Event filtering method, computer device and storage medium

Abstract

The application discloses an event filtering method, a computer device and a storage medium. The method comprises the following steps: acquiring a prompt message of a target event reported by a front-end device, wherein the prompt message comprises a plurality of event image frames; detecting and analyzing the plurality of event image frames to determine scene information and event related information of the target event; determining a target filtering condition for the target event based on the scene information, and filtering the target event by using the target filtering condition and the event related information; and wherein the prompt message of the filtered target event is not pushed for prompting. The above scheme can filter events in an adaptive scene, thereby more accurately prompting events.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to an event filtering method, computer device, and storage medium. Background Technology

[0002] With the development of image processing technology, various image acquisition devices are able to perform preliminary detection on the acquired images and generate events to provide prompts.

[0003] In related technologies, it is common practice to provide alerts for all detected events. However, if there are many events, it can lead to a large volume of alerts or interference from repetitive events, resulting in lower accuracy of event alerts. Summary of the Invention

[0004] The main technical problem addressed by this application is to provide an event filtering method, computer device, and storage medium that can adaptively filter events based on the scene, thereby providing more accurate event prompts.

[0005] The first aspect of this application provides an event filtering method, which includes: obtaining a prompt message of a target event reported by a front-end device, the prompt message containing several event image frames; detecting and analyzing the several event image frames to determine scene information and event-related information of the target event; determining target filtering conditions for the target event based on the scene information, and filtering the target event using the target filtering conditions and event-related information; wherein, the prompt message of the filtered target event is not pushed.

[0006] A second aspect of this application provides a computer device including a memory and a processor coupled to each other, the memory storing program instructions and the processor executing the program instructions to implement the aforementioned event filtering method.

[0007] A third aspect of this application provides a computer-readable storage medium storing program instructions that, when executed by a processor, implement the aforementioned event filtering method.

[0008] The above solution acquires the prompt messages of target events reported by the front-end device. These prompt messages contain several event image frames. The solution detects and analyzes these event image frames to determine the scene information and event-related information of the target event. Based on the scene information, it determines the target filtering conditions for the target event. The solution can adaptively determine the target filtering conditions according to the current scene information. Then, it filters the target event using the target filtering conditions and event-related information. Prompt messages for filtered target events are not pushed. This adaptive filtering of target events and the scene itself allows for more accurate event prompts, avoiding indiscriminate and mechanical event prompts, effectively reducing interference from invalid and redundant event prompts, and improving the accuracy and effectiveness of the prompt information. Furthermore, by not pushing prompt messages for filtered target events, and only performing prompt operations on valid target events that were not filtered, the solution ensures that important events are detected and processed in a timely manner, guaranteeing the timeliness and reliability of event prompts, while also reducing the system's computational and prompt resource consumption, thus improving the overall operational efficiency of event processing.

[0009] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this application. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in this application, the accompanying drawings required in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Among them: Figure 1 This is a schematic diagram of the structure of an embodiment of the event notification system of this application; Figure 2 This is a flowchart illustrating an embodiment of the event filtering method of this application; Figure 3 This application Figure 2 A flowchart illustrating an embodiment of step S13; Figure 4 This is an example schematic diagram of an embodiment of the objective-related information of this application; Figure 5 This is a schematic diagram of the structure of an embodiment of the filtering device for events in this application; Figure 6 This is a schematic diagram of the structure of an embodiment of the computer device of this application; Figure 7 This is a schematic diagram of the structure of an embodiment of the computer-readable storage medium of this application. Detailed Implementation

[0011] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0012] The terms "first" and "second" in this application are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such processes, methods, products, or apparatus.

[0013] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0014] In this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship. Furthermore, "several" in this document means one or more. Moreover, the term "at least one" in this document means any combination of at least two of any one or more of a plurality of elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0015] This application provides the following embodiments, and each embodiment is described in detail below.

[0016] Please see Figure 1 , Figure 1This is a schematic diagram of an embodiment of the event notification system of this application. The event notification system 100 includes a front-end device 101 and a server 102. The front-end device 101 can capture images / videos of a target area and perform event detection on the images / videos to detect whether a target event has occurred, thereby reporting a notification message of the target event to the server 102. The notification message contains several event image frames. After receiving the notification message of the target event and obtaining several event image frames, the server 102 can execute the following event filtering method to determine whether to filter the target event. If filtered, the notification message of the target event will not be pushed. For details, please refer to the following description of the embodiment.

[0017] Please see Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of the event filtering method of this application. This embodiment is applied to the server side, and the method may include the following steps: S11: Obtain the prompt message of the target event reported by the front-end device. The prompt message contains several event image frames.

[0018] The server can obtain prompt information about the target event reported by the front-end device. The prompt information includes several event image frames related to the target event. These event image frames can be captured by the front-end device from the target area. The event image frames can include continuous or discontinuous image or video frames. For example, a set number of image frames can be selected for the target event, or a video segment with a set selection duration can be selected for the target event. The set number of frames and the set selection duration can be related to the target event and can be preset. This application does not impose any restrictions on this.

[0019] Optionally, the front-end device can acquire images / videos of the target area, perform event detection on the images / videos to determine whether a target event has occurred; in response to the occurrence of a target event, it can use several image frames or video clips related to the acquired target event as event image frames, generate a prompt message for the target event, and upload it to the server. Optionally, the target event includes events such as the appearance of an entity target, the disappearance of an entity target, target behavior events, and scene change events. The target events to be detected can be determined according to the specific application scenario, and this application does not limit the target events.

