Security and protection deployment and control management method and device

Abstract

The invention discloses a security and protection deployment and control management method and device. The method comprises the following steps: acquiring an initial deployment and control task parameter which is input by a target object in an interactive interface and comprises a deployment and control scene, determining extension prompt information matched with the initial deployment and control task parameter by utilizing a pre-trained computer vision large model, and displaying the extension prompt information in the interactive interface, obtaining a target deployment and control task parameter which is input by the target object in the interactive interface and obtained after the initial deployment and control task parameter is expanded according to the expansion prompt information, generating a deployment and control task according to the target deployment and control task parameter, and in the process of executing the deployment and control task, obtaining a monitoring image in real time, and comprehensively analyzing the monitoring image and the target deployment and control task parameters by using a computer vision large model to obtain a deployment and control task matching result corresponding to the monitoring image. According to the method, the technical problems that a related model algorithm in a security and protection deployment and control scene can only process a specific visual task and the intelligent degree is not high are solved.

Description

Technical Field

[0001] This application relates to the field of computer vision technology, and more particularly, to a security layout management method and device. Background Art

[0002] With the development of artificial intelligence technology, more and more security monitoring systems begin to use large computer vision models as intelligent auxiliary tools for monitoring to identify or detect target objects in the monitoring scene. Generally, large computer vision models can make relatively accurate identifications or detections of target objects in the monitoring scene under specific scenarios. However, due to the defects of the large computer vision models themselves or algorithm designs, there may still be cases of incorrect identifications or detections. For example, in some complex scenarios, the large computer vision models may give false alarms or miss detections when identifying or detecting target objects, ultimately resulting in the failure of the layout task execution.

[0003] In response to the above problems, no effective solutions have been proposed yet. Summary of the Invention

[0004] Embodiments of this application provide a security layout management method and device to at least solve the technical problem that relevant model algorithms in security layout scenarios can only handle specific vision tasks and have low intelligence.

[0005] According to one aspect of the embodiments of this application, a security layout management method is provided, including: obtaining initial layout task parameters input by a target object in an interaction interface, where the initial layout task parameters at least include a layout scene; using a pre-trained large computer vision model to determine extended prompt information matching the initial layout task parameters, and displaying the extended prompt information in the interaction interface; obtaining target layout task parameters obtained by the target object in the interaction interface by expanding the initial layout task parameters based on the extended prompt information, and generating a layout task based on the target layout task parameters; during the execution of the layout task, obtaining monitoring images in real time, and comprehensively analyzing the monitoring images and the target layout task parameters using the large computer vision model to obtain a layout task matching result corresponding to the monitoring images.

[0006] Optionally, obtaining the initial layout task parameters input by the target object in the interaction interface includes: receiving the initial layout task parameters in text modality input by the target object in the interaction interface; or, receiving the initial layout task parameters in voice modality input by the target object in the interaction interface, and converting the initial layout task parameters into text modality using speech recognition technology; where the initial layout task parameters include at least one of the following: task name, task object handling type, layout time limit, layout time period, layout area, layout threshold, layout scene.

[0007] Optionally, a pre-trained computer vision big model is used to determine extended prompt information that matches the initial control task parameters, including: using the computer vision big model to respectively determine the first matching degree between each historical task in the historical task database and the initial control task parameters, and sorting the obtained multiple first matching degrees from large to small; if there is a first matching degree greater than a preset matching threshold, determining a preset number of target first matching degrees ranked at the top from the first matching degrees greater than the preset matching threshold, and obtaining historical task data of the historical tasks corresponding to the target first matching degrees as extended prompt information, wherein the historical task data includes at least one of the following: historical task description information, historical monitoring video clips, and historical monitoring images; if there is no first matching degree greater than the preset matching threshold, using the computer vision big model to generate extended prompt information that matches the initial control task parameters, wherein the type of the extended prompt information includes at least one of the following: a task image corresponding to the initial control task parameters, a text description of the control task parameters to be supplemented.

[0008] Optionally, a computer vision big model is used to respectively determine the first matching degree between each historical task in the historical task database and the initial control task parameters, including: using the computer vision big model to perform semantic analysis on the initial control task parameters to determine the first subtasks of multiple dimensions corresponding to the initial control task parameters and the association relationship between each first subtask; determining a dependency graph corresponding to multiple first subtasks based on the association relationship, and determining the weight coefficients corresponding to the first subtasks of each dimension based on the dependency graph; using the computer vision big model to analyze each historical task to determine the second subtasks of multiple dimensions corresponding to each historical task; for each historical task, determining the similarity between the second subtask and the first subtask of the same dimension corresponding to the historical task and the initial control task parameters, and performing weighted calculation on each similarity obtained based on the weight coefficient to obtain the first matching degree between the historical task and the initial control task parameters.

[0009] Optionally, weighted calculation is performed on each obtained similarity according to a weight coefficient to obtain a first matching degree between the historical task and the initial control task parameters, including:

[0010] The first matching degree between the historical task and the initial control task parameters is calculated using the following formula:

[0011]

[0012] Where S(T1,T2) represents the first matching degree between the initial control task parameter T1 and the historical task T2; n represents the total number of dimensions of the first subtask corresponding to the initial control task parameter and the second subtask corresponding to the historical task; sim(f 1i ,f 2i ) represents the first subtask f of the i-th dimension1i and the second subtask f of the i-th dimension 2i The similarity between them. If there is no first subtask of the i-th dimension or no second subtask of the i-th dimension, sim(f 1i , f 2i ) = 0; w i represents the weight coefficient of the first subtask of the i-th dimension. If there is no first subtask of the i-th dimension, w i = 0; d i is a dependence factor, indicating whether there is a second subtask with the same upstream dependence task dimension as the first subtask of the i-th dimension. If so, d i = 1. If not, d i = 0.

