A mobile advertisement targeting method and device based on micro scene recognition and a medium

Abstract

The application discloses a mobile advertisement targeting method and device based on micro-scene recognition and a medium, relates to the technical field of advertisement targeting, and comprises the following steps: establishing a micro-scene semantic node graph, a segment dictionary and a trigger bit list, and generating a node semantic constraint set by configuring behavior semantic constraints according to nodes; segmenting and arranging terminal events by using the node semantic constraint set and labeling parallel behavior semantics to obtain a time semantic segment chain and a double-labeled segment set; performing time-aware sequence alignment and parallel candidate disambiguation on the time semantic segment chain and the double-labeled segment set to form a scene stage code and a next node indication; and calling the scene stage code, the next node indication and the node semantic constraint set to perform advertisement bit arrangement and material pool constraint distribution to obtain a delivery execution unit. The application updates a trigger relationship state in a structured manner through a scene transition account book, so that the node semantic constraint set can be continuously corrected according to real scene transitions after delivery.

Description

Technical Field

[0001] This invention relates to the field of advertising targeting technology, and in particular to a mobile advertising targeting method, device and medium based on micro-scene recognition. Background Technology

[0002] As mobile advertising technology has evolved from geofence-based regional targeting to refined targeting based on terminal behavior, related technologies have formed a data-driven process based on location status, movement status, application usage status, and location advertising space management. Existing solutions typically achieve advertising reach across scenarios such as in-store visits, in-store decision-making, and post-payment conversions through business location segmentation, terminal event collection, behavioral semantic annotation, advertising space matching, and campaign execution recording. These solutions are applied in locations such as commercial complexes and transportation hubs.

[0003] Existing technologies for micro-scene-level ad targeting generally suffer from problems such as coarse-grained location units, insufficient correlation between behavioral semantics and segment temporal sequence, and difficulty in expressing parallel behaviors at the same node. This can easily lead to discrepancies between scene stage identification and the judgment of the next node, resulting in inconsistencies between ad placement arrangement and material category, and unstable delivery timing. At the same time, there is a lack of structured linkage updates between the scene transition results after delivery and the front-end identification rules, making it difficult to continuously correct the triggering relationship with the real scene path. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a mobile advertising targeting method based on micro-scene recognition to solve the problem of mismatch between ad placement arrangement and creative delivery caused by inconsistency between micro-scene-level behavioral semantics and temporal stage recognition.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a mobile advertising targeting method based on micro-scene recognition, comprising,

[0008] Establish a micro-scene semantic node graph, fragment dictionary, and trigger bit list, and configure behavioral semantic constraints according to nodes to generate a node semantic constraint set;

[0009] The terminal events are fragmented and arranged using a set of node semantic constraints, and the parallel behavioral semantics are labeled to obtain a chain of temporal semantic fragments and a set of double-labeled fragments.

[0010] Perform time-aware sequence alignment and parallel candidate disambiguation on the temporal semantic fragment chain and the dual-labeled fragment set to form scene stage codes and next node indicators;

[0011] The ad placement orchestration and material pool constraint distribution are executed by calling the scene stage code, the next node indicator, and the node semantic constraint set to obtain the ad execution order;

[0012] By using the node semantic constraint set, the terminal events corresponding to the ad delivery order are further segmented and arranged to generate a chain of subsequent segments. The ad delivery order and the chain of subsequent segments are then recorded for the transfer of execution scenarios and the trigger relationship within the node semantic constraint set, thus generating the ad delivery decision result.

[0013] As a preferred embodiment of the mobile advertising targeting method based on micro-scene recognition described in this invention, the step of establishing a micro-scene semantic node graph, a fragment dictionary, and a trigger position list includes:

[0014] The nodes are divided and numbered according to the location units within the business premises that can trigger advertising decisions, resulting in a node number table;

[0015] Configure the node number table with location boundaries, venue types, and ad slot types to obtain a micro-scene semantic node graph and a list of trigger positions;

[0016] Define segment for terminal events such as entry, stay, pass through, queue, leave, cross floor, and payment, and configure the segment start and end time fields and node number field to obtain the segment dictionary.

[0017] As a preferred embodiment of the mobile advertising targeting method based on micro-scene recognition described in this invention, the step of configuring behavioral semantic constraints by node to generate a node semantic constraint set includes:

[0018] Call the micro-scene semantic node graph and trigger bit list, configure allowed and excluded behavior semantic items for each node, and obtain the node semantic item set;

[0019] By establishing a correspondence between the behavioral semantic items of each node and the fragment type through a fragment dictionary, a node fragment mapping table is obtained;

[0020] The node fragment mapping table is merged with the ad slot type of each node to generate a node semantic constraint set.

