Security system, interaction method thereof and electronic device

Abstract

This application discloses a security system, its interaction method, and an electronic device. The method includes acquiring image data of a target object and determining the target object's identity information based on the image data; acquiring voice data of the target object and parsing the target object's needs information based on the voice data, wherein at least a portion of the voice data parsing is performed by a cloud device; determining a target template from multiple preset templates based on the identity information and the needs information; and executing a corresponding interactive operation according to the target template, the interactive operation including outputting a corresponding voice response and / or driving at least one terminal actuator to perform a corresponding behavioral action. This application solves the technical problem in related technologies that it is impossible to dynamically adjust the interaction strategy according to the visitor's identity, making it difficult to balance differentiated emotional interaction and security control. It achieves the technical effect of adaptively outputting voice responses and driving the terminal to perform corresponding behavioral actions according to identity and needs, realizing differentiated human-computer emotional interaction and security control.

Description

Technical Field

[0001] This application relates to the field of security, and in particular to a security system, its interaction method, and electronic devices. Background Technology

[0002] Currently, intelligent security systems are widely used in smart homes, service robots, security monitoring, and smart offices. However, most existing intelligent security systems are based on fixed script responses or simple voice command recognition, and cannot dynamically adjust interaction strategies according to the visitor's identity, thus making it difficult to achieve differentiated human-computer emotional interaction and security control.

[0003] Therefore, how to provide a solution to the above-mentioned technical problems is a problem that needs to be solved by those skilled in the art. Summary of the Invention

[0004] The purpose of this application is to provide a security system, its interaction method, and electronic device, so as to at least solve the technical problems in the related technologies that are unable to dynamically adjust the interaction strategy according to the visitor's identity and are difficult to balance differentiated emotional interaction and security control.

[0005] This application provides an interaction method for a security system, comprising: acquiring image data of a target object, and determining the identity information of the target object based on the image data; acquiring voice data of the target object, and parsing the target object's demand information based on the voice data, wherein the parsing of at least a portion of the voice data is performed by a cloud device; determining a target template from multiple preset templates based on the identity information and the demand information; and performing a corresponding interaction operation according to the target template, wherein the interaction operation includes outputting a corresponding voice response and / or driving at least one terminal actuator to perform a corresponding behavioral action.

[0006] This application also provides a security system, including an image acquisition unit, a voice acquisition unit, a memory, a processor, a voice output unit, and at least one terminal actuator; the image acquisition unit includes an imaging sensor and an optical lens, which are positioned facing the area to be acquired to generate image data; the voice acquisition unit is used to acquire voice data from a target object; the memory is used to store multiple preset templates; the processor is used to acquire the identity information of the target object determined by analyzing the image data, acquire the demand information of the target object obtained by parsing the voice data, and determine the target template from the multiple preset templates based on the identity information and demand information; the voice output unit is used to output a corresponding voice response according to the target template; and at least one terminal actuator is used to perform a corresponding action according to the target template.

[0007] This application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the interaction method of the security system as described above.

[0008] This application utilizes both identity information obtained from image data and demand information obtained from voice data as decision-making bases. Furthermore, the analysis of emotions and semantics is performed in the cloud, unrestricted by terminal computing power. This ensures real-time terminal response while improving the accuracy of demand recognition. The combination of these two data points is used to match different pre-configured templates, enabling the security system to automatically provide differentiated responses to visitors with different identities and needs, rather than relying on a uniform fixed script or single voice command recognition. Therefore, it solves the technical problems in related technologies that cannot dynamically adjust interaction strategies based on visitor identity and struggle to balance differentiated emotional interaction and security control. It achieves the technical effect of adaptively outputting voice responses based on identity and needs, driving the terminal to execute corresponding actions, thereby realizing differentiated human-computer emotional interaction and security control. Attached Figure Description

[0009] To more clearly illustrate the embodiments of this application, the accompanying drawings used in 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.

[0010] Figure 1 This is a schematic diagram of the structure of the first security system provided in the embodiments of this application; Figure 2 A flowchart illustrating the steps of an interaction method for a security system provided in this application embodiment; Figure 3 This application provides a flowchart of steps for determining the identity information of a target object in an embodiment of the present application. Figure 4 A flowchart illustrating the steps of another interaction method for a security system provided in an embodiment of this application; Figure 5 A flowchart illustrating the steps for determining a target template provided in this application embodiment; Figure 6 This is a schematic diagram of the structure of the second security system provided in the embodiments of this application; Figure 7 This is a schematic diagram of the structure of a third security system provided in an embodiment of this application.

[0011] Terminal equipment - 110, cloud equipment - 120; Image acquisition unit-211, voice acquisition unit-212, memory-213, processor-214, voice output unit-215, terminal actuator-216, terminal image processor-217, terminal image intensifier-218, optical character recognition unit-219, cloud voice processor-220. Terminal image acquisition device-2111, terminal voice acquisition device-2121. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, 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 some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0013] This application provides an interaction method for a security system. This method can be applied to electronic devices with image acquisition and voice interaction capabilities, such as smart doorbells, smart access control systems, service robots, smart speakers, and smart display terminals. It can also be applied to security systems composed of these devices and cloud-based devices. To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0014] Reference Figure 1 As shown, Figure 1 This is a schematic diagram of a security system provided in an embodiment of this application. The security system includes a terminal device 110 and a cloud device 120. The interaction method of the security system provided in this application can be jointly implemented by the terminal device 110 and the cloud device 120.