[0020] In some implementations, the prompt information for the target event also includes first detection information of the target event by the front-end device. This first detection information includes at least one of the following: the number of entity targets (entity count), the location of the entity targets (entity location), etc. For example, the front-end device performs target detection on the acquired image / video, such as using algorithms like YOLO (You Only Look Once) or SSD (Single Shot MultiBox Detector), to extract information such as the coordinate location information of the entity targets (entity location) and the number of entities. Optionally, the front-end device can also determine whether a target event has occurred based on the first detection information.

[0021] S12: Detect and analyze several event image frames to determine the scene information and event-related information of the target event.

[0022] Several event image frames can be detected and analyzed to obtain scene information and event-related information of the target event. Scene information may include scene category, environment-related information, etc. Environment-related information may include region information, lighting information, time information, location information, temperature information, etc. Event-related information includes event information (such as event category, event description information, event feature information, etc.), target-related information (such as target category, image features, contour features, attribute information, behavioral features, etc.), and scene-related information (such as scene description information, scene feature information, etc.), etc., which are not limited in this application. Optionally, large language models, convolutional neural networks, etc., can be used to detect and analyze several event image frames to determine the scene information and event-related information of the target event.

[0023] In some implementations, step S12 above involves detecting and analyzing several event image frames to determine the scene information and event-related information of the target event, including the following steps: S121: Using the first detection information of the target event, select several event image frames to obtain the selected event image frames.

[0024] First, several event image frames are selected using the first detection information of the target event. Event image frames that meet preset selection conditions can be selected to obtain the selected event image frames. Optionally, the total evaluation value of each event image frame can be determined using the first detection information of the target event, and then event image frames whose total evaluation value meets preset selection conditions can be selected to obtain the selected event image frames. The first detection information includes at least one of entity location and entity quantity. For example, the total number of entities in each event image frame can be determined using the entity quantity; entity integrity can be determined using entity location (e.g., if the entity location is at the edge of the event image frame, the entity integrity is low); entity quality can be determined using the region image of the entity location (e.g., if the region image of the entity indicates that the entity is unobstructed and has high clarity, the entity quality is high); the total evaluation value of each event image frame can be obtained based on at least one of the following: total number of entities, entity integrity, entity quality, etc. If the same event image frame contains multiple entity targets, the final total evaluation value of the event image frame can be obtained by combining the total evaluation values ​​of all entity targets.

[0025] Optionally, the preset selection criteria include ranking the total evaluation value as a top preset number, where the preset number can be one or more. For example, event image frames with the highest total evaluation value or ranked as a top preset number can be selected; alternatively, event image frames with the largest number of entities, the highest entity quality evaluation value, or the highest entity completeness can be selected. For instance, the first number (e.g., the top one or more) of event image frames ranked by the number of entities are selected from the target event. Then, based on the entity's coordinate position information (entity position), the entity completeness in the first number of event image frames is determined. The second number (e.g., the top one or more) of event image frames ranked by the highest entity completeness are selected to obtain the selected event image frames, thereby eliminating event image frames where the entity is located at an edge, only a part is included, or the entity is occluded. This application does not limit the method of selecting event image frames.

[0026] S122: Detect the selected event image frame to obtain the second detection information.

[0027] Each event image frame can be inspected (e.g., by object detection, feature extraction, or at least one other method) to obtain second detection information. This second detection information can include object-related information, such as object category, image features, contour features, attribute information, and behavioral features. For example, feature extraction can employ a Person Re-identification (ReID) algorithm. By using object detection algorithms and ReID algorithms to inspect each event image frame, second detection information for the target event can be obtained.

[0028] S123: Use a large language model to detect the first detection information, the second detection information, and the selected event image frame to obtain the third detection information, wherein the third detection information contains at least the scene category.

[0029] The first detection information, the second detection information, and the selected event image frames can be input into a large language model. The large language model is then used to perform large-scale model analysis on each event image frame to obtain third detection information. This third detection information includes at least a scene category. Optionally, the third detection information includes scene category, environment-related information, scene-related information, event information, and target-related information. For example, a multi-scene large language model can be used to perform large-scale model analysis on each event image frame to perform large-scale model scene analysis, event extraction, and other processing to obtain the third detection information of the target event. For example, a visual large model and a large language model can be used to perform large-scale model scene analysis on each event image frame to obtain scene information and event-related information; this application does not impose limitations on this approach.

[0030] Optionally, one can first obtain information such as environment-related information and scene-related information, then perform vectorization encoding (Embedding encoding) on ​​the environment-related information and scene-related information to obtain scene feature information, and then use the scene feature information to classify the scene and obtain the scene category.

[0031] Optionally, the scene category can include multi-level preset categories, which contain subcategories with multiple dimensions. Each dimension can correspond to an influencing factor of a scene. For example, multiple dimensions can include at least one such dimension, such as region, time period, and security level. This application does not limit the number of multi-level preset categories. For example, taking the region dimension as an example, it can be divided into indoor and outdoor subcategories. The indoor subcategories can be further divided into subcategories such as corridors, classrooms, and offices, and the outdoor subcategories can be further divided into subcategories such as roads and doorways. Taking the time period dimension as an example, it can be divided into subcategories such as morning, afternoon, and evening, and each time period subcategory can be further divided into multiple time period subcategories. The security level dimension can represent the security level to which it belongs. For example, multiple security level subcategories can each filter target events to different degrees. For example, the filtering degree corresponding to the first security level subcategory is lower than that of the second security level subcategory (e.g., for target filtering conditions less than a first quantity threshold, the value of the first quantity threshold of the first security level subcategory is smaller). This application does not limit the number of multi-level preset categories.