[0021] As a preferred embodiment of the mobile advertising targeting method based on micro-scene recognition described in this invention, the method involves using a node semantic constraint set to fragment and arrange terminal events and label parallel behavioral semantics to obtain a temporal semantic fragment chain and a dual-labeled fragment set, including:

[0022] Terminal events are read in short time windows, and the terminal events include at least the location status, motion status, screen-on status, foreground application category and wireless access identifier, to obtain a window event set;

[0023] The node event set is classified into node landing points and behavioral semantics using the node semantic constraint set to obtain node event segments;

[0024] The fragment dictionary is invoked to merge similar events of the same node into the node event segments, and the start and end times are filled in to obtain the time semantic fragment;

[0025] Perform parallel annotation on the semantics of behaviors that appear in parallel within the same node and the same time window to obtain a set of double-labeled fragments;

[0026] By concatenating the temporal semantic fragments in the order of occurrence and preserving the set of double-tagged fragments, a temporal semantic fragment chain is obtained.

[0027] As a preferred embodiment of the mobile advertising targeting method based on micro-scene recognition described in this invention, the step of performing time-aware sequence alignment and parallel candidate disambiguation on the temporal semantic fragment chain and the dual-labeled fragment set to form a scene stage code and a next node indicator includes:

[0028] Call the preset micro-scene template sequence and configure the node order, fragment type, fragment duration relationship and fragment interval relationship to obtain the time template set;

[0029] Expand the temporal semantic fragment chain into the main sequence, and insert the double-labeled fragment set into the fragment position with the same node number in the main sequence according to the node number of the double-labeled fragment, to obtain the parallel candidate sequence group;

[0030] The main sequence and the parallel candidate sequence group are aligned segment by segment using a time template set to obtain a candidate alignment result set. The segment alignment simultaneously verifies the node order, segment type, occurrence time, duration and adjacent interval.

[0031] The candidate alignment result set is compared with the coverage consistency and node continuity consistency of the execution phase to obtain the disambiguation candidate path;

[0032] Extract the current template stage and successor node from the disambiguation candidate path to form the scene stage code and the next node indicator.

[0033] As a preferred embodiment of the mobile advertising targeting method based on micro-scene recognition described in this invention, the step of calling the scene stage code, the next node indicator, and the node semantic constraint set to execute the ad placement orchestration and material pool constraint distribution, thereby obtaining the delivery execution order, includes:

[0034] By invoking the scene stage code and the next node indicator, the ad slot types corresponding to the current node and the next node are located in the node semantic constraint set to obtain the candidate ad slot set;

[0035] The candidate ad placement set is matched with a preset material pool category, and the node behavior semantic constraint verification is performed on the matching results to obtain the candidate material set;

[0036] Perform scenario-stage consistency screening on the candidate material set and determine the delivery channel to generate a delivery execution order.

[0037] As a preferred embodiment of the mobile advertising targeting method based on micro-scene recognition described in this invention, the step of performing sequential segmentation and arrangement of terminal events corresponding to the delivery execution unit through a set of node semantic constraints to generate a sequential segment chain includes:

[0038] Extract the terminal identifier, node number, and deployment time from the deployment execution order to obtain the continuation monitoring window;

[0039] The terminal events within the continuation segment monitoring window are filtered out from the node semantic constraint set, and the node landing point and behavior semantic classification are performed to obtain the continuation segment node event segment.

[0040] The fragment dictionary is invoked to perform fragmentation and arrangement of the continuation segment node event segments while preserving the node dwelling relationships, resulting in continuation segment temporal semantic fragments;

[0041] The continuation segment chain is generated by arranging the segments according to the start and end times of the semantic segments.

[0042] As a preferred embodiment of the mobile advertising targeting method based on micro-scene recognition described in this invention, the step of updating the scene transfer accounting and node semantic constraint set trigger relationship of the delivery execution order and subsequent segment chain to generate advertising delivery decision results includes:

[0043] By comparing the delivery execution order and the subsequent segment chain, we can identify node entry, node dwell, post-payment segment, exit segment, and the re-triggering of similar materials to obtain a scene transfer record set;

[0044] The scene transfer record set is merged according to the node number and scene stage code to form a scene transfer ledger;

[0045] The scenario transfer ledger is invoked to adjust the trigger relationship status within the semantic constraint set of the node, and the corresponding associated items in the trigger bit list and fragment dictionary are adjusted simultaneously to generate the advertising placement decision results.

[0046] In a second aspect, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the mobile advertising targeting method based on micro-scene recognition as described in the first aspect of the present invention.

[0047] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the mobile advertising targeting method based on micro-scene recognition as described in the first aspect of the present invention.

