stationary states in the waiting inspection stage correspond to the long window type, and states in the inspection stage correspond to the no-push window type.

[0036] The terminal determines the budget duration and generates a budget credibility identifier based on the window type. The budget duration can be obtained by looking up a table by window type. The budget credibility identifier can be jointly determined by the anchor confidence level and the evidence fingerprint: a high credibility level corresponds to a push level with no conflicting evidence fingerprints, while a medium or low credibility level corresponds to a push level that has been reached or a conflicting evidence fingerprint exists. The window budget includes the budget duration and the budget credibility identifier, and serves as a hard constraint for the subsequent arrangement of educational content units, preventing content from timed out or being pushed at inappropriate stages.

[0037] In one possible implementation, the propaganda content unit sequence adds constraints such as "contract matching and filtering, priority rules, and dependencies" to the window budget, ensuring that the output sequence satisfies business boundaries and can be presented within the budgeted timeframe. The terminal performs contract matching and filtering on propaganda content units based on the propaganda content contract. The propaganda content contract includes at least a set of applicable anchor point identifiers, applicable stage type, prohibition conditions, minimum budget duration, priority rules, and dependencies. Each propaganda content unit can carry an estimated duration, presentation format identifier, content theme identifier, and preceding dependency identifier.

[0038] The terminal obtains a candidate set when the anchor point identifier belongs to the applicable anchor point identifier set, the process stage identifier matches the applicable stage type, the budget duration is not less than the minimum budget duration, and the prohibition condition is not met. The prohibition condition can be defined based on the process stage identifier, behavior status category, or budget credibility identifier. For example, audio playback is prohibited during the inspection stage, and triggering highly interactive content is prohibited when the budget credibility identifier is low. Based on the candidate set, a sequence of educational content units is generated according to priority rules and satisfying dependencies: The terminal first sorts the candidate set according to priority rules, which can be determined by the content theme, risk level, hospital department strategy, or historical click data; the terminal then performs dependency verification, which can be expressed in the form of "preceding content units must appear first" or "preceding business events must have occurred". Educational content units whose dependencies are not satisfied are not included in the sequence.

[0039] The terminal generates a sequence of missionary content units using a loading method. This loading method allows missionary content units to be added sequentially from high priority to low priority, ensuring the total estimated duration does not exceed the budgeted duration. When the budget confidence level is medium or low, the length of the missionary content unit sequence is forcibly limited, and missionary content units with shorter estimated durations or more conservative presentation formats are prioritized. The generated missionary content unit sequence must contain at least a list of missionary content unit identifiers and a corresponding list of presentation format identifiers, serving as direct input for subsequent presentation token applications and push execution.

[0040] When the sequence of propaganda content units includes augmented reality overlay and the anchor confidence level reaches the threshold, a presentation token is requested from the near-field wireless broadcasting device and augmented reality overlay is executed; otherwise, a two-dimensional graphical interface is executed and a push record is generated. After the missionary content unit sequence is generated, the terminal determines whether it includes augmented reality overlay based on the presentation format identifier carried by each missionary content unit in the sequence, and performs presentation mode distribution based on the anchor confidence level. When the missionary content unit sequence includes augmented reality overlay and the anchor confidence level reaches the threshold, the terminal first requests a presentation token from the near-field wireless broadcasting device. After token authorization, it executes the augmented reality overlay and completes the push record writing. When the anchor confidence level does not reach the threshold or the missionary content unit sequence does not include augmented reality overlay, the terminal directly executes the two-dimensional graphical interface presentation and generates a push record. The push record includes at least the anchor identifier, missionary content unit sequence identifier, presentation mode identifier, start time, end time, and presentation result code, which serve as input for subsequent closed-loop evidence package generation and content contract update.

[0041] In one possible implementation, the presentation token application is used to resolve resource contention and interference issues caused by multiple terminals simultaneously presenting augmented reality content within the same micro-zone. The added limitation is the introduction of a micro-zone identifier and the inclusion of presentation duration parameters in the application message, enabling the near-field wireless broadcasting device to perform unified arbitration based on micro-zones. The micro-zone identifier is mapped using an anchor template, which can divide the area into several micro-zones. Each micro-zone corresponds to one or more anchor identifiers, and the device identifier and communication parameters of the micro-zone broadcasting device are configured. After obtaining the anchor identifier, the terminal queries the micro-zone identifier through the anchor template and binds the micro-zone identifier and anchor identifier to the current push context.