[0032] In some implementations, preset scene categories (such as abnormal scene categories, a certain scene category, or all scene categories) and preset event categories (such as abnormal event categories, a certain event category, or all event categories) can be categorized and labeled. This application does not impose any restrictions on this.

[0033] S124: By combining the first detection information, the second detection information, and the third detection information, scene information and event-related information are obtained.

[0034] The first detection information is the first-level feature information detected by the front-end device, the second detection information is the second-level feature information detected by the server on several event image frames, and the third detection information is the third-level feature information obtained by the large language model. The first detection information, the second detection information, and the third detection information can be combined to obtain scene information and event-related information. That is, scene information and event-related information contain three levels of feature information.

[0035] S13: Based on scene information, determine the target filtering conditions for the target event, and filter the target event using the target filtering conditions and event-related information; wherein, the notification message for the filtered target event is not pushed.

[0036] Based on scene information, target filtering conditions for target events can be determined. Thus, target events can be filtered using target filtering conditions and event-related information to determine whether the prompt information for target events needs to be filtered. Specifically, prompt messages for filtered target events are not pushed, while prompt messages for unfiltered target events are pushed. The prompt message for the target event (or a push message generated based on the prompt message for the target event, scene information, and / or event-related information) can be pushed to the target device for prompting. This application does not impose any restrictions on this.

[0037] Optionally, in response to not filtering the target event, a prompt message for the target event is displayed, and the prompt information for the target event is pushed to the target device; or, in response to filtering the target event, a prompt message for the target event is not displayed, and the prompt information for the target event is not pushed to the target device.

[0038] Optionally, event-related information can be used to determine whether a target event meets the target filtering conditions; if the target filtering conditions are met, the target event is filtered; or, if the target filtering conditions are not met, the target event is not filtered.

[0039] In some implementations, the scene category of the target event can be determined based on scene information, such as office scene category, entrance / exit scene category, outdoor scene category, etc. This application does not limit the scene category; the target filtering conditions of the target event can be determined based on the preset filtering conditions corresponding to the scene category, thereby enabling adaptive configuration of the target filtering conditions.

[0040] In some implementations, different target filtering conditions correspond to different scene categories. For example, different parameter values ​​and filtering conditions can be configured. The thresholds or parameters in the following conditions, such as the preset rate of change threshold and the preset same target condition, for example, the first quantity threshold, the first similarity threshold, the second similarity threshold, and the third similarity threshold, can be various thresholds or parameters mentioned below. This application does not limit the parameter values ​​of the target filtering conditions. For example, regarding the parameter values ​​of the target filtering conditions, taking the first quantity threshold corresponding to the number of identical targets included in the preset same target condition as an example, for instance, in a multi-target scene (such as an office scene), the first quantity threshold is relatively large, meaning that some targets are not the same, so filtering can be omitted; for instance, in an entrance scene, the first quantity threshold is relatively small, meaning there are different targets or any different target is different, so filtering can be omitted. This application does not limit the parameter values ​​of the target filtering conditions.

[0041] In some implementations, when multiple levels of preset categories are included, the subcategory to which the scene category belongs can be determined based on scene information from the multiple levels of preset categories. For example, subcategories can be determined separately from the categories of each dimension, and then the subcategories of each dimension can be combined to obtain the subcategory to which the scene category belongs. The smallest subcategory to which the scene category belongs can be determined. For example, multiple dimensions may include a region dimension and a time dimension. Based on scene information, subcategories for the time dimension (e.g., a morning time subcategory) and subcategories for the region dimension (e.g., an entrance / exit subcategory) can be determined separately, and then the subcategories of each dimension can be combined to determine the subcategory to which the scene category belongs (e.g., an entrance / exit morning time subcategory). Thus, based on the preset filtering conditions corresponding to the subcategories, the target filtering conditions for the target event can be determined.

[0042] Optionally, corresponding preset filtering conditions (such as specific parameter values ​​or adjustment weights on basic parameter values) can be set for each dimension's subcategories. Thus, by combining the preset filtering conditions of each dimension's subcategories, the preset filtering conditions for the subcategory to which the scene category belongs can be obtained. The corresponding preset filtering conditions (such as parameter values) can be determined as the target filtering conditions (such as parameter values) for the target event. Alternatively, the total adjustment weight can be obtained by combining all preset filtering conditions (such as adjustment weights on basic parameter values), and this total adjustment weight can be used to adjust the basic parameter values ​​to obtain the target filtering conditions (such as parameter values) for the target event. This application does not impose any limitations on this.

[0043] In some implementations, the event category of the target event can be determined based on event-related information. Based on the event category, a target filtering method for the target event can be determined. The target filtering conditions, the target filtering method, and the event-related information are then used to filter the target event. This allows for adaptive determination of both the target filtering conditions and the target filtering method, enabling adaptive filtering of target events based on different situations. Optionally, the aforementioned target filtering conditions include parameter values ​​corresponding to each target filtering method (such as the target filtering method for each event category), which is not limited in this application.

[0044] Optionally, event categories can be divided into normal event categories and abnormal event categories. Alternatively, event categories can be divided into event categories with preset targets and event categories without preset targets, etc. This application does not limit the event categories. Different target filtering methods can be used to filter target events for different event categories.

[0045] In some embodiments, please refer to Figure 3 For step S13 above, the following steps may be included: S131: Get the event category of the target event.

[0046] S132: Determine whether the event category is an abnormal event category.