[0015] Terminal device 110 refers to a device deployed at the interaction site that directly collects information and outputs responses to target objects, including but not limited to smart doorbells, smart access control systems, integrated access control machines, service robots, smart speakers, or smart display terminals. Terminal device 110 has image acquisition and voice interaction capabilities, used to acquire images and voice data from target objects entering the acquisition range, and output voice responses and drive corresponding actions to the area where the target object is located. Cloud device 120 refers to a device that communicates with terminal device 110 via a network and provides computing support to terminal device 110, including but not limited to servers, server clusters consisting of multiple servers, or cloud platforms. Cloud device 120 can be used to undertake computationally intensive processing tasks, such as parsing the emotional tendencies and semantics in voice data.

[0016] Terminal device 110 and cloud device 120 establish a communication connection via a network, including but not limited to wireless local area networks, mobile communication networks, or wired networks. As an optional embodiment, terminal device 110 uploads the collected voice data to cloud device 120 via this network. Cloud device 120 parses the voice data and returns the parsed results to terminal device 110 via the same network. Based on this, terminal device 110 and cloud device 120 cooperate by dividing tasks. Processes with high real-time requirements, such as image acquisition, voice acquisition, and response output, are performed locally on terminal device 110, while computationally intensive semantic and sentiment analysis is handled by cloud device 120. This ensures real-time response from the terminal while reducing resource consumption on terminal device 110 and improving parsing accuracy by leveraging the computing power of the cloud.

[0017] Reference Figure 2 As shown, the interaction methods of this security system include: S101: Obtain image data of the target object and determine the identity information of the target object based on the image data; In this embodiment, the target object refers to the object that enters the image acquisition range and needs to interact with the device, including but not limited to visitors, such as a person standing in front of the door.

[0018] Specifically, acquiring image data of a target object refers to imaging the target object entering its field of view using a camera module, obtaining one or more frames containing the target object. In some embodiments, the image data is a single-frame static image, such as when the target object approaches and triggers the doorbell, the camera on the doorbell captures a frame containing a close-up of the target object's face. In other embodiments, the image data is a sequence of continuously acquired multi-frame images, such as after the target object appears in the field of view, the camera continuously captures image frames containing the target object at a preset frame rate, forming a sequence reflecting its movement process; the multi-frame image sequence can not only be used to determine identity information, but also, if the target object is subsequently identified as an unauthorized person, analyze its movement trajectory to determine whether it has engaged in suspicious behavior such as loitering or probing, serving as a basis for triggering an alarm.

[0019] Determining the identity of a target object based on image data refers to analyzing and processing image data to obtain information that characterizes who the target object is or what category it belongs to. In some embodiments, the identity information is the identification result of the individual target object, such as identifying the target object as A or B by comparing its facial features with a pre-stored resident facial feature database. In other embodiments, the identity information is the classification result of the category to which the target object belongs, such as classifying it into one of the preset categories such as authorized resident, visitor, courier, or stranger by analyzing its facial features, body shape features, and behavioral features. In still other embodiments, the identity information includes both individual identification results and category classification results, such as identifying the target object as a resident's family member A and classifying it as a visitor. Based on this, a preliminary judgment of the visitor can be obtained before the terminal device engages in voice interaction with the target object. Within a limited time window, targeted interaction strategies can be adopted for visitors with different identities, such as issuing a package pickup reminder to a pre-registered courier and issuing a warning if a stranger lingers for a long time. This improves the flexibility of responding to different visit scenarios by advancing the timing of differentiated responses in the process.

[0020] S102: Acquire the voice data of the target object, and parse the target object's requirement information based on the voice data, wherein at least part of the voice data parsing is performed by a cloud device; Specifically, the system collects the sound signals emitted by the target object through a sound pickup device to obtain speech data. This speech data is then recognized and understood to obtain information that represents what the target object wants to do. For example, the parsed information about the target object's needs includes, but is not limited to, the target object wanting to open a door, have a package delivered, or be looking for someone. At least a portion of the speech data parsing is performed by a cloud device. In some embodiments, the transcription, semantic understanding, and sentiment analysis of the speech data are all completed by the cloud device, with the terminal device only responsible for collection and uploading. In other embodiments, the terminal device first performs voice wake-up and preliminary voice activity detection, and then uploads the valid speech segments to the cloud device for semantic parsing. In still other embodiments, the terminal device performs speech-to-text transcription, while the more computationally intensive semantic understanding and sentiment analysis are handled by the cloud device.

[0021] This embodiment places at least part of the parsing in the cloud. On the one hand, this eliminates the need for terminal devices to store and run large semantic models locally, thereby reducing the flash memory usage and computing power consumption of terminal devices. On the other hand, cloud devices can deploy models with better performance and can be replaced and upgraded as needed, thereby improving the accuracy of demand identification while ensuring real-time terminal response.