[0042] When sending a token request message, the terminal establishes a short connection with the near-field radio broadcasting device or uses a broadcast backhaul channel to send the request. The token request message includes at least a micro-cell identifier, an anchor identifier, and a presentation duration parameter, and may also include a terminal identifier, request priority, and number of retries. The presentation duration parameter is determined by the budgeted duration of the window budget and can be truncated based on the cumulative estimated duration of the propaganda content unit sequence to ensure that the request duration does not exceed the budgeted duration. After receiving the token request message, the near-field radio broadcasting device locates the token status of the current micro-cell according to the micro-cell identifier and responds by authorizing or rejecting according to the arbitration rules. The arbitration rules may include that only one token can be authorized in the same micro-cell at the same time, the token holding time does not exceed the presentation duration parameter, and low-priority requests can be rejected when high-priority requests arrive.

[0043] Upon receiving a token authorization message, the terminal enters the augmented reality overlay presentation process. The augmented reality overlay presentation can use the real-time view captured by the mobile terminal's camera as the background layer, and load the corresponding 3D model or 2D annotation layer from the resource package according to the anchor point identifier, and overlay the educational guidance on the screen. The overlay position can be aligned according to the anchor point reference point coordinates provided by the anchor point template or the image recognition positioning result. When the anchor point reference point coordinates are unavailable, it is allowed to degenerate into a fixed overlay area with the center of the screen as the reference.

[0044] The terminal writes a token identifier to the push record at the start of the augmented reality overlay presentation and releases the token when the presentation ends or the user actively exits. The release method can be sending a token release message or waiting for the token to time out and be reclaimed. Upon receiving a token rejection message, the terminal executes a two-dimensional graphical interface presentation, which may include text explanations, graphic cards, or short video playback, and writes the reason for degradation to the push record. Degradation reasons include at least token occupation, micro-area arbitration rejection, communication failure, or insufficient anchor confidence level. Through the aforementioned token application and degradation path, the terminal can complete the educational presentation even with limited resources or insufficient evidence, and the push record carries traceable presentation decision-making evidence, facilitating subsequent closed-loop analysis and contract parameter adjustments.

[0045] A closed-loop evidence package is generated based on push records, evidence fingerprints, and business events, which is used to update evidence reliability parameters and propaganda content contracts.

[0046] After the push record is generated, the terminal uses the push record, evidence fingerprint, and business event as input to generate a closed-loop evidence package, and writes the closed-loop evidence package into the closed-loop storage area, serving as a unified basis for updating evidence reliability parameters and propaganda content contracts. The closed-loop evidence package fixes the anchor point determination results, push arrangement results, and business stage results within a push cycle into the same structured object, enabling the system to retrieve the complete context using the same association key during subsequent statistics and reconfiguration, thereby achieving continuous correction of evidence weights and contractual constraints. Evidence reliability parameters are used to describe the stability of different evidence types at different process stages and in different micro-areas within the institution, while propaganda content contracts constrain the pushable boundaries of propaganda content units at specific anchor points and stages. Both are updated through data-driven closed-loop evidence packages.

[0047] In one possible implementation, the generation of the closed-loop evidence package adds explicit association keys, field sets, and missing data handling rules, ensuring that the closed-loop data can still be stably stored in the database and used by subsequent batch processing even under actual network fluctuations and event delays. The terminal uses anchor identifiers and timestamps as association keys to associate push records, evidence fingerprints, and business events to obtain the closed-loop evidence package. The timestamp preferably uses the start time of the push record, and allows for updates to the closed-loop evidence package after the end time of the push record has arrived. Multiple business events may occur within a preset time window; the terminal merges them according to timestamp and event type, including business events whose time periods intersect with the push record into the same closed-loop evidence package.