[0048] The beneficial effects of this invention are as follows: It employs a continuous technical path of node semantic constraint set—temporal semantic fragment chain—time-aware sequence alignment—scene transfer ledger update, establishing a unified constraint relationship between location boundaries, ad slot types, behavioral semantic items, and fragment types. Furthermore, it retains parallel behavioral semantics within the same node through a dual-labeled fragment set, thereby improving the micro-scene recognition and expression capabilities. Additionally, it performs segment-by-segment alignment and parallel candidate disambiguation of node order, fragment continuity, and fragment interval relationships through a temporal template set, improving the stability of scene stage codes and next node indications. Finally, it uses a scene transfer ledger to structurally update the trigger relationship state, enabling the node semantic constraint set to be continuously corrected as real-world scene transfers occur after ad placement, thereby improving the accuracy of ad slot arrangement, the consistency of material constraints, and the reusability of ad placement execution. Attached Figure Description

[0049] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a flowchart of a mobile advertising targeting method based on micro-scene recognition.

[0051] Figure 2 A schematic diagram for micro-scene modeling and semantic constraint generation.

[0052] Figure 3 This is a diagram illustrating the fragmented arrangement and semantic annotation of terminal events.

[0053] Figure 4 This is a diagram illustrating scene recognition and ad placement decisions. Detailed Implementation

[0054] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0055] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0056] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0057] Reference Figures 1-4 This is one embodiment of the present invention, which provides a mobile advertising targeting method based on micro-scene recognition, including the following steps:

[0058] S1. Establish a micro-scene semantic node graph, fragment dictionary, and trigger bit list, and configure behavioral semantic constraints according to nodes to generate a node semantic constraint set.

[0059] S1.1. List the location units that can trigger advertising decisions within the business premises. The location units include shopping mall entrances, store fronts, store passages, cashier areas, escalator entrances, waiting areas, parking payment points, and the outside of subway turnstiles. Divide the location units into nodes according to their spatial distribution within the business premises, and assign a unique node number to each node. The node number and the location unit name form a one-to-one correspondence, resulting in a node number table.

[0060] The node number is generated by combining the site number and the location unit number, and the expression is:

[0061] ;

[0062] in, Indicates the node number. Indicates the location number. Indicates the position element number. The length of the number of bits indicating the position unit number.

[0063] S1.2. Configure the location boundaries for each node in the node number table. The location boundaries are expressed using a sequence of polygon boundary points, and this sequence is written into the location boundary field of the node number table, resulting in a node number table with location boundary fields. The location boundaries are validated using polygon area calculation, and the area expression is:

[0064] ;

[0065] in, Represents the area of ​​a polygon. Indicates the number of boundary points of the polygon. Indicates the first The coordinates of the boundary points Pick To close the boundary.

[0066] In the node number table with location boundary field, configure the venue type and ad slot type for each node. The venue type is used to describe the business attributes of the location unit, and the ad slot type is used to describe the ad carrying location. The ad slot type is expressed by the display location category within the business venue. Organize the node number table with location boundary field, venue type field and ad slot type field to form a micro-scene semantic node graph. At the same time, organize the correspondence between node number and ad slot type to form a trigger position list.

[0067] S1.3. Based on the identifiable event types of terminal events, define entry segments, dwell segments, passage segments, queue segments, exit segments, cross-floor segments, and post-payment segments. For entry segments, dwell segments, passage segments, queue segments, exit segments, cross-floor segments, and post-payment segments, uniformly configure segment start and end time fields and node number fields. The segment start and end time fields are used to record the segment start time and segment end time, and the node number field is used to reference the node number in the node number table. The entry segments, dwell segments, passage segments, queue segments, exit segments, cross-floor segments, post-payment segments, and field definitions are organized into a segment dictionary.

[0068] Based on the micro-scene semantic node graph and the trigger bit list, each node is configured with allowed and excluded behavior semantic items. Allowed behavior semantic items are used to limit the allowed behavior semantic categories within the node, while excluded behavior semantic items are used to limit the behavior semantic categories within the node that do not participate in advertising decisions. Allowed and excluded behavior semantic items are associated with the node number to obtain a set of node semantic items. The set of node semantic items is used in subsequent steps to generate a node fragment mapping table.

[0069] S1.4. Based on the fragment dictionary, establish correspondences between the allowed behavior semantic items and excluded behavior semantic items in the node semantic item set and the entry fragment, stay fragment, pass through fragment, queue fragment, leave fragment, cross-level fragment, and post-payment fragment, respectively. The node number, behavior semantic item, and fragment type form a triple record. The set of triple records is organized to form a node fragment mapping table.