[0013] Optionally, obtain the target deployment task parameters after expanding the initial deployment task parameters according to the extended prompt information input by the target object in the interaction interface, including: repeatedly execute the following process: obtain the deployment task parameters expanded according to the extended prompt information input by the target object in the interaction interface, and detect whether the target object inputs a confirmation instruction; if it is detected that the target object inputs a confirmation instruction, terminate the loop process and use the expanded deployment task parameters as the target deployment task parameters; if it is not detected that the target object inputs a confirmation instruction, use the computer vision large model to determine new extended prompt information that matches the expanded deployment task parameters, and display the new extended prompt information in the interaction interface.

[0014] Optionally, obtain monitoring images in real time, including: within a preset deployment period, obtain multiple monitoring videos collected by multiple cameras in a preset deployment area in real time; extract the current frame images of each monitoring video as monitoring images.

[0015] Optionally, comprehensively analyze the monitoring images and the target deployment task parameters by using the computer vision large model to obtain the deployment task matching result corresponding to the monitoring images, including: using the computer vision large model to extract the semantic features corresponding to the target deployment task parameters, where the semantic features are at least used to reflect the deployment intention of the target object; using the computer vision large model to extract the image features corresponding to the monitoring images; determining the second matching degree between the semantic features and the image features; in the case where the second matching degree is not less than a preset deployment threshold, determine that the monitoring images and the deployment tasks match and need to be processed; in the case where the second matching degree is less than the deployment threshold, determine that the monitoring images and the deployment tasks do not match and do not need to be processed.

[0016] Optionally, the security deployment management method further includes: when it is determined that the monitoring image matches the deployment task and needs to be processed, executing a security management strategy corresponding to a preset task object disposal type, where the security management strategy includes at least one of the following: tracking the task object in the monitoring image, generating a warning prompt message, and automatically dialing an alarm call.

[0017] According to another aspect of the embodiments of the present application, there is also provided a security deployment management device, including: an acquisition module, configured to acquire initial deployment task parameters input by a target object in an interaction interface, where the initial deployment task parameters at least include a deployment scenario; a prompt module, configured to use a pre-trained large computer vision model to determine extended prompt information matching the initial deployment task parameters, and display the extended prompt information in the interaction interface; a generation module, configured to acquire target deployment task parameters obtained by the target object expanding the initial deployment task parameters based on the extended prompt information in the interaction interface, and generate a deployment task based on the target deployment task parameters; an analysis module, configured to, during the execution of the deployment task, acquire monitoring images in real time, and comprehensively analyze the monitoring images and the target deployment task parameters using the computer vision model to obtain a deployment task matching result corresponding to the monitoring images.

[0018] According to another aspect of the embodiments of the present application, there is also provided a computer program product, including: a computer program, where when the computer program is executed by a processor, the above-mentioned security deployment management method is implemented.

[0019] According to another aspect of the embodiments of the present application, there is also provided an electronic device, including: a memory and a processor, where a computer program is stored in the memory, and the processor is configured to execute the above-mentioned security deployment management method through the computer program.

[0020] In the embodiments of the present application, a pre-trained large computer vision model is used to expand the initial deployment task parameters, and deployment task parameters with more complete semantics are obtained. Using the expanded deployment task parameters as the target deployment task parameters helps to generate a deployment task that better matches the deployment intention of the target object, laying a foundation for the intelligent execution and matching of subsequent deployment tasks, and thus solving the technical problem that related model algorithms in the security deployment scenario can only process specific vision tasks and have a low degree of intelligence. Description of the Drawings

[0021] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings:

[0022] Figure 1It is a schematic diagram of an optional security layout management method according to an embodiment of the present application;

[0023] Figure 2 It is a schematic diagram of an optional new layout task page according to an embodiment of the present application;

[0024] Figure 3 It is a schematic diagram of an optional process for obtaining target layout task parameters according to an embodiment of the present application;

[0025] Figure 4 It is a schematic diagram of an optional security layout management device according to an embodiment of the present application;

[0026] Figure 5 It is a schematic diagram of an optional electronic device according to an embodiment of the present application. Detailed implementation manners

[0027] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0028] It should be noted that the terms "first", "second", etc. in the specification, claims and drawings of the present application are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily need to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0029] To better understand the embodiments of the present application, some nouns or terms that appear in the description process of the embodiments of the present application are translated and explained as follows:

[0030] Natural semantic understanding: Natural semantic understanding is a core research direction in the field of artificial intelligence, aiming to enable the computer to not only perform simple lexical matching or syntactic analysis on the text, but also delve into the semantic level of the language to understand the true meaning behind the text.

[0031] Computer Vision Large Model: Computer Vision (CV) large models have a large number of parameters and can process complex visual information. Compared with ordinary deep learning models, they have a more complex structure, require more computing resources, consider multi-task learning, can handle multiple visual tasks simultaneously, and have a wider range of applications.

[0032] Generalization: Generalization refers to the ability of a model or algorithm to process new data that the model has not seen during the training process. Generalization is an important indicator for evaluating model performance and reflects the adaptability and effectiveness of the model when facing unknown situations.

[0033] Multi-task Learning: Training a single model to perform multiple related tasks simultaneously. The core idea of the method is that there may be common learning features or patterns among different tasks. By sharing representations, the performance of the model on each task can be improved, making the model more robust.