[0048] The closed-loop evidence package includes at least anchor point identifiers, anchor point confidence levels, window budgets, a sequence of educational content units, evidence fingerprints, and business events. Anchor point identifiers and anchor point confidence levels are used to mark the health check points and credibility levels bound to this push notification; the window budget is used to mark the available budget duration and budget credibility identifier at that time; the educational content unit sequence is used to mark the actual arranged content units and their presentation forms; the evidence fingerprint is used to mark the encoding results of the evidence type set and conflict tag set; and the business event is used to mark the process stage identifier and key event type. To reduce storage overhead, the educational content unit sequence can store only the list of content unit identifiers and the cumulative estimated duration, and the window budget can store only the budget duration and budget credibility identifier. When the evidence fingerprint or business event has not yet arrived within the time window, the closed-loop evidence package allows writing with a missing marker, and records the completion time and source when it is subsequently completed, ensuring that the association key remains unchanged. After the closed-loop evidence package is successfully written, the terminal sets the write status identifier to complete and provides a cursor field for incremental scanning for subsequent statistics, such as the batch number or day sequence number.

[0049] The updated evidence reliability parameters for closed-loop evidence packages include a new statistical definition and parameter adjustment rules for the "frequency of occurrence of conflict label sets," enabling the evidence source and conflict status to be fed back into the consistency calculation of subsequent anchor point determination. The terminal or server traverses the closed-loop evidence packages within a preset statistical period, analyzes the evidence fingerprint to obtain conflict label sets, and groups them by anchor point identifier, process stage identifier, or micro-area identifier to statistically analyze the frequency of occurrence of conflict label sets. The frequency of occurrence can be defined as "the proportion of closed-loop evidence packages containing conflict label sets within that group to the total number of closed-loop evidence packages," and can be further decomposed by conflict type into stage advance frequency, stage lag frequency, and stage skip frequency.

[0050] The evidence reliability parameters include at least evidence type weights and conflict penalty factors: Evidence type weights describe the relative credibility of near-field wireless signal clue evidence, semantic confirmation result evidence, behavioral state data evidence, and business event evidence within the current group; the conflict penalty factor is used to deduct conflicting edges during consistency determination. During updates, the frequency of conflict tag sets is mapped to the conflict penalty factor, and the evidence type weights are adjusted synchronously. For example, when the stage crossing frequency exceeds a threshold, the impact of business event evidence on anchor point convergence is reduced, and the priority of semantic confirmation result evidence is increased; when the wireless clue drift frequency exceeds a threshold, the continuity requirement for near-field wireless signal clue evidence is increased, and its effective time window is shortened. The updated evidence reliability parameters are written to the configuration center in a versioned configuration format. After obtaining the new version, the terminal replaces the local parameters and retains the previous version for rollback, ensuring the traceability of consistency determination behavior.

[0051] In one possible implementation, the updated propaganda content contract adds a calculation method for the "content and stage matching rate" and contract item adjustment rules to the closed-loop evidence package, enabling contractual constraints to gradually converge around the actual business stages. The terminal or server calculates the matching rate between the propaganda content unit sequence and the corresponding process stage identifiers of the business events based on the closed-loop evidence package: for each propaganda content unit in the closed-loop evidence package, its applicable stage type and prohibition clauses are read, and the process stage identifier obtained from the business event parsing is compared to determine whether the applicable stage type is met and the prohibition conditions are not triggered, thus obtaining a matching marker; the matching rate can be calculated as the "proportion of the number of content units with true matching markers and completed presentation to the total number of propaganda content unit sequences," and can be combined with the presentation result code in the push record to exclude content units that failed to be presented.

[0052] In response to a matching rate below a preset threshold, at least one of the prohibition conditions and minimum budget duration in the propaganda content contract is updated: when low matching is mainly concentrated in the inspection stage or prohibited window types, prohibition conditions for the corresponding stages are added, or some presentation forms are restricted to two-dimensional graphical interface presentation; when low matching is mainly concentrated in the short window type in the walking state, the minimum budget duration threshold is increased, or the upper limit of the allowed cumulative estimated duration is lowered, so that short windows prioritize shorter and more conservative content units. Each update generates a contract version identifier, which, along with the effective time and applicable department identifier, is written into the contract configuration, and the updated propaganda content contract is distributed through the configuration center; after receiving the new version, the terminal records the version switch time and writes the contract version identifier into the subsequent push records, so that the closed-loop evidence package can be statistically analyzed by contract version, thereby verifying the change in matching rate before and after the contract update.