[0070] The node fragment mapping table and the trigger position list are merged, with the node number as the merge key. The merged result retains the correspondence between the behavior semantic items and fragment types in the node fragment mapping table, and also retains the ad position types in the trigger position list. The merged result is organized into a node semantic constraint set, which includes the node number, location boundary, venue type, ad position type, fragment type corresponding to the allowed behavior semantic items, and fragment type corresponding to the excluded behavior semantic items.

[0071] S2. The terminal events are fragmented and arranged using the node semantic constraint set, and the parallel behavioral semantics are labeled to obtain the temporal semantic fragment chain and the double-labeled fragment set.

[0072] S2.1. Read terminal events according to a fixed short time window. The fixed short time window length is set to thirty seconds. The fixed short time window adopts a continuous scrolling mode with a continuous scrolling step size of five seconds. The starting point of the fixed short time window is aligned with the five-second scale according to the terminal event timestamp and then proceeds sequentially.

[0073] The 30-second value is used to cover the complete occurrence process of the entry segment, the passage segment, and the exit segment, and simultaneously accommodates the continuous recording of the dwell segment, the queuing segment, and the post-payment segment within the window. The 5-second value is used to ensure that there is an overlapping interval between adjacent fixed short-time windows to avoid segment breaks when terminal events are distributed across windows. The terminal events within the fixed short-time window record at least the location status, movement status, screen-on status, foreground application category, and wireless access identifier. The location status is used to indicate the location where the terminal event occurs, the movement status is used to indicate the movement status corresponding to the terminal event, the screen-on status is used to indicate the screen status corresponding to the terminal event, the foreground application category is used to indicate the application category corresponding to the terminal event, and the wireless access identifier is used to indicate the access identifier corresponding to the terminal event. The terminal events within the fixed short-time window are arranged in chronological order to form a window event set.

[0074] Based on the location boundary and node number in the node semantic constraint set, node landing processing is performed on the window event set. The node landing processing adopts polygon inclusion judgment between the positioning state and the location boundary, and performs location unit verification in combination with the wireless access identifier. The node number is labeled for terminal events whose positioning state falls into the location boundary and whose wireless access identifier satisfies the location unit correspondence relationship, thus obtaining the node landing event set.

[0075] S2.2. Based on the allowed and excluded behavior semantic items in the node semantic constraint set, perform behavior semantic classification on the node landing event set. Behavior semantic classification uses the correspondence between motion state, screen-on state, foreground application category and allowed behavior semantic items for matching, and uses excluded behavior semantic items to filter terminal events that do not participate in advertising decisions, retaining terminal events that pass the matching and filtering; terminal events that pass the matching and filtering are aggregated according to node number and occurrence time to form node event segments.

[0076] The fragment dictionary is invoked to merge events of the same type and node type on the node event segment. Merging events of the same type and node type is based on the same node number and the same fragment type as merging conditions, and the continuous occurrence time of terminal events is used as the connection condition. The merged record is supplemented with the fragment start and end time field and the node number field to obtain the time semantic fragment. The fragment type in the time semantic fragment is selected from the following: entry fragment, stay fragment, passage fragment, queue fragment, departure fragment, cross-floor fragment, and post-payment fragment.

[0077] S2.3. Within the same node and the same fixed short time window, check the behavioral semantic items corresponding to the time semantic segments. If the behavioral semantic items simultaneously satisfy two allowed behavioral semantic items and the segment types corresponding to the two allowed behavioral semantic items are both retained, perform parallel behavioral semantic annotation. The parallel behavioral semantic annotation records the node number, segment start and end time, two behavioral semantic items and two segment types to obtain a double-labeled segment set.

[0078] The time semantic segments are connected according to their occurrence order. The occurrence order is arranged according to the order of the start time of the segments. Time semantic segments with the same start time are arranged according to the order of the end time of the segments. The arranged time semantic segments are organized into a time semantic segment chain, and the association between the node number corresponding to the double-tagged segment set and the start and end times of the segments is preserved.

[0079] Resident segments and queued segments are filtered from the temporal semantic segment chain and organized into a node dwell segment list according to node number, segment start time, and segment end time. The node dwell segment list is used to represent the continuous time range of dwelling behavior within a node. The temporal semantic segment chain is organized into the current segment chain, and the double-labeled segment set is organized into a double-labeled segment list.

[0080] S3. Perform time-aware sequence alignment and parallel candidate disambiguation on the temporal semantic fragment chain and the dual-labeled fragment set to form scene stage codes and next node indicators.