[0034] In the embodiments of this application, the information collected is information and data authorized by the user or fully authorized by all parties. Moreover, for the processing of relevant data such as collection, storage, use, processing, transmission, provision, disclosure, and application, all comply with the relevant laws, regulations, and standards of the relevant countries and regions, necessary confidentiality measures are taken, it does not violate public order and good customs, and a corresponding operation entry is provided for users to choose to authorize or reject.

[0035] Embodiment 1

[0036] According to the embodiments of this application, a security deployment management method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0037] Figure 1 is a schematic flowchart of a security deployment management method provided according to the embodiments of this application, as Figure 1 shown, the method includes the following steps:

[0038] Step S102, obtain the initial deployment task parameters input by the target object in the interaction interface, where the initial deployment task parameters at least include the deployment scenario.

[0039] When initially creating a surveillance task, an interactive interface is provided to the target object. The interactive interface may include various surveillance task parameters. The surveillance scenario describes the specific objectives and background of the monitoring or surveillance task. For example, "looking for a person suspected of theft in a mall". This description is directly related to the objects or behaviors that the system needs to detect and identify, and is a prerequisite for executing the surveillance task. Without a scenario description, the system will have no idea what to focus on and thus cannot conduct effective monitoring and analysis. Therefore, the surveillance scenario is a required parameter item.

[0040] Step S104: Use a pre-trained large computer vision model to determine extended prompt information that matches the initial surveillance task parameters and display the extended prompt information in the interactive interface.

[0041] When the user enters the initial surveillance task parameters on the interactive interface, these parameters at least include a text description of the scene to be monitored. For example, "looking for a middle-aged man who walks quickly and keeps looking back in the mall". At this time, the task of the system is to convert this natural language description into a machine-executable surveillance instruction.

[0042] Through the pre-trained large computer vision model, the text description is converted into specific visual features and behavior patterns. For example, for the above-mentioned surveillance scenario description, the large model can identify key information points such as "middle-aged man", "walking quickly", and "looking back", and further analyze the visual features behind these information, such as the age, walking speed, and head movements of the individual.

[0043] Using its built-in multi-task learning mechanism, the large model can simultaneously process various visual tasks such as image classification, object detection, and behavior recognition. This enables it to quickly generate a series of extended prompt information related to the initial surveillance task parameters after receiving them. These information may be specific object recognition instructions ("identify individuals wearing red tops"), behavior detection requirements (such as "detect fast movement"), or even environmental conditions (such as "in a high-traffic area"). Together, they form a more comprehensive and specific definition of the surveillance task.

[0044] Step S106: Obtain the target surveillance task parameters after the target object expands the initial surveillance task parameters based on the extended prompt information in the interactive interface, and generate a surveillance task based on the target surveillance task parameters.

[0045] Suppose the surveillance scenario is: "Search for a middle-aged man who is walking quickly in the mall and constantly looking back." Based on the above surveillance scenario, the pre-trained large computer vision model may expand the following information: the age range of the middle-aged man, the color of his clothing, specific areas of the mall, walking speed, and other extended information. Based on the initial parameters of the initial surveillance task, the pre-trained large computer vision model will generate a series of extended prompt messages. These prompt messages are designed to supplement and improve the initial description, providing more specific and detailed surveillance elements. These extended information are presented to the user in an intuitive way on the interaction interface, such as through drop-down lists, sliders, or text input boxes, allowing the user to adjust and confirm details according to actual needs. When the user completes the feedback and confirmation of the extended prompt messages, the system then determines the target surveillance task parameters corresponding to the extended prompt messages and intelligently generates the task.

[0046] Step S108, during the execution of the surveillance task, continuously obtain surveillance images, and comprehensively analyze the surveillance images and the target surveillance task parameters using the large computer vision model to obtain the surveillance task matching result corresponding to the surveillance images.

[0047] When the obtained surveillance image matches the surveillance task, it indicates that the surveillance task has discovered the target to be searched, and correspondingly, relevant warning measures need to be taken.

[0048] Next, combined with Figure 1 and the specific implementation process, each step of the security surveillance management method will be described.

[0049] When obtaining the initial surveillance task parameters input by the target object in the interaction interface, the following situations may be included: receiving the initial surveillance task parameters in text mode input by the target object in the interaction interface; receiving the initial surveillance task parameters in voice mode input by the target object in the interaction interface, and using speech recognition technology to convert the initial surveillance task parameters into text mode;

[0050] First of all, the initial surveillance task parameters may include various types of parameters. Common parameters include but are not limited to task name, task object disposal type, surveillance time limit, surveillance period, surveillance area, surveillance threshold, surveillance scenario. Figure 2The task interface when creating a new control task is shown. The interface contains various common control task parameters, such as task name, disposal type, control validity period, control threshold, control time period, etc. And the right side of the interface shows the operation prompt for inputting the control task scenario. Generally speaking, the task name provides a clear identification label for the task, facilitating management and retrieval in the system; the task object disposal type is used to define the type of action that the system should take when the task object is detected; the control validity period is used to set the effective time of the task, determining how long the control will continue; the control time period is used to limit the specific execution time of the control task, applicable to security concerns during a specific time period; the control area is used to specify the geographical scope of the control task to ensure reasonable allocation of resources; the control threshold is used to set the minimum matching standard for the model during comparison or detection; detection results below this value will be ignored, and the control scenario provides a detailed description of the specific monitoring target, which is the core information of the control task.