[0053] In one specific embodiment, a set of reproducible data is constructed using log field formats consistent with the on-site system of the physical examination center. This data is used to demonstrate the complete execution chain and verifiable statistical criteria from anchor point triggering to closed-loop update. The scenario includes six types of anchor points: information desk posters, ECG signs, blood collection windows, ultrasound waiting screens, CT registration areas, and physical examination form item recognition. Each type of anchor point is configured with an anchor point template, which includes anchor point identifiers, micro-area identifiers, stage types, near-field wireless broadcast device identifiers, a set of available presentation modes, and a default budget range. The educational content contract contains 24 educational content units. Each educational content unit provides an applicable set of anchor point identifiers, applicable stage types, prohibition conditions, minimum budget duration, priority rules, and dependencies. Data sources include: near-field wireless signal cues from BLE beacons and NFC card readers; semantic confirmation results from camera-recognized text fragments and output matching scores; behavioral status data from the mobile phone's inertial sensor and output walking or stationary status; and business events from the physical examination business system and output events such as check-in, call number, and completion, as well as process stage identifiers. The example data contains a total of 878 arrival events, counted by anchor type as 165, 158, 152, 145, 138, and 120 respectively.

[0054] Evidence recording is generated by organizing multi-source observations on a per-arrival-event basis. Taking the arrival event at the blood collection window as an example, four types of key evidence are accessed within 6 seconds: BLE received signal strength is -63dBm, NFC card reading successfully returns a tag identifier; semantic recognition identifies "blood collection window 3" and outputs a matching score of 0.92; the business event provides a "test-blood collection check-in" process stage identifier; and the behavioral state is determined to be walking. Time alignment uses a sliding time window method, mapping near-field wireless signal clues, semantic confirmation results, behavioral state data, and business events to the same time axis. The time window is 8 seconds, allowing for 1 second of jitter before and after to cover network and sensor sampling delays. Field mapping rules convert fields from different sources into structured evidence entries. Each structured evidence entry includes at least a timestamp, evidence type identifier, and anchor point candidate identifier, and may include strength or score. For example, this arrival yields: 09:21:35.120 Near-field wireless evidence anchor candidate identifier A03 strength -63dBm; 09:21:36.005 Semantic evidence anchor candidate identifier A03 score 0.92; 09:21:40.000 Business event evidence process stage identifier verification - blood collection check-in; 09:21:34.910 Behavioral status evidence walking status. Structured evidence entries with the same anchor candidate identifier within the time window are aggregated to obtain evidence records. The aggregated fields include the candidate set, evidence type set, aligned time range, and observation set for each evidence type, providing a unified input for subsequent graph structure modeling and consistency determination.

[0055] The evidence interaction graph organizes structured evidence items into a graph structure to explicitly express consistency and conflict. Nodes are structured evidence items, and edges are divided into consistency relationships and conflict relationships. Consistency relationships are used to express the same direction of semantic confirmation results and near-field wireless signal clues on anchor candidate identifiers. For example, if semantic evidence gives A03 and near-field wireless evidence also points to A03, a consistency relationship is established between the corresponding nodes. Conflict relationships are used to express inconsistencies between stage types and process stage identifiers: stage types come from anchor templates, while process stage identifiers come from business events; for example, if the anchor template labels the stage type as "image examination waiting," but the business event is "inspection-blood collection check-in," a conflict relationship is established and a conflict label is generated. To ensure engineering feasibility, conflict labels are implemented using an enumerated set method, which at least includes categories such as "stage inconsistency," "arrival outside the time window," and "candidate not unique."

[0056] Consistency determination outputs anchor point identifiers by comprehensively considering consistency and conflict relationships within a preset time window, and updates the anchor point confidence level in a tiered manner. The anchor point confidence level must include at least the arrival level and the push level. A feasible confidence calculation employs weighted aggregation and adds conflict penalties: in, For anchor point confidence score, For near-field wireless consistency score, The semantic matching score, As a score for consistency of business events, For the number of conflicting tags, , , As weight, This represents the conflict penalty coefficient. Example: [Example value would be inserted here] , , , In the blood collection window arrival event, the near-field wireless consistency score is 0.78, the semantic matching score is 0.92, the service event consistency score is 1.00, and the number of conflicting tags is 0. When the arrival level threshold is set to 0.75 and the push level threshold is set to 0.85, the event is determined to have arrived and is ready to be pushed.