[0022] S103: Based on identity information and demand information, determine the target template from multiple preset templates; In this embodiment, identity information and demand information are used together as the decision-making basis. One template matching the current situation is selected from multiple pre-configured templates as the target template. The preset templates are pre-configured data sets that specify how the device should respond to the target object. As an optional embodiment, multiple preset templates are stored in the form of a mapping table, with different combinations of identity information and demand information pointing to different templates. Alternatively, multiple preset templates are stored in the form of a rule set, and the target template is selected by matching the conditions of the rules. Because both identity information obtained from image data and demand information obtained from voice data are used as the decision-making basis, and their combination is used to match different pre-configured templates, the security system can automatically provide differentiated responses to visitors with different identities and demands, no longer relying on a uniform fixed script or a single voice command recognition.

[0023] S104: Execute the corresponding interactive operation according to the target template. The interactive operation includes outputting the corresponding voice response and / or driving at least one terminal actuator to perform the corresponding action.

[0024] In some embodiments, the interactive operation only includes outputting a voice response, such as playing a greeting or prompt through a speaker. In other embodiments, the interactive operation only includes driving a terminal actuator to perform a behavioral action, such as controlling a door lock to open or adjusting lighting. In still other embodiments, the interactive operation includes both outputting a voice response and driving a terminal actuator to perform a behavioral action, such as opening a door lock while playing a reception script. A terminal actuator refers to an actuator capable of performing actions at the physical or display level. Based on this, the security system can adaptively output a voice response and drive the terminal to perform corresponding behavioral actions according to the visitor's identity and needs, thereby achieving differentiated human-computer emotional interaction and security control. This solves the problems in related technologies where it is impossible to dynamically adjust the interaction strategy according to the visitor's identity and it is difficult to balance differentiated emotional interaction and security control.

[0025] In one exemplary embodiment, the identity information includes a target identity category; Acquire image data of the target object, and determine the target object's identity information based on the image data, including: Image data is obtained by acquiring images of the target object through a terminal device; Based on the image recognition model, the facial features and morphological features of the target object in the image data are analyzed; Utilize facial features and / or morphological features to determine the target identity category of the target object.

[0026] In this embodiment, the identity information includes the target identity category, which is the classification result of the category to which the target object belongs. Compared with only providing the individual identification result, using the target identity category as the identity information enables the subsequent template selection to be organized based on the category, which facilitates the configuration of a unified response strategy for the same category.

[0027] like Figure 3 As shown, acquiring image data of a target object through a terminal device refers to the imaging process performed by the locally deployed terminal device using its image acquisition module. In some embodiments, the terminal device is a smart doorbell, and its image acquisition module is a camera mounted on the doorbell panel. In other embodiments, the terminal device is a stand-alone camera device or an integrated access control system. By performing image acquisition locally on the terminal device, preliminary identity verification can be completed before the target object approaches and before voice interaction begins.

[0028] Specifically, based on image recognition models, the facial and morphological features of target objects in image data are analyzed. Facial features refer to information that characterizes the face of a target object, including but not limited to facial key point features and facial feature vectors extracted by deep neural networks. Morphological features refer to information that characterizes the appearance and shape of a target object, including but not limited to body contour features, clothing appearance features, and gait features. Image recognition models can employ structures such as convolutional neural networks and visual Transformers to detect target objects and extract the above features from image data. In some embodiments, the image recognition model first performs object detection to identify target types such as people, vehicles, and pets in the image data, and then performs feature extraction on targets belonging to the category of people.

[0029] In some embodiments, the target identity category can be determined using only facial features; in other embodiments, it can be determined using only morphological features; and in still other embodiments, both facial and morphological features can be used simultaneously, for example, using facial features as the primary feature and morphological features as a secondary feature to jointly determine the target identity category. Based on this, even if one type of feature of the target object is incomplete, another type of feature can still be used to determine the identity category, thereby improving the reliability of identity determination and providing a basis for subsequent template selection by category.

[0030] In one exemplary embodiment, determining the target identity category of a target object using facial features and / or morphological features includes: Using facial features and / or morphological features, determine the target identity category that matches the target object from multiple preset identity categories.

[0031] Specifically, multiple preset identity categories are pre-defined sets of categories used to categorize visitors. The process of determining the target identity category involves finding the category that best matches the target object within this set. In some embodiments, this is achieved through comparison, where the extracted facial features and / or morphological features are compared one by one with the reference features corresponding to each preset identity category, and the category corresponding to the highest degree of matching is taken as the target identity category. In other embodiments, this is achieved through classification, where the extracted features are input into a classifier, and the classifier directly outputs the preset identity category to which the target object belongs. Based on this, continuous feature representations are mapped to discrete identity categories, thereby enabling the organization of subsequent response strategies in units of a finite and clear set of categories, avoiding the redundancy caused by configuring strategies individually for each individual.

[0032] In one exemplary embodiment, the multiple preset identity categories include authorized identity categories and unauthorized identity categories; wherein, the authorized identity categories include at least two of the following: owner, family member, acquaintance, courier, and visitor; wherein, the unauthorized identity categories include user-defined dangerous persons.