[0047] The abnormal event categories include at least one of the following: abnormal behavior of the preset target (such as intrusion behavior, falling behavior, etc.), the occurrence of a preset security event (such as a dangerous event) in the scene, the first appearance of the preset target (such as the first appearance of the preset target within a preset time period, or the first appearance of a certain or a certain type of interested entity target), etc. For example, the first appearance of the preset target is such as the first appearance of the preset target within 10 minutes, etc. This application does not limit the abnormal event categories.

[0048] The preset target is the entity category or entity target that the user is interested in. Since an image frame may contain at least one type of entity target, at least one type of entity target or at least one of the entity targets can be identified as the preset target. For example, entity targets include living beings, objects, moving objects, etc. The preset target can be a type of entity target or entity target that can be pre-configured or determined according to the application scenario. This application does not limit the preset target.

[0049] In some implementations, event categories can be determined using event-related information, and it can be determined whether the event category is an abnormal event category, thereby allowing the event category to be labeled as a normal event category or an abnormal event category.

[0050] In some implementations, scene information and event-related information can be used to determine the event category and whether the event category is an abnormal event category. In this case, the event category can be divided into an abnormal event category or a normal event category. Alternatively, the event category can be determined adaptively by combining scene information. For example, in region 1, the action behavior 1 of the preset target is a normal behavior (normal event category), while in region 2, the action behavior 1 of the preset target is an abnormal behavior (abnormal event category). The event category and whether it is an abnormal event category can be determined adaptively. This application does not limit this.

[0051] If the event category is an abnormal event category, then execute step S133 below; if the event category is not an abnormal event category, that is, a normal event category, then execute step S134 below or execute step 135 below.

[0052] S133: In response to an event category being an abnormal event category, determine that the target event will not be filtered.

[0053] In response to the event category being an abnormal event category, if it is determined that the target event will not be filtered, it is necessary to provide a prompt for the target event of this abnormal event category. This application does not restrict the abnormal event scenario.

[0054] S134: In response to an event category being a normal event category, determine several preset historical events within a preset duration period.

[0055] In response to the event category being classified as a normal event, several preset historical events within a preset duration period are determined. These preset historical events include historical target events reported by the front-end device within a preset duration period prior to the current time of the target event. These acquired preset historical events are used to perform the following steps and compare them with the target event.

[0056] Optionally, each target event reported by a preset device can be added to a preset historical event set to serve as a preset historical event for subsequent target events within a preset time period. This method allows for cyclical comparison of prompt messages for target events within a preset time period (e.g., 1 minute, 5 minutes, 10 minutes, 30 minutes, etc.) to detect periodically repeating events within the preset time period window, thereby filtering out repeating events. Optionally, if at least one of several preset historical events determines that the target event meets the target filtering condition, i.e., the target event is a repeating event within the preset time period, then filtering of the target event is determined. Optionally, several historical events within the preset time period can be sequentially used as the current preset historical events, and the target event can be compared sequentially with the current preset historical events to determine whether the target event meets the target filtering condition. If it does, the target event is filtered; if it does not meet the target filtering condition, the next historical event is used as the current preset historical event, and the above steps are repeated until all historical events have been compared. If none of the historical events meet the target filtering condition, the target event is not filtered.

[0057] S135: Determine whether the event category is an event category with a preset target.

[0058] Determine whether the event category is an event category with a preset goal. For normal event categories, they can be further divided into event categories with preset goals and event categories without preset goals.

[0059] Optionally, if the event category is an event category with a preset target, then perform step S136 below; if the event category is an event category without a preset target, then perform step S137 below.

[0060] Optionally, the event-related information detected for several event image frames includes at least one of target-related information and scene-related information. Different information can be used to compare different event categories to determine whether the target event meets the target filtering conditions.

[0061] S136: In response to an event category having a preset target, the system uses the target-related information of the preset target to determine the first filtering comparison information between the target event and preset historical events, and determines whether to filter the target event based on the first filtering comparison information and the target filtering conditions.

[0062] For event categories with preset targets, in response to the event category having preset targets, the system uses target-related information of the preset targets to determine first filtering comparison information between the target event and preset historical events. Based on the first filtering comparison information and the target filtering conditions, it determines whether to filter the target event. Optionally, it determines whether the first filtering comparison information meets the target filtering conditions. In response to the first filtering comparison information meeting the target filtering conditions, it determines whether to filter the target event.

[0063] Optionally, the target filtering conditions include a preset rate of change threshold and / or a preset identical target condition, wherein the preset rate of change threshold and / or the preset identical target condition are set for the corresponding scene category, that is, the preset rate of change threshold and / or the preset identical target condition can be adaptively set for different scene categories.

[0064] In some implementations, please refer to Figure 4 The target-related information includes entity target features (such as entity features, attribute features, behavioral features, etc.), entity location, entity quantity, etc., and can include features of a 128-dimensional vector. The target-related information of the preset target can be obtained from both the target event and preset historical events to obtain first-level filtering comparison information between the two. For example, the first-level filtering comparison information includes at least one of the following: the rate of change of the number of entities of the preset target, the similarity of the preset targets, etc.

[0065] In some implementations, step S136 may include at least one of the following steps: (1) Determine the rate of change of the number of entities of the target event and the preset historical events.

[0066] The rate of change of the number of entities of the preset target can be obtained by using the difference between the number of entities of the preset target in the target event and the number of entities of the preset target in the preset historical events.

[0067] (2) Determine whether the rate of change of the number of entities is less than or equal to the preset rate of change threshold.