[0057] Anchor point determination effectiveness is assessed using precision, recall, and F1 score statistics. Definitions: This represents the number of true positives. This refers to the number of false positives. The number of false negatives; For accuracy; Recall rate; For harmonic average indicators.

[0058] In the 878 arrival events, "single-source triggering" was used as a comparison: The information desk posters were received 165 times, compared to (TP=128, FP=30, FN=37) in the control group and (TP=145, FP=11, FN=20) in this invention. The ECG readings reached 158 times, compared to (TP=122, FP=31, FN=36) in the control group and (TP=141, FP=12, FN=17) in this invention. The blood collection window reached 152 times. The control group was (TP=124, FP=25, FN=28), while the present invention was (TP=139, FP=9, FN=13). The ultrasound waiting screen reached 145 times, with the control group having (TP=112, FP=31, FN=33) and the present invention having (TP=130, FP=13, FN=15); The CT registration site reached 138 times, with the control group being (TP=110, FP=23, FN=28) and the present invention being (TP=124, FP=9, FN=14); The physical examination item recognition reached 120 times, the control was (TP=100,FP=19,FN=20), and the present invention was (TP=106,FP=6,FN=14).

[0059] After aggregation, the precision of the control group was 0.816, the recall was 0.791, and the F1 score was 0.804; the precision of this invention was 0.928, the recall was 0.894, and the F1 score was 0.912. The total number of false triggers decreased from 159 to 60, a decrease of approximately 62.3%. F1 scores for each anchor type are compared below. Figure 2 .

[0060] Evidence fingerprints are used to solidify the combination of evidence sources and conflict states, facilitating closed-loop statistics and traceability. In implementation, the evidence type set and conflict label set are encoded as controlled-length fields: the evidence type set uses a 4-bit bitmap to represent the occurrence of near-field wireless, semantic, behavioral, and business events; the conflict label set uses a predefined label bitmap to represent conflict categories. The two are concatenated to generate a short string fingerprint and written to the push record. Taking a blood collection window event as an example, when all four types of evidence are complete and there are no conflicts, the evidence type bitmap is 1111 and the conflict bitmap is 0000, which can be encoded as "F0"; when a stage inconsistency conflict occurs, the conflict bitmap is set accordingly to form "F1", etc.

[0061] The window budget is jointly determined by the anchor confidence level, behavioral state data, and business events. Behavioral state categories are limited to walking or stationary states; business events provide process stage identifiers; these, combined with the stage type of the anchor template, yield the window type, such as "Stationary Waiting for Inspection" or "Walking Through the Passage." The window type maps to the budget duration and budget confidence identifier, which reflects the combined result of the anchor confidence level and behavioral stability. Example statistics show the budget duration distribution: in the walking state, the control mean is 6.20 seconds, median is 5.93 seconds, and 90th percentile is 9.14 seconds; the present invention's mean is 5.56 seconds, median is 5.31 seconds, and 90th percentile is 8.12 seconds. In the stationary state, the control mean is 24.66 seconds, median is 24.69 seconds, and 90th percentile is 34.18 seconds; the present invention's mean is 28.07 seconds, median is 28.29 seconds, and 90th percentile is 37.06 seconds. The walking budget has been tightened to reduce forced loading during the walking process, while the static budget has been increased to allow for a stable duration of augmented reality overlay in the waiting area, and the distribution is as follows: Figure 3 .

[0062] The generation of educational content unit sequences consists of two steps: contract matching and screening, and sequence assembly. Contract matching and screening iterates through educational content units within the contract, checking each unit to ensure its anchor point falls within the applicable anchor point set, its process stage matches the applicable stage type, its budget duration is not less than the minimum budget duration, and its prohibition conditions are met. Units meeting these conditions are added to the candidate set. Sequence assembly executes priority rules and satisfies dependencies on the candidate set: priority rules use integer levels and time order within the same level; dependencies use directed edges to indicate that preceding units must be presented first; during assembly, dependencies are topologically sorted, and tails are pruned within the budget duration limit. Taking the ECG doorplate scenario as an example, the process stage identifier is "ECG check waiting," the stage type is "ECG check," the budget duration is approximately 28 seconds, and the budget credibility is high. The final assembled sequence is a three-segment sequence: a short note card, an augmented reality overlay of lead pasting, and a post-check notification. If the business event notification indicates that the ECG check is complete, the prohibition condition is met, and the corresponding unit in the dependency chain does not enter the candidate set.