[0033] In this embodiment, the preset identity categories are divided into two main categories: authorized identity categories and unauthorized identity categories. Authorized identity categories refer to those recognized by the security system and subject to relatively lenient interaction policies, while unauthorized identity categories refer to those not recognized by the security system and subject to relatively cautious or vigilant interaction policies.

[0034] In some embodiments, the authorized identity categories include two types: owner and family member. In other embodiments, the authorized identity categories include five types: owner, family member, acquaintance, courier, and visitor. In still other embodiments, the authorized identity categories are any two, three, or four combinations of the above five types. It is understood that categories such as courier and visitor do not refer to a fixed individual, but rather to a group of people with corresponding attributes. For example, an individual initially identified as a stranger can be labeled as a courier or visitor by an authorized user, thus falling into the corresponding authorized identity category.

[0035] Specifically, the unauthorized identity category includes user-defined dangerous individuals. In some embodiments, authorized users can register specific individuals as dangerous individuals, and handle them according to the unauthorized identity category when the individual is identified again; in other embodiments, any object that cannot be classified into any authorized identity category is classified into the unauthorized identity category. Based on this, by distinguishing between authorized and unauthorized identity categories, differentiated interaction and control strategies can be applied to visitors of different risk levels, facilitating authorized personnel while maintaining necessary vigilance towards unauthorized personnel.

[0036] In one exemplary embodiment, determining a target identity category matching the target object from multiple preset identity categories using facial features and / or morphological features includes: The facial features are matched with each reference facial feature in the first feature library, which includes the correspondence between authorized identity categories and reference facial features. The target identity category of the target object is determined based on the matching results.

[0037] Specifically, the first feature library is used to store the correspondence between authorized identity categories and reference facial features. In some embodiments, the first feature library is cached locally on the terminal device, wherein the stored reference facial features are facial key point features; in other embodiments, the reference facial features stored in the first feature library are facial feature vectors extracted by a neural network. In the first feature library, each registered authorized object corresponds to a set of reference facial features and is labeled with its authorized identity category, for example, one set of reference facial features is labeled as owner and another set is labeled as family member.

[0038] Specifically, such as Figure 3 As shown, the facial features of the target object extracted from the image data are compared one by one with the reference facial features stored in the first feature library to obtain the matching result. In this embodiment, the matching result is represented by an identity confidence score. The identity confidence score is used to characterize the degree of matching between the facial features of the target object and a certain reference facial feature; the higher the score, the higher the degree of matching.

[0039] In some embodiments, the identity confidence score is obtained by calculating the similarity between the facial feature vector of the target object and the reference facial feature vector. For example, cosine similarity or normalized Euclidean distance is used as the identity confidence score. In other embodiments, the probability value of the target object belonging to each authorized identity category is directly output by the classification model as the identity confidence score.

[0040] Specifically, determining the target identity category of a target object based on the matching results means determining the identity category to which the target object belongs based on the identity confidence score. In this embodiment, the identity confidence score is compared with a pre-set threshold. When the identity confidence score corresponding to a certain reference facial feature is higher than the threshold, it is considered that the target object and the authorized object corresponding to the reference facial feature are the same person, and thus the authorized identity category marked by the reference facial feature is determined as the target identity category of the target object.

[0041] In some embodiments, when multiple identity confidence scores are all higher than a threshold, the authorized identity category corresponding to the identity with the highest confidence score is selected as the target identity category.

[0042] Multiple authorized identity categories can use a unified threshold, or different authorized identity categories can use different thresholds. For example, a higher threshold can be set for categories with higher security requirements to reduce the risk of false identification. The threshold in this embodiment can be determined through testing, selecting values ​​that meet product requirements in various scenarios. In some embodiments, the threshold value is also associated with the specific recognition algorithm used. In other embodiments, the threshold can be adjusted by authorized users, allowing them to choose between acceptable levels of leniency and strictness. Furthermore, when the identity confidence score corresponding to all reference facial features is not higher than the threshold, it is determined that the target object has not matched any authorized identity category. Based on this, determining the target identity category by comparing the identity confidence score with the threshold provides a quantifiable and reproducible basis for judgment, and allows for a flexible balance between recognition accuracy and tolerance through threshold setting.

[0043] In one exemplary embodiment, the interaction method of the security system further includes: When a facial feature fails to match a corresponding authorized identity category in the first feature library, the morphological feature is matched with each reference morphological feature in the second feature library, which includes the correspondence between authorized identity categories and reference morphological features. The target identity category of the target object is determined based on the matching results.

[0044] When facial features cannot be effectively extracted due to the target being wearing a mask, having their back to the camera, or being obscured, or when the confidence scores of the extracted facial features in the first feature database are all below the threshold, it is determined that the facial features failed to match the corresponding authorized identity category in the first feature database. In this case, morphological features are used for matching.