[0051] Among the numerous control task parameters, the control scenario is the core of the entire control task. Specifically, the control scenario parameters are input in the following ways. For example, directly edit the text in the control scenario parameter input box: "Look for a middle-aged man who is walking quickly in the mall and constantly looking back", or first input through voice: "Look for a middle-aged man who is walking quickly in the mall and constantly looking back", and then convert the above voice into text content through speech recognition technology.

[0052] To further obtain a more comprehensive and specific definition of the control task, first, based on the initial security control parameters, obtain extended prompt information that matches the initial security control parameters. The extended prompt information can be obtained through the following steps S1 - S2:

[0053] Step S1, use the computer vision large model to respectively determine the first matching degree between each historical task in the historical task database and the initial control task parameters, and sort the obtained multiple first matching degrees from large to small.

[0054] Specifically, when calculating the first matching degree, it can be calculated through the following steps S111 - S114:

[0055] Step S111, use the computer vision large model to perform semantic analysis on the initial control task parameters, determine the first subtasks of multiple dimensions corresponding to the initial control task parameters and the association relationship between each first subtask;

[0056] Step S112, determine the dependency graph corresponding to the multiple first subtasks according to the association relationship, and determine the weight coefficient corresponding to each first subtask of each dimension according to the dependency graph;

[0057] Step S113: Analyze each historical task using a large computer vision model to determine multiple dimensions of second subtasks corresponding to each historical task;

[0058] Step S114: For each historical task, determine the similarity between the second subtask and the first subtask of the same dimension corresponding to the historical task and the initial deployment task parameters, and perform weighted calculation on each obtained similarity according to the weight coefficient to obtain the first matching degree between the historical task and the initial deployment task parameters.

[0059] Specifically, performing weighted calculation on each obtained similarity according to the weighted coefficient to obtain the first matching degree between the historical task and the initial deployment task parameters can be calculated by the following formula:

[0060]

[0061] In the formula, S(T1, T2) represents the first matching degree between the initial deployment task parameter T1 and the historical task T2; n represents the total number of dimensions of the first subtask corresponding to the initial deployment task parameter and the second subtask corresponding to the historical task; sim(f 1i , f 2i ) represents the similarity between the first subtask f 1i and the second subtask f 2i of the i-th dimension. If there is no first subtask of the i-th dimension or no second subtask of the i-th dimension, sim(f 1i , f 2i ) = 0; w i represents the weight coefficient of the first subtask of the i-th dimension. If there is no first subtask of the i-th dimension, w i = 0; d i is a dependence factor, indicating whether there is a second subtask with the same upstream dependence task dimension as the first subtask of the i-th dimension. If so, d i = 1. If not, d i = 0.

[0062] Calculate the first matching degree between each historical task and the initial deployment task parameters in the above manner, and sort the first matching degrees from largest to smallest.

[0063] Step S2: Determine whether there is a first matching degree greater than the preset matching threshold. If so, determine a preset number of target first matching degrees with the top rankings from the first matching degrees greater than the preset matching threshold, and obtain the historical task data of the historical tasks corresponding to the target first matching degrees as extended prompt information. The historical task data includes at least one of the following: historical task description information, historical monitoring video clips, and historical monitoring images. If not, use the computer vision large model to generate extended prompt information that matches the initial deployment control task parameters. The types of the extended prompt information include at least one of the following: task images corresponding to the initial deployment control task parameters, and text descriptions of the deployment control task parameters to be supplemented.

[0064] Among them, the preset number can be defined as needed. Assuming the preset number is 5, the above value of the preset number is only for example and does not constitute a specific limitation. When calculating the first matching degree for 10 historical tasks and the calculation results show that only the first matching degrees of 3 historical tasks are greater than the preset matching threshold, then directly use the historical task data of the above 3 historical tasks as the extended prompt information.

[0065] Furthermore, after obtaining the extended prompt information, it is necessary to expand the initial deployment control task parameters based on the extended prompt information, and use the obtained expanded initial deployment control task parameters as the target deployment control task parameters. The following combines Figure 3 to explain in detail the process of obtaining the target deployment control task parameters.

[0066] First, obtain the deployment control task parameters expanded based on the extended prompt information input by the target object in the interaction interface, and detect whether the target object inputs a confirmation instruction.

[0067] Secondly, detect whether the target object inputs a confirmation instruction. If it is detected that the target object inputs a confirmation instruction, terminate the loop process and use the expanded deployment control task parameters as the target deployment control task parameters; if it is not detected that the target object inputs a confirmation instruction, use the computer vision large model to determine new extended prompt information that matches the expanded deployment control task parameters, and display the new extended prompt information in the interaction interface, and repeat the above process.

[0068] From the above process, it can be seen that obtaining the target deployment control task parameters is based on the adjustment loop of the extended prompt information according to the user feedback. This is essentially a process of retraining the model, which allows the computer vision large model to perform self-learning and optimization while executing tasks, thereby enhancing the model's processing ability for new data and unknown situations and improving its generalization performance.

[0069] The following is an example to illustrate the process of determining the target surveillance task parameters. Suppose the surveillance scenario parameter is: "Monitor suspicious persons wearing black coats in the mall". The system displays extended prompt information on the interaction interface that matches the initial surveillance task parameters, which includes possible task images and text descriptions to be supplemented. By observing this information, the user can further supplement or adjust the details of the surveillance task, such as the precise age range or specific suspicious behavior patterns. If the user does not click the confirmation instruction after viewing the extended information, the pre-trained large computer vision model needs to intervene and generate new extended prompt information again based on the surveillance task parameters supplemented by the user. For example, if the user has provided a style description of "black coat", the model can further generate refined prompt information such as "Does the coat have a hat?" and "Accessory color" to help the user more comprehensively define the surveillance target. Throughout the process, the user continuously adjusts the surveillance task parameters according to the extended prompt information displayed on the interface until all key information points are clearly defined or the user believes that the current level of extension is sufficient. This mechanism ensures the accuracy of the surveillance description, reduces false alarms and missed alarms during the surveillance process, and improves the system execution efficiency.