[0063] The presentation mode is determined based on whether the sequence of educational content units includes augmented reality overlay and whether the anchor confidence level reaches a threshold. Augmented reality overlay uses a presentation token controlled by a near-field wireless broadcasting device to control concurrent presentations within the same micro-area: the micro-area identifier is determined by the anchor template and the anchor identifier; the token request message carries the micro-area identifier, anchor identifier, and presentation duration parameters; upon receiving an authorization message, augmented reality overlay is executed and the token identifier is written to the push record; upon receiving a rejection message, it degenerates to a two-dimensional graphical interface presentation and the reason for the degeneration is written to the push record. In a 7-day example, the proportion of augmented reality overlay presentations in the control group increased from 0.52 to 0.59, while the proportion of this invention increased from 0.55 to 0.74; the proportion of two-dimensional graphical interface presentations decreased accordingly. Token rejections mainly arise from concurrent arrivals within micro-areas. This invention calculates the window budget before initiating a token request, reducing invalid requests due to insufficient budget, and the daily proportion is as follows: Figure 4 .

[0064] The closed-loop phase will push back records, evidence fingerprints, and business events for updating evidence reliability parameters and educational content contracts. The closed-loop evidence package is linked using anchor identifiers and timestamps as association keys. The association keys use timestamps rounded to the nearest 1 second and with content mismatches within a 2-second range before and after. The closed-loop evidence package must include at least anchor identifiers, anchor confidence levels, window budgets, educational content unit sequences, evidence fingerprints, business events, presentation modes, degradation reasons, dwell time, and interaction counts. Evidence reliability parameter updates use the frequency of conflict tag sets as the core statistic and can employ exponential smoothing for stable updates. in, This is the reliability parameter for evidence on that day. This represents the number of times conflict labels appear in every 100 closed-loop evidence packages. Taking "phase inconsistency" conflict as an example, the number of conflicts was approximately 22.0 times / 100 on day 1, decreasing to 17.5 times / 100 each day; for this invention, it decreased from 21.5 times / 100 to 9.5 times / 100. The update of the missionary content contract is triggered by the matching rate between the missionary content unit sequence and the process phase identifier: the matching rate can be calculated as "the number of units in the sequence that meet the current process phase / the total number of units in the sequence" and aggregated at the daily granularity. The matching rate of the control sample increased from 0.705 to 0.752, and for this invention, it increased from 0.735 to 0.872; when the matching rate is below the threshold, the prohibition conditions or the shortest budget duration are updated and a contract version identifier is generated, which takes effect on the next calendar day after issuance. The trend is as follows: Figure 5 .

[0065] Below is an example snippet showing the mapping between confidence scores and ratings, with inputs directly from the evidence records and consistency judgment output fields: defconfidence_score(near_field_score,semantic_score,event_score,conflict_count, wn=0.35,wm=0.35,we=0.30,lam=0.08): returnwn*near_field_score+wm*semantic_score+we*event_score-lam*conflict_count defgrade_map(score,arrive_th=0.75,push_th=0.85): arrive_level="Arrived" if score>=arrive_thelse" Not arrived push_level="pushable" if score>=push_thelse" only reachable returnarrive_level,push_level The verifiable points of this embodiment are concentrated in four areas: evidence recording solidifies multi-source observations into unified entries and maintains time traceability; the evidence interaction graph makes consistency and conflict explicit and makes conflict labels statistically quantifiable; the window budget constrains "when to push and for how long" into calculable results and forms a closed loop with behavior state categories and process stage identifiers; the closed-loop evidence package feeds back the push behavior to update the evidence reliability parameters and the propaganda content contract, forming a continuously converging self-calibration mechanism. Overall statistical results and... Figures 2 to 5 It is consistent and can be directly used as a standard for reproduction experiments and on-site joint commissioning and acceptance.