[0045] Specifically, the second feature library is used to store the correspondence between authorized identity categories and reference morphological features. In some embodiments, the second feature library is a pedestrian re-identification feature library, wherein the stored reference morphological features include at least one of body contour features, clothing appearance features, and gait features. Figure 3 As shown, matching morphological features with each reference morphological feature in the second feature library refers to comparing the morphological features of the target object with each reference morphological feature in the second feature library one by one based on the person re-identification algorithm to obtain a matching result. In some embodiments, the matching result is also represented by a morphological similarity score, and whether the match is successful is determined by comparing it with a corresponding threshold. The target identity category of the target object is determined based on the matching result, that is, the authorized identity category marked by the successfully matched reference morphological features is determined as the target identity category of the target object. Based on this, even when facial features are unavailable, morphological features can still be used to provide an identity category judgment, using person re-identification as a supplement to facial recognition, thereby maintaining the usability of identity determination in scenarios such as wearing masks or facing away from the camera, and expanding the scope of application of identity recognition.

[0046] In one exemplary embodiment, the interaction method of the security system further includes: When the target identity category of the target object is determined to be an unauthorized identity category, the control terminal device enters the automatic enhanced monitoring mode. The automatic enhanced monitoring mode includes at least one of the following: increasing the image acquisition frame rate, starting multi-angle camera tracking, and increasing the sampling frequency of environmental sensors.

[0047] Specifically, when the target object's identity category is determined to be unauthorized, it indicates that the target object is not recognized or requires vigilance. At this time, the control terminal device enters an automatic enhanced monitoring mode to obtain more comprehensive monitoring information about the target object. In some embodiments, the automatic enhanced monitoring mode increases the image acquisition frame rate, that is, it increases the terminal device's image acquisition frame rate from the normal frame rate to a higher frame rate to capture more continuous and detailed movements of the target object. In other embodiments, the automatic enhanced monitoring mode activates multi-angle camera tracking, that is, it uses multiple cameras arranged at different angles to track and capture images of the target object to obtain multi-view images. In still other embodiments, the automatic enhanced monitoring mode increases the sampling frequency of environmental sensors, that is, it increases the sampling frequency of environmental sensors such as temperature, infrared, and sound to more sensitively perceive the target object's activities within the monitoring area. In still other embodiments, the automatic enhanced monitoring mode includes two or three of the above. Based on this, for target objects with unauthorized identity categories, the monitoring intensity is automatically increased, thereby retaining more sufficient evidence information and improving security protection capabilities in situations where potential risks exist.

[0048] In one exemplary embodiment, the interaction method of the security system further includes: When the target identity category of the target object is determined to be the unauthorized identity category, the behavior trajectory of the target object is obtained, and the behavior trajectory is used to determine whether the target object has loitering behavior; An alarm is triggered when the target object exhibits loitering behavior.

[0049] In this embodiment, when the target object's identity category is determined to be an unauthorized identity category, the target object's behavioral trajectory is obtained. The behavioral trajectory refers to the target object's position and status information changing over time within the monitored area.

[0050] In some embodiments, the behavior trajectory is obtained by continuously tracking the position of the target object in a multi-frame image sequence. In other embodiments, the behavior trajectory is also described in conjunction with information such as dwell time.

[0051] In some embodiments, when the time a target object spends in the monitored area exceeds a preset time threshold, it is determined that it has loitering behavior. In other embodiments, when a target object moves back and forth in the monitored area and its movement path shows a repeated back-and-forth pattern, it is determined that it has loitering behavior. In still other embodiments, the loitering behavior is determined by combining the time spent in the monitoring area and the path pattern.

[0052] An alarm is triggered when a target object is observed loitering. In some embodiments, the alarm may be a push notification to an authorized user or a warning voice message played through the device's speaker to warn and deter strangers who linger at the door for an extended period. Based on this, by analyzing the behavioral patterns of unauthorized individuals and triggering an alarm when they loiter, timely warnings can be issued regarding potentially suspicious loitering behavior, further reducing the risk of unauthorized intrusion.

[0053] In an exemplary embodiment, before the terminal device acquires an image of the target object, the interaction method of the security system further includes: Based on the current ambient lighting conditions, at least one of the following methods is activated to enhance image quality: artificial intelligence image signal processor enhancement, infrared illumination, and high dynamic range imaging.

[0054] Specifically, such as Figure 3 As shown, before acquiring images of the target object through the terminal device, the current ambient lighting conditions are first obtained. In some embodiments, ambient illuminance is detected by a light sensor; in other embodiments, ambient lighting conditions are estimated by statistically analyzing the brightness of the pre-acquired images. Based on the current ambient lighting conditions, corresponding image quality enhancement processing is initiated. In some embodiments, image quality enhancement processing involves activating an artificial intelligence image signal processor (AI Image Signal Processor) to enhance the image signal, improving imaging quality in low-light, no-light, or backlit environments. In other embodiments, image quality enhancement processing involves activating infrared illumination, supplementing the acquisition area with infrared light. In still other embodiments, image quality enhancement processing involves activating high dynamic range imaging, improving imaging effects in high-contrast scenes such as backlighting by synthesizing images with different exposures. In yet another embodiment, two or three of the above methods are activated simultaneously. Based on this, image quality enhancement processing improves image quality under unfavorable lighting conditions such as low light, no light, and backlighting, thereby increasing the accuracy of target object identification in these environments and providing a more reliable image data foundation for subsequent identity determination.