[0068] The method determines whether the rate of change of the number of entities is less than or equal to a preset rate of change threshold. The preset rate of change threshold is 0, 2, 3 or 5, etc. The specific preset rate of change threshold is not limited in this application.

[0069] Optionally, if the rate of change of the number of entities is less than or equal to a preset rate of change threshold, then step (3) is executed; if the rate of change of the number of entities is greater than the preset rate of change threshold, then step (4) is executed.

[0070] When the first filtering comparison information includes the rate of change of the number of entities of the preset target, the target filtering condition includes that the rate of change of the number of entities is less than or equal to the preset rate of change threshold. For example, if the number of entities in the target event and the preset historical event are the same or the difference in the number of entities is less than or equal to the preset rate of change threshold (such as 0, 2, 3, 5, etc.), then the target event meets the target filtering condition, and it is determined to filter the target event.

[0071] (3) In response to the change rate of the number of entities being less than or equal to the preset change rate threshold, the target event is filtered.

[0072] If the rate of change of the number of entities is less than or equal to the preset rate of change threshold, it means that the rate of change of the target event and the preset historical events is small or unchanged, the target event meets the target filtering condition, and it is determined to filter the target event. Therefore, no push notification will be sent for the target event.

[0073] (4) In response to the change rate of the number of entities being greater than the preset change rate threshold, determine the similarity between the target event and the preset target in the preset historical events.

[0074] If the rate of change of the number of entities is greater than the preset rate of change threshold, it indicates that the target event and the preset historical events have changed or the rate of change is large. Then, the similarity between the target event and the preset historical events is determined to decide whether to filter the target event.

[0075] (5) Determine whether the similarity of the preset target satisfies the preset same target condition.

[0076] Optionally, the similarity conditions of the preset targets include at least one of the following: the number of identical preset targets, the target similarity between entity target features of the preset targets, and the event similarity determined by target-related information of the preset targets. In this case, the corresponding target filtering condition is the preset identical target condition.

[0077] Optionally, the preset conditions for the same target include at least one of the following: the number of the same preset targets is greater than a first quantity threshold, the target similarity is greater than a first similarity threshold, and the event similarity is greater than a second similarity threshold.

[0078] In some implementations, for steps (4) to (5) above: Optionally, the similarity between each preset target in the target event and the preset historical event can be obtained by utilizing the entity target features (such as entity features, attribute features, behavioral features, etc.) in both the target event and the preset historical event. If the similarity is greater than a set similarity threshold, the preset targets in both events are determined to be the same target, thereby obtaining the number of the same preset targets in both events. Optionally, the similarity between each preset target can be obtained by utilizing the entity target features in both the target event and the preset historical event, and then the target similarity between the target event and the preset historical event can be obtained based on the similarity between each preset target. Optionally, the similarity can be calculated by using the target-related information of the preset targets in the target event and the preset historical event to obtain the event similarity between the target-related information of the two events. The event similarity is calculated using the feature similarity between the target-related information and at least one feature change rate and the corresponding change rate weight coefficient. Each change rate weight coefficient is set according to the scene category, that is, the change rate weight coefficient corresponding to each change rate can be adaptively determined for each scene category. Optionally, at least one feature change rate includes the change rate of the number of entities, the change rate of attributes, the change rate of location, etc., wherein each feature change rate corresponds to a change rate weighting coefficient.

[0079] In some implementations, the event similarity between the target event and the preset historical events can be obtained by using the feature similarity between the target event and target-related information of the preset target in preset historical events, along with at least one feature change rate and its corresponding change rate weighting coefficient. For example, the feature change value can be obtained by weighted summing of at least one feature change rate and its corresponding change rate weighting coefficient; then, the difference between the feature similarity and the feature change value can be obtained to obtain the event similarity between the target event and the preset historical events. The at least one feature change rate includes the change rate of the number of entities, the change rate of attributes, and the change rate of location. The change rate of the number of entities corresponds to a first weighting coefficient, the change rate of attributes corresponds to a second weighting coefficient, and the change rate of location corresponds to a third weighting coefficient.

[0080] Alternatively, the formula for calculating event similarity can be expressed as follows: Event similarity = Feature similarity - Rate of change in number of entities * First weight coefficient - Rate of change in attribute * Second weight coefficient - Rate of change in location * Third weight coefficient.

[0081] Specifically, taking event A as the target event and event B as a preset historical event as an example, the specific calculations for each part are explained below: Feature similarity: The cosine similarity algorithm can be used to calculate the similarity between the target-related information of two events, thus obtaining the feature similarity, which is the similarity at the semantic level.

[0082] For example, target-related information is represented by feature vectors. The feature similarity can be calculated by performing a cosine similarity calculation on the values ​​of the 128-dimensional feature vectors of the two events. For instance, the vectorized feature vector of event A is: {10101010......10101010......10101010}; the vectorized feature vector of event B is: {11111110......10101010......10101010}; feature similarity = cosine similarity (A, B).

[0083] Rate of change in the number of entities: The rate of change in the number of entities can be used to assess the differences between two events. For example, the percentage change in the number of preset targets in event A and event B can be obtained to get the rate of change in the number of entities.

[0084] For example, if the number of entities in event A is 2 and the number of entities in event B is 1, the rate of change of the number of entities can be determined to be 1.

[0085] Attribute change rate: The attribute change rate can be determined by using the attribute characteristics of the preset target in both events (such as color attribute, appearance type, appearance feature, stage attribute, etc.). Specifically, the attribute characteristics in both events can be compared to obtain the ratio of the number of changed attribute characteristics (i.e., different attribute characteristics) to the total number of attributes, thus obtaining the attribute change rate.