[0070] When the user expresses satisfaction with the extended surveillance task parameters by clicking "Confirm" or a similar instruction, the system will terminate the loop process, use these parameters as the final target surveillance task parameters, generate specific surveillance strategies. Subsequently, the system will, based on these parameters, activate relevant surveillance devices and analysis algorithms to conduct real-time surveillance of the designated area, identify and track suspicious persons or behaviors that match the description, ensuring the effective execution of the surveillance task.

[0071] Specifically, when conducting real-time surveillance, the comprehensive analysis of the surveillance image and the target surveillance task parameters by using the large computer vision model to obtain the matching result of the surveillance task corresponding to the surveillance image can be achieved in the following way: Use the large computer vision model to extract the semantic features corresponding to the target surveillance task parameters, where the semantic features are at least used to reflect the surveillance intention of the target object; Use the large computer vision model to extract the image features corresponding to the surveillance image; Determine the second matching degree between the semantic features and the image features; In the case where the second matching degree is not less than the preset surveillance threshold, determine that the surveillance image needs to be processed for matching with the surveillance task; In the case where the second matching degree is less than the surveillance threshold, determine that the surveillance image does not match the surveillance task and no processing is required.

[0072] When determining the second matching degree between the semantic features and the image features, it involves the comprehensive processing of the task results in two different fields of natural language understanding and computer vision. By using the multi-task learning ability of the large model, the model can simultaneously process multiple related tasks, such as object detection, behavior recognition, scene understanding, etc., rather than just a single visual recognition or text understanding task.

[0073] Specifically, when the large computer vision model extracts semantic features of the target deployment task parameters, it can first preprocess the information in the text modality. Common preprocessing methods include Tokenization, RemoveStopwords, etc. Taking the analysis of the semantic features corresponding to the deployment scenario as an example below.

[0074] After obtaining the preprocessed deployment scenario parameters, the large computer vision model first performs syntactic analysis on the preprocessed text and constructs a syntax tree T to represent the structure of the sentence; then extracts key information from the text information. The extraction of key information can be achieved by identifying the dependency relationship R and the semantic role SRL:

[0075] Key_Information = NER(T, R, SRL)

[0076] Among them, Key_Information represents the extracted key information, NER is the named entity recognition function, T is the syntax tree, R is the dependency relationship, and SRL is the semantic role annotation.

[0077] After obtaining the extracted key information, perform word embedding operations, convert the words in the text into numerical vector forms, and use Transformer to encode the word vectors in the sentence into a single vector representation of the sentence, identify the dependency relationship between the words in the sentence, and construct a dependency graph. By constructing a complex dependency graph, the depth and breadth of the understanding of the deployment instruction can be enhanced. And the attention mechanism plays a core role in this link, enabling the model to focus on the key parts of the input sequence and improving the processing effect. The self-attention mechanism can be expressed as:

[0078]

[0079] Among them, Q, K, and V are the query, key, and value matrices respectively, and d k is the dimension of the key, indicating that the model only focuses on the vocabulary information in the input sequence that is most relevant to the deployment task during encoding or decoding, so as to efficiently perform semantic understanding and intention recognition of the deployment task.

[0080] After completing word embedding and Transformer encoding, the system will use a classification model to further identify the intention of the deployment instruction submitted by the user. This model has learned a large number of features of the deployment scenario through deep learning technology and can convert the deployment instruction into a probability distribution with a clear intention, which can be specifically expressed by the following formula:

[0081] o = Softmax(W i H + b i )

[0082] Among them, W i and b i are model parameters, and o is the probability distribution of the intention. The probability distribution of the intention reflects the most likely intention direction of the control instruction. Through this step, the system can ensure that the execution of the control task conforms to the original intention of the user, improving the accuracy and effectiveness of the control.

[0083] When the calculated second matching degree is not less than the preset control threshold, it indicates that the monitored image matches the control task. At this time, processing measures need to be taken, and the security management strategy corresponding to the preset task object disposal type is executed. Common security management strategies include the following: tracking the task object in the monitored image, generating early warning prompt information, and automatically dialing the alarm phone.

[0084] Through the above steps, the pre-trained computer vision large model is used to expand the initial control task parameters, obtaining more complete semantic control task parameters. Using the expanded control task parameters as the target control task parameters helps generate a control task that better fits the control intention of the target object, laying a foundation for the intelligent execution and matching of subsequent control tasks, and thus solving the technical problem that related model algorithms in the security control scenario can only handle specific vision tasks and have low intelligence.

[0085] Embodiment 2

[0086] According to the embodiment of the present application, there is also provided a security control management device for implementing the security control management method in Embodiment 1, as Figure 4 shown. The security control management device at least includes: an acquisition module 41, a prompt module 42, and a generation module 43, where:

[0087] The acquisition module 41 is used to acquire the initial control task parameters input by the target object in the interaction interface. Among them, the initial control task parameters at least include the control scenario.

[0088] The prompt module 42 is used to determine the extended prompt information that matches the initial control task parameters by using the pre-trained computer vision large model, and display the extended prompt information in the interaction interface.

[0089] The generation module 43 is used to acquire the target control task parameters obtained by expanding the initial control task parameters according to the extended prompt information input by the target object in the interaction interface, and generate a control task based on the target control task parameters.