[0055] In an exemplary embodiment, after determining the target identity category of the target object using facial features and / or morphological features, the interaction method of the security system further includes: Based on the target identity category, output guiding voice prompts; Acquire the target object's voice data, and parse the target object's requirements information based on the voice data, including: The system acquires voice data from the target audience based on guided voice prompts, and analyzes the emotional tendencies and semantics of the voice data through cloud devices to obtain demand information. The emotional tendencies analysis results include at least one of urgency, politeness, and threat; the semantics analysis results include at least one of door opening requests, package retrieval requests, and person finding requests.

[0056] Specifically, such as Figure 4 As shown, after determining the target object's identity category using facial features and / or morphological features, a guiding voice prompt is output based on the target identity category. The guiding voice prompt is used to guide the target object to provide voice feedback. In some embodiments, when the target identity category is an authorized identity category, a greeting-like guiding voice prompt is output; in other embodiments, when the target identity category is an unauthorized identity category, a neutral guiding voice prompt inquiring about the purpose of the visit is output. In this embodiment, the identity category is first determined by image recognition, and then the guiding voice prompt is output. This ensures that the image side first recognizes the presence of the target object, and then initiates the voice prompt accordingly, making the initiation of voice interaction a trigger condition based on the image recognition result.

[0057] Specifically, after hearing the guiding voice prompt, the target user responds, and their response is captured in real time by the microphone of the terminal device, resulting in feedback voice data. The cloud device then analyzes the sentiment and semantics of the voice data to obtain demand information. In some embodiments, the cloud device performs sentiment analysis based on the target user's word choice; in other embodiments, it performs sentiment analysis based on the target user's tone of voice; and in still other embodiments, it performs sentiment analysis based on both word choice and tone of voice. The specific features used depend on the dataset used during the training of the language model. The sentiment analysis results include at least one of eagerness, politeness, and threat.

[0058] In some embodiments, the cloud device extracts keywords and recognizes intent from the text obtained by speech-to-text feedback to obtain semantic keywords such as "open the door," "pick up the package," and "who to look for," and categorizes them into the corresponding request types. Based on this, emotional tendency and semantics are incorporated into the demand information, so that the demand information not only describes what the target object wants to do, but also describes the emotions they express, thereby providing a richer reference for subsequent template selection based on identity and needs.

[0059] In one exemplary embodiment, the interaction method of the security system further includes: Optical character recognition is performed on image data to extract express delivery waybill information or package information from the image data, and corresponding voice prompts are generated based on the recognition results.

[0060] Specifically, optical character recognition (OCR) of image data refers to text recognition of text regions appearing in image data. In some embodiments, when the target is a courier holding a package, OCR is performed on the waybill on the surface of the package in the image data to extract waybill information, such as the waybill number and recipient information. In other embodiments, text on the outer packaging of the package is recognized to extract package information. Based on the recognition results, corresponding voice prompts are generated, such as a prompt to the recipient to pick up the package or a voice confirming package information, which can be played through a speaker. Therefore, by utilizing OCR to extract waybill or package text from image data, package-related information can be automatically obtained in scenarios such as express delivery, and targeted voice prompts can be generated accordingly, further enriching the interactive content and scenario adaptability.

[0061] In one exemplary embodiment, the plurality of preset templates includes a serious warning template; Based on identity information and requirement information, a target template is determined from multiple preset templates, including: When the sentiment analysis result is threatening, the serious warning template is identified as the target template; According to the target template, perform the corresponding interactive operations, including: Following the serious warning template, control the door lock to lock and generate an alarm notification.

[0062] Specifically, among the multiple preset templates is a serious warning template. This template is used to issue a serious warning to the target in situations where a threat exists. When the emotional tendency analysis result obtained from feedback voice interpretation indicates a threat, the serious warning template is selected as the target template, and the door lock is controlled to lock and an alarm notification is generated according to this template. In some embodiments, the alarm notification is a notification message pushed to an authorized user or security platform; in other embodiments, the alarm notification is accompanied by the playback of an alarm sound. Based on this, when a threatening emotional tendency is detected, the door lock is automatically locked and an alarm is triggered, thereby enabling timely security control of threatening visitors.

[0063] Considering that the analysis results of sentiment tendencies may be inconsistent with identity information, in this embodiment, the determination of the target template is based on both identity information and demand information, and when the two are inconsistent, the target identity category in the identity information is given priority.

[0064] In some embodiments, when the target identity category obtained from image recognition is an authorized identity category such as a family member, while the emotional tendency obtained from feedback voice analysis is threatening, the target identity category obtained from image recognition is prioritized, and the target object is still determined to be an authorized identity category such as a family member. In this case, the serious warning template is not determined as the target template, but rather a template suitable for reassurance is determined, thereby outputting reassuring remarks. That is to say, when the emotional tendency analysis result is threatening, and the identity information does not point to an authorized identity category that conflicts with it, the serious warning template is determined as the target template. Based on this, by prioritizing image recognition when there is a clear contradiction between the image recognition result and the voice intent analysis, it avoids accidentally triggering serious warnings and door locks due to a family member's momentary words, while still being able to impose serious warnings and security controls on visitors who are indeed threatening, thus achieving a balance between the appropriateness of emotional interaction and the necessity of security control in the security system.