[0081] S3.1. Organize the micro-scene template sequence according to the advertising decision path of the business venue. The micro-scene template sequence adopts a combination recording method of "node order, segment type, segment duration relationship, and segment interval relationship", and configures a template number and stage number for each micro-scene template sequence. The node order is used to indicate the order of node numbers in the micro-scene template sequence. The segment type is used to indicate the entry segment, dwell segment, passage segment, queue segment, departure segment, cross-floor segment, and post-payment segment corresponding to each stage in the micro-scene template sequence. The segment duration relationship is used to indicate the duration range formed by the start and end times of the segment in the same stage. The segment interval relationship is used to indicate the time interval range between adjacent stage segments. The organized micro-scene template sequence forms a time template set.

[0082] The temporal semantic fragment chain is expanded into a main sequence according to the fragment start time. Each record in the main sequence retains the node number, fragment type, fragment start time, and fragment end time. The double-labeled fragment set is located at the fragment positions in the main sequence with the same node number and overlapping time according to the node number and fragment start and end time of the double-labeled fragments. Branch records corresponding to the two fragment types are retained at the same fragment positions. When the double-labeled fragment set contains multiple double-labeled fragments, the branch records are expanded one by one according to the order of the double-labeled fragments in the temporal semantic fragment chain. The expanded record set forms a parallel candidate sequence group.

[0083] S3.2. Use time template sets to perform segment-by-segment alignment on the main sequence and parallel candidate sequence groups respectively. Segment-by-segment alignment compares node order, segment type, occurrence time, duration, and adjacent intervals in the order of stage number in the time template set. The occurrence time is compared by the segment start time, the duration is compared by the segment end time minus the segment start time, and the adjacent interval is compared by the time interval obtained by the subtraction of the start time of the next segment from the end time of the previous segment. Records with consistent node order, consistent segment type, duration falling within the segment duration relationship range, and adjacent interval falling within the segment interval relationship range are retained as valid matching records. Valid matching records are organized into a candidate alignment result set according to template number and stage number.

[0084] The candidate alignment result set is compared for stage coverage consistency and node continuity consistency. Stage coverage consistency comparison is used to check whether the stage numbers of the candidate alignment result set are continuous under the same template number. Candidate paths with continuous stage numbers and a large number of consecutive stage numbers are retained. Node continuity consistency comparison is used to check the continuity relationship between the node number of the last segment of the candidate path and the corresponding subsequent stage node number in the time template set. Candidate paths whose last segment node number and subsequent stage node number conform to the node order of the time template set are retained. After the stage coverage consistency comparison and node continuity consistency comparison are completed, the retained results are organized to form disambiguation candidate paths.

[0085] Read the template number and the stage number corresponding to the current valid matching record from the disambiguation candidate path. The template number and the stage number are combined to form the scene stage code. Read the node numbers of the subsequent nodes of the current stage from the time template set corresponding to the disambiguation candidate path. The subsequent node numbers form the next node indicator.

[0086] S4. Call the scene stage code, next node indicator and node semantic constraint set to execute the ad placement arrangement and material pool constraint distribution to obtain the delivery execution order.

[0087] S4.1. Determine the current node number based on the scene stage code, and determine the next node number based on the next node indication. The scene stage code uses a combination of template number and stage sequence number for encoding. The template number and stage sequence number correspond to the node number of the current stage. Based on the current node number and the next node number, retrieve the ad slot type in the node semantic constraint set. The ad slot types in the node semantic constraint set come from the trigger position list formed in S1. The retrieval results are organized into a candidate ad slot set according to the current node number and the next node number.

[0088] The candidate ad placement set is matched with the material pool category. The material pool category is classified and recorded as in-store guidance, in-store decision-making, queue reduction, post-payment related, and out-of-store recall. The ad placement type in the candidate ad placement set is matched with the ad placement type applicable to the ad material in the material pool category. After the matching, the ad material that matches the candidate ad placement set is retained to obtain the first candidate material set.

[0089] S4.2. Perform node behavior semantic constraint verification on the first candidate material set based on the node semantic constraint set. The node behavior semantic constraint verification uses the allowed behavior semantic items and excluded behavior semantic items corresponding to the current node number for checking, and combines the fragment type of the scene stage code to perform consistency verification. Advertisements whose material pool category is consistent with the fragment type corresponding to the allowed behavior semantic item of the current node number and do not fall into the fragment type corresponding to the excluded behavior semantic item are retained. Advertisements whose material pool category is consistent with the fragment type corresponding to the excluded behavior semantic item are removed. The verification results are organized to form a candidate material set.

