KR-20260064344-A - Method And Device for Managing Recurring Events

Abstract

A method and apparatus for managing repeatedly occurring events are disclosed. According to one aspect of the present disclosure, a computer-implemented method for managing a repeatedly occurring event comprises: a process of extracting a plurality of feature expressions from one or more video clips in which a target event is detected; a process of clustering the plurality of feature expressions to generate one or more clusters; a process of selecting a candidate feature expression representing the cause of occurrence of the target event from each of the one or more clusters; and a process of storing each of the selected one or more candidate feature expressions together with identification information for the target event in a database.

Inventors

• 박정환
• 마춘페이
• 최준향
• 이병원

Assignees

• 에스케이텔레콤 주식회사

Dates

Publication Date

20260507

Application Date

20241031

Claims (10)

1. As a computer implementation method for managing repeatedly occurring events, A process of extracting multiple feature representations from one or more video clips in which a target event is detected; A process of clustering the above plurality of feature expressions to generate one or more clusters; A process of selecting candidate feature expressions representing the cause of occurrence of the target event from each of the one or more clusters; and The process of storing one or more selected candidate feature expressions in a database, each along with identification information for the aforementioned target event. A computer implementation method including
2. In paragraph 1, A process of generating text explaining the cause of occurrence of the target event from at least one candidate feature representation using a vision-language model. A computer implementation method including further
3. In paragraph 1, The above database further stores weights assigned to each candidate feature representation, and A computer implementation method in which the above weights are updated based on one or more first novel feature representations extracted from a new video clip in which the target event is detected.
4. In paragraph 3, The process of selecting the candidate feature expression with the highest weight among the candidate feature expressions corresponding to the above target event as the representative feature expression with high explanatory power for the above target event. A computer implementation method including further
5. In paragraph 4, A process of extracting one or more second novel feature representations from one or more new video frames; A process of querying a representative feature expression similar to at least some of the one or more second novel feature expressions in the above database; and A process of generating an alarm that indicates the possibility of an event occurring corresponding to the representative feature expression retrieved above. A computer implementation method including further
6. In paragraph 3, For each of the above one or more candidate feature expressions, the method further includes a process for determining the likelihood of each candidate feature expression appearing at the time of the occurrence of the target event, based on at least one of the following: the number of times the candidate feature expression appears among the above one or more video clips, the time interval between the frame in which the target event is detected and the frame in which the candidate feature expression is extracted, and the distribution of the cluster containing the candidate feature expression. The above-mentioned storage process is a computer implementation method that stores the likelihood determined for each candidate feature representation as an initial weight for each candidate feature representation.
7. In paragraph 1, The above extraction process is, A computer-implemented method for extracting one or more feature representations in each video clip using one or more frames that are temporally preceding the frame in which the target event is detected.
8. In paragraph 1, Prior to the above extraction process, The process further includes receiving input from the user regarding the type and detailed conditions of the target event to be analyzed, and selecting one or more of the above video clips, The above detailed conditions are set based on object detection results or classification results for at least one frame within each video clip, and A computer-implemented method wherein identification information for the above target event includes information capable of identifying the type of the above target event and the above detailed conditions.
9. Memory for storing instructions; and at least one processor, comprising The above at least one processor executes the above instructions, Extract multiple feature representations from one or more video clips in which a target event is detected, and Clustering the above multiple feature expressions to generate one or more clusters, and Select candidate feature expressions representing the cause of occurrence of the target event from each of the above one or more clusters, and A device that stores one or more selected candidate feature expressions, each along with identification information for the target event, in a database.
10. A computer program stored on a computer-readable recording medium to execute the processes included in the method according to any one of paragraphs 1 through 8.

Description

Method and Device for Managing Recurring Events The present disclosure relates to a method and apparatus for managing repeatedly occurring events. The following description merely provides background information related to the present embodiment and does not constitute prior art. As the importance of security and surveillance increases in modern society, the volume of video data collected from devices such as CCTVs and dashcams is skyrocketing. These devices are used as a primary means to detect various crimes and accidents, including theft, intrusion, and violence; in particular, intelligent CCTV systems provide features that detect specific events in real time and enable administrators to quickly review the relevant footage. Conventional intelligent CCTV systems operate by storing specific time periods or frames before and after the moment a pre-defined event (e.g., intrusion, theft, injury) occurs, thereby providing context for the event. While this method assists managers in tracking incidents and taking necessary actions, it still has limitations in that human review is required during the analysis process following event detection. In other words, whenever an event occurs, managers must manually examine pre- and post-event video clips to identify the cause and circumstances. This process is time-consuming and inefficient, and there is a risk of missing critical information if managers do not review every single clip. Furthermore, since the prediction and response to recurring events are not automated, it is difficult to take timely follow-up actions, and there is a risk that proactive responses to foreseeable risks may not be implemented. FIG. 1 is a block diagram schematically showing an event management device according to one embodiment of the present disclosure. FIG. 2 is a drawing referenced to explain the operation of extracting a feature representation from a video clip according to one embodiment of the present disclosure. FIGS. 3a and 3b are drawings referenced to explain the operation of selecting candidate feature representations for an event according to one embodiment of the present disclosure. FIG. 4 is a drawing referenced to explain the operation of generating text explaining the cause of an event according to one embodiment of the present disclosure. FIG. 5 is a flowchart illustrating an event management method according to one embodiment of the present disclosure. FIG. 6 is a schematic block diagram of an exemplary computing device that can be used to implement the devices and methods described in the present disclosure. Some embodiments of the present disclosure are described in detail below with reference to exemplary drawings. It should be noted that in assigning reference numerals to the components of each drawing, the same components are given the same reference numeral whenever possible, even if they are shown in different drawings. Furthermore, in describing the present disclosure, if it is determined that a detailed description of related known components or functions could obscure the essence of the present disclosure, such detailed description is omitted. In describing the components of the embodiments according to the present disclosure, symbols such as first, second, i), ii), a), b), etc., may be used. These symbols are intended only to distinguish the components from other components, and the essence, order, or sequence of the components is not limited by the symbols. When a part in the specification is described as 'comprising' or 'having' a component, this means that, unless explicitly stated otherwise, it does not exclude other components but may include additional components. The detailed description set forth below, together with the accompanying drawings, is intended to describe exemplary embodiments of the present disclosure and is not intended to represent the only embodiment in which the present disclosure can be practiced. FIG. 1 is a block diagram schematically showing an event management device according to one embodiment of the present disclosure. The event management device (10) may include all or part of an event detection module (100), an analysis module (120), a coordination module (140), and a prediction module (150). Not all blocks shown in FIG. 1 are essential components, and some blocks included in other embodiments may be added, changed, or deleted. Meanwhile, the components shown in FIG. 1 represent functionally distinct elements, and at least one component may be implemented in a form that is integrated with one another in an actual physical environment. The event detection module (100) can detect the occurrence of a predefined event from the input video. Here, the predefined event may include, for example, assault, falling, intrusion, theft, etc., but is not limited to these examples. The event detection module (100) can detect a predefined event by using various known image processing algorithms and/or voice analysis algorithms. For example, the