[0065] In one exemplary embodiment, determining a target template from multiple preset templates based on identity information and demand information includes: Based on the combination of identity information and demand information, a target template that matches the combination is selected from multiple preset templates.

[0066] Specifically, selecting a target template that matches a combination of identity information and requirement information from multiple preset templates means using both identity information and requirement information together as the basis for template selection, rather than relying on only one of them. In some embodiments, this is achieved through a mapping table, which records the target templates corresponding to various combinations of identity information and requirement information, and the target template matching the current combination is obtained by looking up the table. In other embodiments, the matching of combinations and templates is achieved through vectorization.

[0067] Specifically, such as Figure 5As shown, in embodiments where matching is achieved through vectorization, identity information and demand information are encoded and concatenated to construct an identity-intent joint vector. In some embodiments, the target identity category is encoded as an identity component, and the sentiment and semantics in the demand information are encoded as intent components. The identity components and intent components are then concatenated into an identity-intent joint vector. For each of the multiple preset templates, a corresponding reference vector is pre-configured. The similarity between the identity-intent joint vector and the reference vectors corresponding to each preset template is calculated. In some embodiments, cosine similarity is used, that is, the cosine value of the angle between two vectors represents the degree of similarity between them. The larger the cosine similarity, the closer they are. In other embodiments, the similarity is achieved using a normalized inner product or distance metric. The preset template corresponding to the one with the highest similarity is selected as the target template, that is, the target template that best matches the identity-intent joint vector is selected from multiple preset templates by taking the maximum similarity. Based on this, by constructing an identity-intent joint vector, calculating its similarity with each preset template reference vector, and taking the template corresponding to the one with the largest similarity, the combination of identity information and demand information can be uniformly measured and matched, so that a suitable target template can be stably selected under various combinations of identity and demand.

[0068] In one exemplary embodiment, selecting a target template that matches a combination of identity information and demand information from a plurality of preset templates includes: Obtain current time information and the geographical location information of the terminal device; Based on a combination of current time, geographic location, demand, and identity information, a matching target template is selected from multiple preset templates.

[0069] In some embodiments, the current time information includes the specific time, while in others it includes the date, day of the week, or time period (e.g., daytime, nighttime). Geographic location information represents the location of the terminal device. Based on a combination of current time, geographic location, demand information, and identity information, a matching target template is selected from multiple preset templates. That is, building upon the aforementioned identity and demand combination embodiments, time and geographic location are further introduced as components of the combination. As an optional embodiment, for the same identity and demand, different target templates can be selected at night and during the day. As another optional embodiment, for the same identity and demand, different target templates can be selected when the terminal device is in different geographic locations. Therefore, by incorporating current time and geographic location information into the template selection criteria, the determination of the target template can adapt to differences in time and location, further improving the adaptability of the response to specific situations.

[0070] In one exemplary embodiment, the preset template includes at least response tone parameters, semantic content library parameters, and behavior action parameters.

[0071] The preset template, as a data set specifying the device's response method, includes at least three types of parameters: response tone parameters, semantic content library parameters, and behavioral action parameters. The response tone parameters specify the tone used in the voice response; the semantic content library parameters specify the selected dialogue content; in some embodiments, the semantic content library parameters point to a semantic content library containing greetings, reception scripts, warnings, etc.; and the behavioral action parameters specify the actions to be performed by the terminal actuator. In some embodiments, the preset template only includes the above three types of parameters; in other embodiments, the preset template includes other parameters such as speech synthesis timbre and response speed in addition to the above three parameters. Based on this, by simultaneously specifying tone, content, and behavioral actions in the preset template, when performing interactive operations based on the target template, the tone expression, dialogue content, and terminal actions can be determined simultaneously, thereby outputting a customized response that coordinates content and behavior.

[0072] In one exemplary embodiment, the tone type defined by the response tone parameter includes at least one of gentle and friendly tone, formal and polite tone, serious warning tone and cold rejection tone, and different tone types correspond to different speech synthesis timbre and speech rate.

[0073] Specifically, the tone type defined by the response tone parameter includes, in some embodiments, a gentle and friendly tone for responding warmly to hosts, family members, etc.; in other embodiments, a formal and polite tone for responding politely to visitors, delivery personnel, etc.; a serious warning tone for issuing a serious warning to threatening individuals; and a cold rejection tone for responding coldly to those who refuse to receive visitors. Furthermore, different tone types correspond to different synthesized voice timbre and speech rate. That is, during speech synthesis, the gentle and friendly, formal and polite, serious warning, and cold rejection tones each employ a corresponding timbre and speech rate. For example, the gentle and friendly tone uses a soft timbre and a relaxed speech rate, while the serious warning tone uses a deep timbre and a faster speech rate. Based on this, by configuring different timbres and speech rates for different tone types, the speech response is made consistent with the intended tone at the auditory level, enhancing the emotional expressiveness of the interaction.