[0090] The candidate creative materials set undergoes scenario-stage consistency screening to determine the delivery channel. Scenario-stage consistency screening checks the correspondence between the scenario stage code and the applicable stage of the creative material pool category. Creative materials that match the stage corresponding to the scenario stage code are retained. The delivery channel is determined according to the ad placement type in the candidate ad placement set, and the ad placement type is consistent with the display location category in the business venue. The scenario-stage consistency screening results and the delivery channel are combined to form a delivery execution order. The delivery execution order includes at least the terminal identifier, node number, scenario stage code, creative material number, delivery channel, and delivery time.

[0091] S5. By using the node semantic constraint set, the terminal events corresponding to the ad delivery order are further segmented and arranged to generate a chain of subsequent segments. The ad delivery order and the chain of subsequent segments are then recorded for the transfer of execution scenarios and the trigger relationship within the node semantic constraint set, and the ad delivery decision results are generated.

[0092] S5.1. Extract the terminal identifier, node number, and delivery time from the delivery execution order. The terminal identifier is used to limit the source of the terminal event, the node number is used to limit the node where the delivery occurred, and the delivery time is used to limit the starting point of the follow-up monitoring. Generate a follow-up monitoring window based on the delivery time. The follow-up monitoring window adopts a fixed short time window continuous scrolling method. The fixed short time window length is the same as the thirty seconds in S2.1, and the continuous scrolling step size is the same as the five seconds in S2.1. The starting point of the follow-up monitoring window is the delivery time, and the ending point of the follow-up monitoring window is 180 seconds after the delivery time. The follow-up monitoring window is used to cover the subsequent changes of node entry, node stay, post-payment segment, and departure segment.

[0093] The monitoring window for the subsequent segment is calculated based on the delivery time, as expressed by the following expression:

[0094] ;

[0095] in, Indicates the start time of the continued monitoring window. Indicates the end time of the continued monitoring window. The value indicates the time of deployment, and 180 indicates the duration of the continuous monitoring window, in seconds.

[0096] Based on the terminal identifier and the continuation monitoring window, terminal events corresponding to the delivery execution order are screened from the terminal events. The screening results are organized in chronological order to form a continuation window event set. Node landing processing is performed on the continuation window event set according to the position boundary and node number in the node semantic constraint set. The node landing processing follows the polygon inclusion judgment and wireless access identifier location unit verification rules in S2.1. The terminal events after retaining the node number form a continuation node landing event set. Behavioral semantic classification is performed on the continuation node landing event set according to the allowed and excluded behavioral semantic items in the node semantic constraint set. Behavioral semantic classification follows the motion state, screen-on state, and foreground application category matching rules in S2.2. The classification results are aggregated by node number and occurrence time to form a continuation node event segment.

[0097] S5.2. Call the fragment dictionary to perform fragmented arrangement on the continuation node event segments. The fragmented arrangement follows the same rules for merging events of the same type and node in S2.2. The merging conditions are the same node number and the same fragment type, and the connection condition is the continuous occurrence time of the terminal event. The merged records are supplemented with the fragment start and end time fields and the node number field. The fragment type is selected from the entry fragment, dwell fragment, passage fragment, queue fragment, departure fragment, cross-floor fragment, and post-payment fragment. The fragmented arrangement result forms the continuation time semantic fragment. Records belonging to dwell fragment and queue fragment in the continuation time semantic fragment retain the node dwelling relationship according to the node number.

[0098] The duration of the continuation segment is calculated using the start and end times fields of the continuation time semantic segment. The expression is as follows:

[0099] ;

[0100] in, Indicates the duration of a semantic segment that continues for a period of time. Indicates the start time of the segment. Indicates the end time of the segment.

[0101] The temporal sequence of the semantic segments is arranged according to the start and end times of the segments. The segments are arranged in chronological order of their start times. For segments with the same start time, the segments are arranged in chronological order of their end times. The arranged segments are organized into a chain of segments, which retains the node number, segment type, segment start and end times, and node stopping relationship.

[0102] S5.3. Identify scene transfer events by comparing the delivery execution order and the subsequent segment chain. Scene transfer events include at least node entry, node dwell, post-payment segment, departure segment, and re-triggering of similar materials. Node entry is identified by matching the node number in the subsequent segment chain with the node number indicated by the next node in the delivery execution order. Node dwell is identified by combining the dwell segment and queue segment in the subsequent segment chain with the node dwell relationship. Post-payment segment is identified by the post-payment segment in the subsequent segment chain. Departure segment is identified by the departure segment in the subsequent segment chain. Re-triggering of similar materials is identified by forming a new delivery execution order in the subsequent monitoring window and the material number matching the material number in the current delivery execution order. The identification results are organized into a scene transfer record set according to terminal identifier, node number, scene stage code, material number, event type, and event occurrence time.