[0086] For example, the definition of attribute features may include a first part color type (such as the upper part of the entity's exterior), a first part length type, a second part color type (such as the lower part of the entity's exterior), a second part length type, a stage classification (such as the entity's time stage classification), a third part length type (such as the entity's head appendages), a third part color type, etc. Each type of attribute feature definition can be represented by a corresponding value.

[0087] For example, the definitions of attribute features are shown in Table 1 below.

[0088] Table 1 Definition of Attribute Features

[0089] Based on the aforementioned attribute characteristics, the rate of change of attribute characteristics between the two events can be calculated, as shown in Table 2 below.

[0090] Table 2. Attribute Change Rate Table for Attribute Features

[0091] Location change rate: Determine the location change rate by using the entity position of the preset target in both events.

[0092] For example, the location of an entity can be represented by the location box corresponding to the location coordinates. The location change rate is obtained by calculating the overlap ratio of the entity locations (location boxes) of the preset target in events A and B. The location change rate ranges from 0 to 1, where 0 means no repetition and 1 means all repetitions.

[0093] Weighting coefficients for each rate of change: Each weighting coefficient can be set based on preset statistical experience, or determined according to the scenario type. This application does not restrict the method of obtaining the weighting coefficients for each rate of change or their specific values.

[0094] In some implementations, the above method can be used to obtain the similarity between the target event and preset targets in preset historical events, thereby determining whether the similarity between the preset targets meets the preset same target condition. For example, if the number of identical preset targets is greater than a first quantity threshold, the target similarity is greater than a first similarity threshold, and the event similarity is greater than a second similarity threshold, then it is determined that the preset same condition is met, that is, the target event meets the target filtering condition.

[0095] Optionally, if the similarity of the preset target satisfies the preset same target condition, the following step (6) is executed; or, if the similarity of the preset target does not satisfy the preset same target condition, the following step (7) is executed.

[0096] (6) In response to similar situations of the preset target satisfying the preset same target conditions, determine to filter the target event.

[0097] If a similar situation to a preset target meets the same preset target condition, and it is determined that the target event will be filtered, then no push notification will be sent for the target event.

[0098] (7) If similar situations to the preset target do not meet the preset same target conditions, it is determined that the target event will not be filtered.

[0099] If a similar situation to a preset target does not meet the preset conditions for the same target, and it is determined that the target event will not be filtered, then a notification message for the target event will be pushed to the target event.

[0100] S137: In response to an event category with no preset target, the second filtering comparison information between the target event and preset historical events is determined using the scene-related information of the target event, and based on the second filtering comparison information and the target filtering conditions, it is determined whether to filter the target event.

[0101] For event categories without preset targets, the target filtering conditions include a third similarity threshold, which is set for the corresponding scene category. That is, different scene categories have corresponding third similarity thresholds.

[0102] The second filtering comparison information between the target event and preset historical events can be determined using the scene-related information of the target event. Based on the second filtering comparison information and the target filtering conditions, it can be determined whether to filter the target event. Optionally, it can be determined whether the second filtering comparison information meets the target filtering conditions. If the target filtering conditions are met, the target event is filtered; otherwise, the target event is not filtered.

[0103] In some implementations, the scene similarity between the scene-related information of the target event and the scene-related information of a preset historical event can be determined, and it can be determined whether the scene similarity is greater than a third similarity threshold. If the scene similarity is greater than the third similarity threshold, it can be determined to filter the target event; or, if the scene similarity is not greater than the third similarity threshold, it can be determined not to filter the target event.

[0104] Optionally, the scene similarity can be obtained by utilizing the similarity between the target event and scene-related information (such as scene description information, scene feature information, etc.) in preset historical events. This scene similarity can refer to the above-mentioned similarity calculation method, and this application does not impose any restrictions on it.

[0105] In some embodiments, after step S13 above, such as after filtering the target event, historically filtered target events can also be recorded. In response to the reappearance of a historically filtered target event after a preset time (i.e., the same target event as the historically filtered target event reappears), the historically filtered target event may not be filtered. For example, the first filtering time of a historically filtered target event can be recorded. Then, if the historically filtered target event reappears, it can be determined whether the first filtering time exceeds the preset time based on the first filtering time and the time of reappearance. If the first filtering time exceeds the preset time, i.e., the historically filtered target event reappears after the preset time, the historically filtered target event may not be filtered in response to the reappearance of the historically filtered target event after the preset time, i.e., a push notification may be sent to the target event.

[0106] The above solution acquires the prompt messages of target events reported by the front-end device. These prompt messages contain several event image frames. The solution detects and analyzes these event image frames to determine the scene information and event-related information of the target event. Based on the scene information, it determines the target filtering conditions for the target event. The solution can adaptively determine the target filtering conditions according to the current scene information. Then, it filters the target event using the target filtering conditions and event-related information. Prompt messages for filtered target events are not pushed. This adaptive filtering of target events and the scene itself allows for more accurate event prompts, avoiding indiscriminate and mechanical event prompts, effectively reducing interference from invalid and redundant event prompts, and improving the accuracy and effectiveness of the prompt information. Furthermore, by not pushing prompt messages for filtered target events, and only performing prompt operations on valid target events that were not filtered, the solution ensures that important events are detected and processed in a timely manner, guaranteeing the timeliness and reliability of event prompts, while also reducing the system's computational and prompt resource consumption, thus improving the overall operational efficiency of event processing.