[0090] The analysis module 44 is used to, during the execution of the control task, acquire the monitored image in real time, and comprehensively analyze the monitored image and the target control task parameters by using the computer vision large model to obtain the control task matching result corresponding to the monitored image.

[0091] The functions of each module of the security deployment management device will be described below in combination with specific implementation processes.

[0092] Optionally, the acquisition module can obtain the initial deployment task parameters input by the target object in the interaction interface in the following ways: receive the initial deployment task parameters in text mode input by the target object in the interaction interface; or, receive the initial deployment task parameters in voice mode input by the target object in the interaction interface, and use speech recognition technology to convert the initial deployment task parameters into text mode; where the initial deployment task parameters include at least one of the following: task name, task object handling type, deployment time limit, deployment time period, deployment area, deployment threshold, deployment scenario.

[0093] Optionally, the prompt module can determine the extended prompt information matching the initial deployment task parameters by the following methods: use the pre-trained computer vision large model to respectively determine the first matching degrees of each historical task in the historical task database with the initial deployment task parameters, and sort the obtained multiple first matching degrees from large to small; if there are first matching degrees greater than the preset matching threshold, determine the preset number of target first matching degrees with the highest rankings from the first matching degrees greater than the preset matching threshold, and obtain the historical task data of the historical tasks corresponding to the target first matching degrees as the extended prompt information, where the historical task data includes at least one of the following: historical task description information, historical monitoring video clips, historical monitoring images; if there are no first matching degrees greater than the preset matching threshold, use the computer vision large model to generate extended prompt information matching the initial deployment task parameters, where the types of the extended prompt information include at least one of the following: task images corresponding to the initial deployment task parameters, text descriptions of the deployment task parameters to be supplemented.

[0094] Optionally, when the prompt module determines the first matching degrees of each historical task with the initial deployment task parameters, it can be implemented through the following steps: use the computer vision large model to perform semantic analysis on the initial deployment task parameters to determine multiple dimensions of first subtasks corresponding to the initial deployment task parameters and the association relationships between the first subtasks; determine the dependency graph corresponding to the multiple first subtasks according to the association relationships, and determine the weight coefficients corresponding to the first subtasks of each dimension according to the dependency graph; use the computer vision large model to analyze each historical task to determine multiple dimensions of second subtasks corresponding to each historical task; for each historical task, determine the similarity between the second subtasks and the first subtasks of the same dimension corresponding to the historical task and the initial deployment task parameters, and perform weighted calculation on the obtained similarities according to the weight coefficients to obtain the first matching degree of the historical task with the initial deployment task parameters.

[0095] Optionally, the hint module can be implemented by the following formula when calculating the first matching degree between the historical task and the initial deployment task parameters:

[0096]

[0097] In the formula, S(T1,T2) represents the first matching degree between the initial deployment task parameter T1 and the historical task T2; n represents the total number of dimensions of the first subtask corresponding to the initial deployment task parameter and the second subtask corresponding to the historical task; sim(f 1i ,f 2i ) represents the similarity between the first subtask f 1i of the i-th dimension and the second subtask f 2i of the i-th dimension. If there is no first subtask of the i-th dimension or no second subtask of the i-th dimension, sim(f 1i ,f 2i ) = 0; w i represents the weight coefficient of the first subtask of the i-th dimension. If there is no first subtask of the i-th dimension, w i = 0; d i is a dependence factor, indicating whether there is a second subtask with the same upstream dependence task dimension as the first subtask of the i-th dimension. If so, d i = 1. If not, d i = 0.

[0098] Optionally, the generation module can be implemented by the following steps when obtaining the target deployment task parameters: Loop through the following process: Obtain the deployment task parameters expanded according to the extended hint information input by the target object in the interaction interface, and detect whether the target object inputs a confirmation instruction; if it is detected that the target object inputs a confirmation instruction, terminate the loop process and use the expanded deployment task parameters as the target deployment task parameters; if it is not detected that the target object inputs a confirmation instruction, use the computer vision large model to determine new extended hint information that matches the expanded deployment task parameters, and display the new extended hint information in the interaction interface.

[0099] Optionally, the analysis module can obtain the monitoring images in real time when performing the deployment task in the following way: During the preset deployment period, obtain multiple monitoring videos collected by multiple cameras in the preset deployment area in real time; extract the current frame image of each monitoring video as the monitoring image.

[0100] Optionally, when the analysis module obtains the matching result of the control task corresponding to the monitoring image by using the computer vision large model, it can be done in the following ways: extracting the semantic features corresponding to the target control task parameters by using the computer vision large model, where the semantic features are at least used to reflect the control intention of the target object; extracting the image features corresponding to the monitoring image by using the computer vision large model; determining the second matching degree between the semantic features and the image features; when the second matching degree is not less than the preset control threshold, determining that the monitoring image needs to be processed for matching with the control task; when the second matching degree is less than the control threshold, determining that the monitoring image does not match the control task and does not need to be processed.

[0101] Optionally, the above security control management device further includes an execution module. When the analysis module determines that the monitoring image needs to be processed for matching with the control task, the execution module is used to execute the security management strategy corresponding to the preset task object disposal type, and the security management strategy includes at least one of the following: tracking the task object in the monitoring image, generating a warning prompt message, and automatically dialing the alarm phone.

[0102] It should be noted that each module in the security control management device in the embodiments of the present application corresponds one by one to each implementation step of the security control management method in Embodiment 1. Since the description in Embodiment 1 is already detailed, some details not shown in this embodiment can be referred to Embodiment 1 and will not be elaborated here.