[0103] The scene transfer record set is merged according to the node number and scene stage code. During the merging, the node entry record, node dwell record, post-payment fragment record, departure fragment record, and re-trigger record of the same material under the same node number and scene stage code are counted separately. The merging results are compiled into a scene transfer ledger. The scene transfer ledger contains at least the node number, scene stage code, number of node entry, number of node dwell, number of post-payment fragments, number of departure fragments, and number of re-trigger of the same material.

[0104] The entry and exit ratios of nodes are calculated using the counting results in the scene transfer ledger. The expression is as follows:

[0105] ;

[0106] in, Indicates the percentage of nodes entering. Indicates the percentage of people leaving. Indicates the number of times a node has been entered. Indicates the number of times the segment is left. This indicates the total number of records corresponding to the same node number and the same scene stage code.

[0107] S5.4. Call the scene transfer ledger to adjust the trigger relationship status within the node semantic constraint set. The trigger relationship status is recorded using a combination of node number, scene stage code, ad slot type, and fragment type. A combination of node entry ratio greater than exit ratio and fragment count greater than zero after payment is marked as priority retention. A combination of node exit ratio greater than node entry ratio is marked as suspended. A combination of node entry ratio and exit ratio both exist and fragment count is zero after payment is marked as normal retention. The trigger relationship status adjustment results are used to synchronously adjust the corresponding associated items in the trigger slot list and fragment dictionary.

[0108] Based on the results of the trigger relationship status adjustment, the correspondence between the node number in the trigger position list and the ad position type is adjusted synchronously, and the available association items of the fragment type in the fragment dictionary are adjusted synchronously. The adjusted trigger position list and the adjusted fragment dictionary are merged and organized with the node number, location boundary, location type, fragment type corresponding to the allowed behavior semantic item, and fragment type corresponding to the excluded behavior semantic item in the node semantic constraint set to form the ad placement decision result.

[0109] This embodiment also provides a computer device applicable to the mobile advertising targeting method based on micro-scene recognition, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the mobile advertising targeting method based on micro-scene recognition as proposed in the above embodiment.

[0110] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0111] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the mobile advertising targeting method based on micro-scene recognition as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0112] In summary, this invention employs a continuous technical path of node semantic constraint set—temporal semantic fragment chain—time-aware sequence alignment—scene transfer ledger update. It establishes a unified constraint relationship between location boundaries, ad slot types, behavioral semantic items, and fragment types. Furthermore, it preserves parallel behavioral semantics within the same node through a dual-labeled fragment set, thereby improving the micro-scene recognition and expression capabilities. It further enhances the stability of scene stage codes and next node indications by performing segment-by-segment alignment and parallel candidate disambiguation on node order, fragment continuity relationships, and fragment interval relationships through a temporal template set. Finally, it structures and updates the trigger relationship state through the scene transfer ledger, enabling the node semantic constraint set to be continuously corrected as real-world scene transfers occur after ad placement, thus improving the accuracy of ad slot arrangement, the consistency of material constraints, and the reusability of ad placement execution.

[0113] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A mobile advertising targeting method based on micro-scene recognition, characterized in that: include, Establish a micro-scene semantic node graph, fragment dictionary, and trigger bit list, and configure behavioral semantic constraints according to nodes to generate a node semantic constraint set; The terminal events are fragmented and arranged using a set of node semantic constraints, and the parallel behavioral semantics are labeled to obtain a chain of temporal semantic fragments and a set of double-labeled fragments. Perform time-aware sequence alignment and parallel candidate disambiguation on the temporal semantic fragment chain and the dual-labeled fragment set to form scene stage codes and next node indicators; The ad placement orchestration and material pool constraint distribution are executed by calling the scene stage code, the next node indicator, and the node semantic constraint set to obtain the ad execution order; By using the node semantic constraint set, the terminal events corresponding to the ad delivery order are further segmented and arranged to generate a chain of subsequent segments. The ad delivery order and the chain of subsequent segments are then recorded for the transfer of execution scenarios and the trigger relationship within the node semantic constraint set, thus generating the ad delivery decision result.

2. The mobile advertising targeting method based on micro-scene recognition as described in claim 1, characterized in that: The establishment of the micro-scene semantic node graph, fragment dictionary, and trigger bit list includes: The nodes are divided and numbered according to the location units within the business premises that can trigger advertising decisions, resulting in a node number table; Configure the node number table with location boundaries, venue types, and ad slot types to obtain a micro-scene semantic node graph and a list of trigger positions; Define segment for terminal events such as entry, stay, pass through, queue, leave, cross floor, and payment, and configure the segment start and end time fields and node number field to obtain the segment dictionary.