[0107] In addition, different target filtering methods can be used to filter target events for different event categories. Different feature information can also be used to compare and filter different event categories. By comparing multi-dimensional vector features, the accuracy and speed of target event prompt message filtering can be improved.

[0108] In addition, it can also solve problems such as frequent prompts for irrelevant or invalid events, resulting in a large number of prompts, interference from reconstructed events, continuous false prompts, semantic repetition but different content, and difficulty in screening important prompts. Among them, repeated event interference means that the same event triggers prompts multiple times in a short period of time; continuous false alarms are such as a movement prompt triggered once per minute due to the continuous swaying of tree branches, which will still be pushed once every period of time by traditional deduplication; semantic repetition but different content is such as the same entity moving indoors, capturing images from different angles multiple times, which is identified as multiple independent events and prompted multiple times.

[0109] It is understood that in the above method of specific implementation, the order in which each step is written does not mean a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0110] In some embodiments, this application also provides an event filtering apparatus for implementing the event filtering method of any of the above embodiments.

[0111] Please see Figure 5 , Figure 5This is a schematic diagram of an embodiment of the event filtering device of this application. The event filtering device 20 includes an acquisition module 21, an analysis module 22, and a filtering module. All modules are interconnected.

[0112] The acquisition module 21 is used to acquire the prompt message of the target event reported by the front-end device. The prompt message contains several event image frames.

[0113] The analysis module 22 is used to detect and analyze several event image frames to determine the scene information and event-related information of the target event.

[0114] The filtering module 23 is used to determine the target filtering conditions for the target event based on scene information, and to filter the target event using the target filtering conditions and event-related information. Notably, no push notifications are sent for the filtered target events.

[0115] Optionally, the event filtering device 20 further includes a prompting module 24, which is used to push a prompting message for the target event in response to not filtering the target event.

[0116] It should be noted that the event filtering device and the event filtering method provided in the above embodiments belong to the same concept. The specific ways in which each module and unit performs its operation have been described in detail in the method embodiments, and will not be repeated here. In practical applications, the event filtering device provided in the above embodiments can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. This application does not impose any limitations on this.

[0117] It is understood that the event filtering method in this application can be executed by a computer device, which can be any device with processing capabilities, such as a mobile device, computer, server, etc., and this application does not impose any restrictions on it. In some possible implementations, the event filtering method can be implemented by the processor calling program instructions stored in memory.

[0118] Please see Figure 6 , Figure 6 This is a schematic diagram of the structure of a computer device according to an embodiment of this application. The computer device 30 includes a memory 31 and a processor 32 coupled to each other. The memory 31 stores program instructions, and the processor 32 executes the program instructions to implement the steps of any embodiment of the event filtering method described above. In a specific implementation scenario, the computer device 30 may include, but is not limited to, a microcomputer or a server. In addition, the computer device 30 may also include mobile devices such as laptops and tablets, which are not limited here.

[0119] In this embodiment, processor 32 can also be referred to as a CPU (Central Processing Unit). Processor 32 may be an integrated circuit chip with signal processing capabilities. Processor 32 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor, or processor 32 can be any conventional processor. Furthermore, processor 32 can be implemented using integrated circuit chips.

[0120] The methods described in the above embodiments can be implemented as computer programs; therefore, this application proposes a computer-readable storage medium. Please refer to [link to relevant documentation]. Figure 7 , Figure 7 This is a schematic diagram of a computer-readable storage medium according to an embodiment of the present application. The computer-readable storage medium 40 stores program instructions 41 that can be executed by a processor to implement the steps of any embodiment of the event filtering method described above.

[0121] The computer-readable storage medium 40 in this embodiment includes: a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc., which can store program instructions 41. Alternatively, it can be a server storing the program instructions 41. The server can send the stored program instructions 41 to other devices for execution, or it can execute the stored program instructions 41 itself.

[0122] This application also provides a computer program product comprising a computer program that, when executed by a processor, can implement the steps of the methods described in any of the foregoing embodiments. Specifically, the computer program product can be a software or program product containing a computer program, capable of running on a computing device or stored on any available medium.

[0123] In some embodiments, the functions or modules of the apparatus provided in this application can be used to perform the methods described in the above method embodiments. The specific implementation can be referred to the description of the above method embodiments. For the sake of brevity, this application will not repeat the details here.

[0124] The description of the various embodiments above tends to emphasize the differences between the various embodiments. The similarities or similarities between them can be referred to. For the sake of brevity, the present application will not repeat them here.

[0125] In the several embodiments provided in this application, it should be understood that the disclosed methods and apparatus can be implemented in other ways. For example, the apparatus implementations described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0126] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0127] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0128] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application.

[0129] Obviously, those skilled in the art should understand that the modules or steps of this application described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, and thus stored in a computer-readable storage medium for execution by a computing device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Therefore, this application is not limited to any particular hardware and software combination.

[0130] If the technical solution of this application involves personal information, the product using this technical solution has clearly informed the user of the personal information processing rules and obtained the user's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using this technical solution has obtained the user's separate consent before processing the sensitive personal information, and also meets the requirement of "express consent". For example, at personal information collection devices such as cameras, clear and prominent signs are set up to inform users that they have entered the scope of personal information collection and that personal information will be collected. If an individual voluntarily enters the collection scope, it is deemed that they have agreed to the collection of their personal information; or on the personal information processing device, with clear signs / information informing users of the personal information processing rules, authorization is obtained from the individual through pop-up information or by asking the individual to upload their personal information; wherein, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.