[0103] Embodiment 3

[0104] According to the embodiments of the present application, there is also provided a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the security control management method in Embodiment 1.

[0105] According to the embodiments of the present application, there is also provided a non-volatile storage medium, which includes a stored computer program. The device where the non-volatile storage medium is located executes the security control management method in Embodiment 1 by running the computer program.

[0106] According to the embodiments of the present application, there is also provided a processor, which is used to run a computer program. When the computer program runs, it executes the security control management method in Embodiment 1.

[0107] According to the embodiments of the present application, there is also provided an electronic device, which includes: a memory and a processor. The memory stores a computer program, and the processor is configured to execute the security control management method in Embodiment 1 through the computer program.

[0108] Specifically, when the computer program runs, it executes the following steps:

[0109] Obtain the initial deployment task parameters input by the target object in the interaction interface, where the initial deployment task parameters at least include the deployment scenario; use the pre-trained large computer vision model to determine the extended prompt information that matches the initial deployment task parameters, and display the extended prompt information in the interaction interface; obtain the target deployment task parameters input by the target object in the interaction interface after expanding the initial deployment task parameters based on the extended prompt information, and generate a deployment task based on the target deployment task parameters; during the execution of the deployment task, obtain the monitoring image in real time, and comprehensively analyze the monitoring image and the target deployment task parameters using the computer vision model to obtain the deployment task matching result corresponding to the monitoring image.

[0110] As an alternative implementation, the above electronic device may exist in the form of a mobile terminal, a computer terminal, or a similar computing device. Figure 5 Shows a hardware structure block diagram of an electronic device for implementing the security deployment management method. As Figure 5 shown, the electronic device 50 may include one or more (shown as 502a, 502b,..., 502n in the figure) processors 502 (the processor 502 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 504 for storing data, and a transmission device 506 for communication functions. In addition, it may further include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply, and / or a camera. Those of ordinary skill in the art can understand that Figure 5 the structure shown is only illustrative and does not limit the structure of the above electronic device. For example, the electronic device 50 may further include more or fewer components than those Figure 5 shown, or have a different configuration from that Figure 5 shown.

[0111] It should be noted that the above one or more processors 502 and / or other data processing circuits are generally referred to as "data processing circuits" in this article. The data processing circuit may be embodied in whole or in part as software, hardware, firmware, or any other combination. In addition, the data processing circuit may be a single independent processing module, or be fully or partially incorporated into any one of the other elements in the electronic device 50. As involved in the embodiments of the present application, the data processing circuit is a processor control (such as the selection of a variable resistance terminal path connected to an interface).

[0112] The memory 504 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the security layout management method in the embodiments of the present application. The processor 502 executes various functional applications and data processing by running the software programs and modules stored in the memory 504, that is, implements the vulnerability detection method of the above application program. The memory 504 may include a high-speed random access memory, and may further include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory 504 may further include a memory remotely disposed relative to the processor 502, and these remote memories can be connected to the electronic device 50 through a network. Examples of the above network include but are not limited to the Internet, enterprise intranet, local area network, mobile communication network, and combinations thereof.

[0113] The transmission device 506 is used to receive or send data via a network. Specific examples of the above network may include the wireless network provided by the communication provider of the electronic device 50. In one instance, the transmission device 506 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station and thus communicate with the Internet. In one instance, the transmission device 506 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0114] The display can be, for example, a touch-screen liquid crystal display (LCD), which enables the user to interact with the user interface of the electronic device 50.

[0115] The above serial numbers of the embodiments are only for description and do not represent the advantages or disadvantages of the embodiments.

[0116] In the above embodiments of the present application, the descriptions of each embodiment have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0117] In several embodiments provided in the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only illustrative. For example, the division of units can be a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of units or modules can be in electrical or other forms.

[0118] The unit described as a separating component may or may not be physically separated, and the component shown as a unit may or may not be a physical unit, that is, it may be located in one place or distributed to multiple units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0119] In addition, each functional unit in various embodiments of the present application may be integrated in a processing unit, may exist separately as individual physical units, or two or more units may be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0120] If the integrated unit is implemented in the form of 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 the present 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 for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in various embodiments of the present application. The foregoing storage medium includes: USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs, etc., which can store program codes.

[0121] The above is only the preferred implementation manner of the present application. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A security deployment management method, characterized in that, include: Acquire initial control task parameters input by the target object in the interactive interface, wherein the initial control task parameters at least include a control scene; Determine extended prompt information matching the initial control task parameters using a pre-trained computer vision large model, and display the extended prompt information in the interactive interface; Acquire the target control task parameters input by the target object in the interactive interface after the initial control task parameters are expanded according to the extended prompt information, and generate a control task according to the target control task parameters; In the process of executing the control task, a monitoring image is acquired in real time, and the monitoring image and the target control task parameters are comprehensively analyzed using the computer vision large model to obtain a control task matching result corresponding to the monitoring image.

2. The method according to claim 1, characterized in that, Get the initial control task parameters entered by the target object in the interactive interface, including: receiving initial control task parameters in text mode input by the target object in the interactive interface; or, Receiving initial control task parameters in voice mode input by the target object in the interactive interface, and converting the initial control task parameters into text mode using voice recognition technology; The initial control task parameters include at least one of the following: task name, task object handling type, control time limit, control period, control area, control threshold, and control scenario.