3. The mobile advertising targeting method based on micro-scene recognition as described in claim 2, characterized in that: The configuration of behavioral semantic constraints by node generates a set of node semantic constraints, including: Call the micro-scene semantic node graph and trigger bit list, configure allowed and excluded behavior semantic items for each node, and obtain the node semantic item set; By establishing a correspondence between the behavioral semantic items of each node and the fragment type through a fragment dictionary, a node fragment mapping table is obtained; The node fragment mapping table is merged with the ad slot type of each node to generate a node semantic constraint set.

4. The mobile advertising targeting method based on micro-scene recognition as described in claim 3, characterized in that: The method involves using a set of node semantic constraints to fragment and arrange terminal events, and labeling parallel behavioral semantics to obtain a chain of temporal semantic fragments and a set of double-labeled fragments, including: Terminal events are read in short time windows, and the terminal events include at least the location status, motion status, screen-on status, foreground application category and wireless access identifier, to obtain a window event set; The node event set is classified into node landing points and behavioral semantics using the node semantic constraint set to obtain node event segments; The fragment dictionary is invoked to merge similar events of the same node into the node event segments, and the start and end times are filled in to obtain the time semantic fragment; Perform parallel annotation on the semantics of behaviors that appear in parallel within the same node and the same time window to obtain a set of double-labeled fragments; By concatenating the temporal semantic fragments in the order of occurrence and preserving the set of double-tagged fragments, a temporal semantic fragment chain is obtained.

5. The mobile advertising targeting method based on micro-scene recognition as described in claim 4, characterized in that: The process of performing time-aware sequence alignment and parallel candidate disambiguation on the temporal semantic fragment chain and the dual-labeled fragment set to form a scene stage code and a next node indicator includes: Call the preset micro-scene template sequence and configure the node order, fragment type, fragment duration relationship and fragment interval relationship to obtain the time template set; Expand the temporal semantic fragment chain into the main sequence, and insert the double-labeled fragment set into the fragment position with the same node number in the main sequence according to the node number of the double-labeled fragment, to obtain the parallel candidate sequence group; The main sequence and the parallel candidate sequence group are aligned segment by segment using a time template set to obtain a candidate alignment result set. The segment alignment simultaneously verifies the node order, segment type, occurrence time, duration and adjacent interval. The candidate alignment result set is compared with the coverage consistency and node continuity consistency of the execution phase to obtain the disambiguation candidate path; Extract the current template stage and successor node from the disambiguation candidate path to form the scene stage code and the next node indicator.

6. The mobile advertising targeting method based on micro-scene recognition as described in claim 5, characterized in that: The invocation scenario stage code, next node indicator, and node semantic constraint set are used to execute ad placement orchestration and material pool constraint distribution, resulting in a delivery execution order, including: By invoking the scene stage code and the next node indicator, the ad slot types corresponding to the current node and the next node are located in the node semantic constraint set to obtain the candidate ad slot set; The candidate ad placement set is matched with a preset material pool category, and the node behavior semantic constraint verification is performed on the matching results to obtain the candidate material set; Perform scenario-stage consistency screening on the candidate material set and determine the delivery channel to generate a delivery execution order.

7. The mobile advertising targeting method based on micro-scene recognition as described in claim 6, characterized in that: The step of performing segmented orchestration of terminal events corresponding to the delivery execution unit through a set of node semantic constraints to generate a chain of segmented events includes: Extract the terminal identifier, node number, and deployment time from the deployment execution order to obtain the continuation monitoring window; The terminal events within the continuation segment monitoring window are filtered out from the node semantic constraint set, and the node landing point and behavior semantic classification are performed to obtain the continuation segment node event segment. The fragment dictionary is invoked to perform fragmentation and arrangement of the continuation segment node event segments while preserving the node dwelling relationships, resulting in continuation segment temporal semantic fragments; The continuation segment chain is generated by arranging the segments according to the start and end times of the semantic segments.

8. The mobile advertising targeting method based on micro-scene recognition as described in claim 7, characterized in that: The process of transferring and recording the execution scenarios of the ad delivery execution order and the subsequent segment chain, and updating the trigger relationships within the node semantic constraint set to generate ad delivery decision results includes: By comparing the delivery execution order and the subsequent segment chain, we can identify node entry, node dwell, post-payment segment, exit segment, and the re-triggering of similar materials to obtain a scene transfer record set; The scene transfer record set is merged according to the node number and scene stage code to form a scene transfer ledger; The scenario transfer ledger is invoked to adjust the trigger relationship status within the semantic constraint set of nodes, and the corresponding associated items in the trigger bit list and fragment dictionary are adjusted simultaneously to generate the advertising placement decision results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the mobile advertising targeting method based on micro-scene recognition as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the mobile advertising targeting method based on micro-scene recognition as described in any one of claims 1 to 8.

Images