[0131] The above description is merely an embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. An event filtering method, characterized in that, include: Obtain a prompt message for the target event reported by the front-end device, wherein the prompt message contains several event image frames; The event image frames are detected and analyzed to determine the scene information and event-related information of the target event; Based on the scenario information, target filtering conditions for the target event are determined, and the target event is filtered using the target filtering conditions and the event-related information; wherein, no push notification is sent for the filtered target event.

2. The method according to claim 1, characterized in that, The step of determining the target filtering conditions for the target event based on the scene information includes: Based on the scene information, determine the scene category of the target event; Based on the preset filtering conditions corresponding to the scene category, the target filtering conditions for the target event are determined.

3. The method according to claim 2, characterized in that, Determining the scene category of the target event based on the scene information includes: Based on the scene information, determine the sub-category to which the scene category belongs from the multi-level preset categories; The step of determining the target filtering conditions for the target event based on the preset filtering conditions corresponding to the scene category includes: Based on the preset filtering conditions corresponding to the sub-category, the target filtering conditions for the target event are determined.

4. The method according to claim 1 or 2, characterized in that, The filtering of the target event using the target filtering conditions and the event-related information includes: Based on the event-related information, determine the event category of the target event; Based on the event category, a target filtering method for the target event is determined, and the target event is filtered using the target filtering conditions, the target filtering method, and the event-related information.

5. The method according to claim 4, characterized in that, The event-related information includes at least one of target-related information and scenario-related information; The step of determining the target filtering method for the target event based on the event category, and filtering the target event using the target filtering conditions, the target filtering method, and the event-related information, includes: Determine whether the event category is an event category with a preset target; In response to the event category being an event category with a preset target, the system uses target-related information of the preset target to determine first filtering comparison information between the target event and preset historical events, and based on the first filtering comparison information and the target filtering conditions, determines whether to filter the target event; or In response to the event category being an event category without a preset target, the second filtering comparison information between the target event and preset historical events is determined using the scene-related information of the target event, and based on the second filtering comparison information and the target filtering conditions, it is determined whether to filter the target event.

6. The method according to claim 5, characterized in that, The target filtering conditions include a preset rate of change threshold and / or a preset identical target condition, wherein the preset rate of change threshold and / or the preset identical target condition are set for the corresponding scene category; The step of determining first filtering comparison information between the target event and preset historical events using target-related information of a preset target, and determining whether to filter the target event based on the first filtering comparison information and the target filtering conditions, includes: Determine the rate of change of the number of entities representing the target event compared to the preset historical events; In response to the rate of change of the number of entities being less than or equal to a preset rate of change threshold, it is determined that the target event should be filtered; or... In response to the rate of change of the number of entities being greater than a preset rate of change threshold, the similarity between the target event and a preset target in a preset historical event is determined; In response to situations where similarities to the preset target satisfy the preset identical target condition, it is determined that the target event will be filtered. And / or, the target filtering condition includes a third similarity threshold, which is set according to the corresponding scene category; The step of determining second filtering comparison information between the target event and preset historical events using scene-related information of the target event, and determining whether to filter the target event based on the second filtering comparison information and the target filtering conditions, includes: Determine the scene similarity between the scene-related information of the target event and the scene-related information of a preset historical event; In response to the scene similarity being greater than a third similarity threshold, it is determined that the target event will be filtered.

7. The method according to claim 6, characterized in that, The similarity of the preset targets includes at least one of the following: the number of identical preset targets, the target similarity between the entity target features of the preset targets, and the event similarity determined by the target-related information of the preset targets. The event similarity is calculated by using the feature similarity between the target-related information and at least one feature change rate and the corresponding change rate weight coefficient. The change rate weight coefficient is set according to the scene category. The preset conditions for identical targets include at least one of the following: the number of identical preset targets is greater than a first quantity threshold, the target similarity is greater than a first similarity threshold, and the event similarity is greater than a second similarity threshold.

8. The method according to claim 5, characterized in that, Before determining whether the event category is an event category with a preset target, the following steps are included: In response to the event category being a normal event category, several preset historical events within a preset time period are determined; wherein, the several preset historical events within the preset time period include historical target events reported by the front-end device within a preset time period prior to the current time of the target event; or, In response to the event category being an abnormal event category, it is determined that the target event will not be filtered; wherein, the abnormal event category includes at least one of the following: the preset target exhibits abnormal behavior, a preset security event occurs in the scene, or the preset target appears for the first time; And / or, after filtering the target event, it includes: Record the target events of the historical filtering; If the target event of the historical filtering occurs again after a preset time, the historical filtering target event will not be filtered.

9. The method according to claim 5, characterized in that, The notification message also includes first detection information of the target event by the front-end device, wherein the first detection information includes at least one of the number of entity targets and the location of entity targets; The detection and analysis of the plurality of event image frames to determine the scene information and event-related information of the target event includes: Using the first detection information of the target event, select from the plurality of event image frames to obtain the selected event image frames; The selected event image frames are detected to obtain second detection information; The first detection information, the second detection information, and the selected event image frame are detected using a large language model to obtain third detection information, wherein the third detection information includes at least a scene category; By combining the first detection information, the second detection information, and the third detection information, scene information and event-related information are obtained.

10. A computer device, characterized in that, The method includes a memory and a processor coupled to each other, the memory storing program instructions, and the processor executing the program instructions to implement the method according to any one of claims 1 to 9.

11. A computer-readable storage medium, characterized in that, The system stores program instructions that, when executed by a processor, implement the method described in any one of claims 1 to 9.

Images