[0066] In one possible implementation, a system is provided for executing the aforementioned anchor-triggered propaganda push method. This system includes: an evidence collection and recording module, an evidence interaction graph construction module, a consistency determination module, an evidence fingerprint generation module, a window budget determination module, a propaganda content unit sequence generation module, a presentation control and push recording module, a closed-loop evidence package generation module, and a parameter and contract update module. The evidence collection and recording module is used to access anchor templates, propaganda content contracts, near-field wireless signal clues, semantic confirmation results, behavioral state data, and business events, and writes them into structured evidence entries according to a unified field specification; the evidence interaction graph construction module establishes relationship edges using structured evidence entries as nodes; the consistency determination module merges consistency and conflict relationships within a preset time window and outputs anchor identifiers and anchor confidence levels; the evidence fingerprint generation module encodes the evidence type set and conflict label set to form an evidence fingerprint; the window budget determination module combines the anchor confidence level, behavioral state data, and business event mapping... The system obtains the budget duration and budget credibility identifier; the propaganda content unit sequence generation module generates a sequence of propaganda content units based on contractual constraints and dependencies; the presentation control and push record module requests a presentation token from the near-field wireless broadcasting device and executes the overlay presentation when the augmented reality overlay presentation conditions are met; otherwise, it executes the two-dimensional graphical interface presentation and generates push records; the closed-loop evidence package generation module associates push records, evidence fingerprints, and business events with anchor point identifiers and timestamps to form a closed-loop evidence package; the parameter and contract update module updates the evidence reliability parameters based on the closed-loop evidence package and generates versioned versions of the propaganda content contract. The system can be deployed in a terminal-side and service-side collaborative form, with the terminal side responsible for near-field signal acquisition, semantic confirmation interaction, presentation, and recording, and the service side responsible for evidence interaction graph computation, contract matching, and closed-loop updates; each module can also be integrated into the same computing device to adapt to single-machine offline scenarios.

[0067] In one possible implementation, an electronic device is provided, comprising a processor and a memory; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory, causing the electronic device to execute the aforementioned anchor-triggered propaganda push method. The processor and memory can be connected via a bus. The memory is used to hold data objects such as anchor templates, propaganda content contracts, evidence records, closed-loop evidence packages, and contract version identifiers; the processor is used to perform computational steps such as constructing an evidence interaction graph, determining consistency, determining window budgets, contract matching and filtering, and writing token applications and degradation strategies.

[0068] Accordingly, a computer-readable storage medium is provided, which stores computer-executable instructions. When these instructions are executed by a processor, they are used to implement the aforementioned anchor-triggered educational push method. The computer-readable storage medium can be a non-transitory medium used to store evidence record field specifications, contract matching rules, dependency configurations, conflict label encoding rules, and closed-loop update strategies, enabling the same version of the strategy to be consistently reproduced across different devices.

[0069] Accordingly, a computer program product is also provided. The computer program product includes a computer program, which, when executed by a processor, implements the aforementioned anchor-triggered propaganda and education push method. The computer program product can be released in a versioned form, with version information associated with the version identifier of the propaganda and education content contract. This supports gray-scale distribution and rollback switching by department, physical examination item, or terminal model, thereby ensuring the maintainability and consistency of the closed-loop update process.

[0070] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects.

[0071] The above are merely embodiments of the present invention and are not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A method for pushing AR-based health checkup education based on anchor point triggering, characterized in that, Includes the following steps: Acquire anchor point templates, educational content contracts, near-field wireless signal clues, semantic confirmation results, behavioral status data, and business events to generate evidence records; Based on the evidence records, an evidence interaction graph is constructed and consistency is determined to obtain anchor point identifiers, anchor point confidence levels, and evidence fingerprints. Based on the anchor point confidence levels, the behavioral state data, and the business events, a window budget is determined, and a sequence of propaganda content units is generated based on the window budget and the propaganda content contract. When the sequence of propaganda content units includes augmented reality overlay and the confidence level of the anchor point reaches the threshold, a presentation token is requested from the near-field wireless broadcasting device and augmented reality overlay is executed; otherwise, a two-dimensional graphical interface is executed and a push record is generated. A closed-loop evidence package is generated based on the push record, the evidence fingerprint, and the business event, which is used to update the evidence reliability parameters and the propaganda content contract.