3. The method according to claim 1, wherein The extended prompt information matching the initial control task parameters is determined by using a pre-trained computer vision large model, including: Determine the first matching degree between each historical task in the historical task database and the initial control task parameter by using the computer vision large model, and sort the obtained multiple first matching degrees from large to small; If there is a first matching degree greater than a preset matching threshold, determine a preset number of target first matching degrees ranked top from the first matching degrees greater than the preset matching threshold, and obtain historical task data of historical tasks corresponding to the target first matching degrees as the extended prompt information, wherein the historical task data includes at least one of the following: historical task description information, historical monitoring video clips, and historical monitoring images; If there is no first match degree greater than a preset matching threshold, the computer vision large model is used to generate extended prompt information that matches the initial control task parameters, wherein the type of the extended prompt information includes at least one of the following: a task image corresponding to the initial control task parameters, and a text description of the control task parameters to be supplemented.

4. The method according to claim 3, wherein Determining the first matching degree between each historical task in the historical task database and the initial control task parameter by using the computer vision large model respectively includes: Using the computer vision big model to perform semantic analysis on the initial control task parameters, determine the first subtasks of multiple dimensions corresponding to the initial control task parameters and the correlation between the first subtasks; Determine a dependency graph corresponding to a plurality of the first subtasks according to the association relationship, and determine a weight coefficient corresponding to the first subtasks of each dimension according to the dependency graph; Analyze each historical task using the computer vision large model to determine multiple dimensions of second subtasks corresponding to each historical task; For each historical task, determine the similarity between the second subtasks and the first subtasks in the same dimension corresponding to the historical task and the initial deployment task parameters, and perform weighted calculation on the obtained similarities according to the weight coefficient to obtain the first matching degree between the historical task and the initial deployment task parameters.

5. The method according to claim 4, wherein Performing weighted calculation on the obtained similarities according to the weight coefficient to obtain the first matching degree between the historical task and the initial deployment task parameters includes: Calculate the first matching degree between the historical task and the initial deployment task parameters using the following formula: Wherein, S(T1, T2) represents the first matching degree between the initial deployment task parameter T1 and the historical task T2; n represents the total dimension number of the first subtask corresponding to the initial deployment task parameter and the second subtask corresponding to the historical task; sim(f 1i , f 2i ) represents the similarity between the first subtask f 1i of the i-th dimension and the second subtask f 2i of the i-th dimension. If there is no first subtask of the i-th dimension or no second subtask of the i-th dimension, sim(f 1i , f 2i ) = 0; w i represents the weight coefficient of the first subtask of the i-th dimension. If there is no first subtask of the i-th dimension, w i = 0; d i is a dependency factor, indicating whether there is a second subtask with the same upstream dependency task dimension as the first subtask of the i-th dimension. If so, d i = 1. If not, d i = 0.

6. The method according to claim 1, wherein Obtain the target deployment task parameters obtained by the target object in the interaction interface by expanding the initial deployment task parameters according to the extended prompt information, including: Loop through the following process: Obtain the deployment task parameters expanded by the target object in the interaction interface according to the extended prompt information, and detect whether the target object inputs a confirmation instruction; If it is detected that the target object inputs a confirmation instruction, terminate the loop process, and use the expanded deployment task parameters as the target deployment task parameters; If it is not detected that the target object inputs a confirmation instruction, use the computer vision large model to determine new extended prompt information that matches the expanded deployment task parameters, and display the new extended prompt information in the interaction interface.

7. The method according to claim 1, characterized in that Obtain monitoring images in real time, including: During a preset deployment period, obtain multiple monitoring videos collected by multiple cameras in a preset deployment area in real time; Extract the current frame image of each monitoring video as the monitoring image.

8. The method according to claim 1, wherein Comprehensively analyze the monitoring image and the target deployment task parameters using the computer vision large model to obtain the deployment task matching result corresponding to the monitoring image, including: Use the computer vision large model to extract the semantic features corresponding to the target deployment task parameters, where the semantic features are at least used to reflect the deployment intention of the target object; Use the computer vision large model to extract the image features corresponding to the monitoring image; Determine the second matching degree between the semantic features and the image features; When the second matching degree is not less than a preset deployment threshold, determine that the monitoring image needs to be processed for matching with the deployment task; When the second matching degree is less than the deployment threshold, determine that the monitoring image does not match the deployment task and does not need to be processed.

9. The method according to claim 8, characterized in that, The method further includes: When it is determined that the monitoring image needs to be processed for matching with the deployment task, execute a security management strategy corresponding to a preset task object handling type, where the security management strategy includes at least one of the following: tracking the task object in the monitoring image, generating a warning prompt message, and automatically dialing an alarm call.

10. A security deployment management device, characterized in that, including: An acquisition module, configured to acquire initial deployment task parameters input by a target object in an interaction interface, where at least a deployment scenario is included in the initial deployment task parameters; A prompting module, configured to use a pre-trained large computer vision model to determine extended prompting information that matches the initial deployment task parameters, and display the extended prompting information in the interaction interface; A generating module, configured to obtain target deployment task parameters obtained by extending the initial deployment task parameters according to the extended prompting information input by the target object in the interaction interface, and generate a deployment task according to the target deployment task parameters; An analyzing module, configured to, during the execution of the deployment task, obtain monitoring images in real time, and comprehensively analyze the monitoring images and the target deployment task parameters by using the computer vision model to obtain a deployment task matching result corresponding to the monitoring images.

11. A computer program product, characterized in that, Comprising: A computer program, wherein when the computer program is executed by a processor, the security deployment management method according to any one of claims 1 to 9 is implemented.

12. An electronic device, characterized in that, Comprising: A memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to execute the security deployment management method according to any one of claims 1 to 9 through the computer program.

Images