2. The method according to claim 1, characterized in that, Generating the evidence record includes: Time alignment is performed on the near-field wireless signal cues, the semantic confirmation results, the behavioral state data, and the service events; Based on field mapping rules, the near-field wireless signal clues, the semantic confirmation results, the behavioral state data, and the service events are converted into structured evidence entries; The structured evidence entry includes at least a timestamp, an evidence type identifier, and an anchor point candidate identifier; the structured evidence entry is aggregated to obtain the evidence record.

3. The method according to claim 2, characterized in that, Constructing the evidence interaction graph includes: The evidence interaction graph is established using the structured evidence items as nodes; A consistency relationship is established between the anchor candidate identifier corresponding to the semantic confirmation result and the anchor candidate identifier corresponding to the near-field wireless signal clue. Based on the anchor template, the stage type corresponding to the anchor candidate identifier is determined, and based on the business event, the process stage identifier is determined, and a conflict relationship is established. The conflict relationship is used to indicate that the stage type is inconsistent with the process stage identifier.

4. The method according to claim 3, characterized in that, The consistency determination includes: Within a preset time window, the anchor point identifier is determined based on the consistency relationship and the conflict relationship, and a set of conflict tags is generated; The anchor confidence level is updated hierarchically based on the set of conflicting tags, and the anchor confidence level includes at least arrival level and push level.

5. The method according to claim 4, characterized in that, Generating the evidence fingerprint includes: Based on the evidence records, a set of evidence types is determined, which includes at least near-field wireless signal clue evidence, semantic confirmation result evidence, behavioral state data evidence, and business event evidence. Obtain the set of conflicting tags; The evidence fingerprint is generated by encoding the evidence type set and the conflict label set. The evidence fingerprint is used to indicate the source of evidence and the conflict status of the anchor confidence level.

6. The method according to claim 1, characterized in that, Determining the window budget includes: The stage type corresponding to the anchor point identifier is determined based on the anchor point template; Based on the business event, the stage type is mapped to a process stage identifier; Based on the behavioral state data, a behavioral state category is determined, which is either walking state or stationary state. The window type is determined based on the process stage identifier and the behavior status category; The budget duration is determined based on the window type, and a budget credibility identifier is generated. The window budget includes the budget duration and the budget credibility identifier.

7. The method according to claim 6, characterized in that, Generating the sequence of missionary content units based on the window budget and the missionary content contract includes: Based on the aforementioned missionary content contract, the missionary content units are matched and screened. The missionary content contract includes at least the set of applicable anchor points, the type of applicable stage, the prohibition conditions, the minimum budget duration, the priority rules, and the dependency relationships. A candidate set is obtained when the anchor point identifier belongs to the set of applicable anchor point identifiers, the process stage identifier matches the applicable stage type, the budget duration is not less than the minimum budget duration, and the prohibition condition is not met. The missionary content unit sequence is generated based on the candidate set, according to the priority rules and satisfying the dependency relationships.

8. The method according to claim 1, characterized in that, Requesting the presentation token from the near-field wireless broadcasting device includes: The micro-area identifier is determined based on the anchor point template and the anchor point identifier; Send a token request message, which includes at least the micro-area identifier, the anchor point identifier, and the presentation duration parameter; Upon receiving a token authorization message, the augmented reality overlay is executed, and the token identifier is written into the push record; Upon receiving a token rejection message, the two-dimensional graphical interface is rendered and the degradation reason is written into the push record.

9. The method according to claim 5, characterized in that, Generating the closed-loop evidence package includes: The push record, the evidence fingerprint, and the business event are associated using the anchor point identifier and timestamp as association keys to obtain the closed-loop evidence package; The closed-loop evidence package includes at least the anchor point identifier, the anchor point confidence level, the window budget, the propaganda content unit sequence, the evidence fingerprint, and the business event; Updating the evidence reliability parameter includes: statistically analyzing the occurrence frequency of conflict label sets in the evidence fingerprint based on the closed-loop evidence package, and updating the evidence reliability parameter accordingly.

10. The method according to claim 9, characterized in that, The updated mission content contract includes: Based on the closed-loop evidence package, the matching rate between the sequence of propaganda content units and the process stage identifiers corresponding to the business events is calculated. In response to the matching rate being lower than a preset threshold, at least one of the prohibition conditions and the minimum budget duration of the missionary content contract is updated, and a contract version identifier is generated; the updated missionary content contract is issued based on the contract version identifier.

Images