[0074] In an exemplary embodiment, the behavioral actions defined by the behavioral action parameters include at least one of the following: adjusting the locked or unlocked state of a door lock, adjusting the color and brightness of a lighting device, controlling the tracking direction of a camera, and adjusting the display content of a display screen.

[0075] Specifically, the actions defined by the behavior parameters can be varied in several ways. In some embodiments, this includes adjusting the lock or unlock status of a door lock, such as unlocking for an authorized object and locking for a threatening object. In other embodiments, it includes adjusting the color and brightness of lighting equipment, such as turning on the door light or using a specific color to indicate a visitor at night. In still other embodiments, it includes controlling the tracking direction of a camera, such as making the camera rotate to continuously record as the target object moves. In yet another embodiment, it includes adjusting the content displayed on the screen, such as displaying a welcome message or a prompt message. In still another embodiment, the behavior parameters define two or more of the above-mentioned actions simultaneously. Based on this, by flexibly specifying the actions of terminal actuators such as door locks, lighting, cameras, and screens through behavior parameters, interactive operations can be coordinated with physical execution and visual feedback in addition to voice response, achieving a more comprehensive and differentiated interaction.

[0076] Please refer to Figure 6 The embodiments of this application also provide a security system, including an image acquisition unit 211, a voice acquisition unit 212, a memory 213, a processor 214, a voice output unit 215, and at least one terminal actuator 216. Image acquisition device 211 is used to acquire images of a target object and obtain image data; The voice acquisition device 212 is used to acquire voice data from a target object. Memory 213 is used to store multiple preset templates; The processor 214 is used to acquire the identity information of the target object determined by analyzing the image data, acquire the demand information of the target object obtained by parsing the voice data, and determine the target template from multiple preset templates based on the identity information and demand information. The voice output device 215 is used to output a corresponding voice response based on the target template. At least one terminal executor 216 is used to perform corresponding behavioral actions according to the target template.

[0077] Reference Figure 6As shown, the security system includes an image acquisition unit 211, a voice acquisition unit 212, a memory 213, a processor 214, a voice output unit 215, and at least one terminal actuator 216. The image acquisition unit 211, voice acquisition unit 212, memory 213, voice output unit 215, and terminal actuator 216 are electrically connected to the processor 214. In some embodiments, the security system further includes a data bus, through which the image acquisition unit 211, voice acquisition unit 212, memory 213, processor 214, voice output unit 215, and terminal actuator 216 are coupled to transmit image data, voice data, control commands, and template data between the components. In some embodiments, the security system further includes a communication interface, through which the processor 214 establishes a communication connection with a cloud device to upload voice data to the cloud device and receive the parsing results returned by the cloud device.

[0078] Specifically, the image acquisition unit 211 is located on one side of the terminal device's panel, with its imaging optical axis facing the area to be acquired. It is used to image the target object entering its field of view to obtain image data. The voice acquisition unit 212 is located on the terminal device and faces the area to be acquired. It is used to acquire the sound signal emitted by the target object to obtain voice data. The voice output unit 215 is located on the terminal device and is used to play a voice response to the area where the target object is located. The terminal actuator 216 is located on the terminal device itself or a controlled device linked to the terminal device. It is used to execute actions under the drive of the processor 214. The memory 213 stores multiple preset templates and computer programs executable by the processor 214. The processor 214 reads and executes the computer programs in the memory 213 to acquire the identity information and demand information of the target object, determine the target template, and drive the voice output unit 215 and the terminal actuator 216. Thus, the security system forms a complete signal link from data acquisition and decision-making to execution. The processor 214 combines the acquired identity information and demand information with the preset template in the memory 213 to make a decision, and then sends the decision result to the voice output device 215 and the terminal executor 216, so that the security system can automatically give different responses to target objects with different identities and different needs.

[0079] Please refer to Figure 7 , Figure 7 This is a schematic diagram of another security system provided in this application. Based on the above embodiments, the image acquisition device 211 is a terminal image acquisition device 2111, and the security system also includes a terminal image processor 217 and a terminal image enhancer 218. The terminal image processor 217 is used to analyze the facial features and morphological features of a target object in image data based on an image recognition model, and to determine the target identity category of the target object using facial features and / or morphological features. The terminal image intensifier 218 is used to perform image quality enhancement processing according to the current ambient lighting conditions. The terminal image intensifier 218 includes at least one of an artificial intelligence image signal processor, an infrared fill light, and a high dynamic range imager.

[0080] The terminal image acquisition unit 2111 is located on one side of the panel of the terminal device, with its imaging optical axis facing the area to be acquired, meaning that image acquisition is completed locally on the terminal device. The terminal image processor 217 and the terminal image intensifier 218 are integrated on the main control board of the terminal device and are electrically connected to the terminal image acquisition unit 2111 and the processor 214, respectively. They can exist as independent chips or be integrated into the same system-on-a-chip.

[0081] The terminal image processor 217 analyzes the facial and morphological features of the target object in the image data based on an image recognition model, and determines the target's identity category using facial and / or morphological features, thus placing the identity category determination locally on the terminal device. The terminal image enhancer 218 is used to enhance image quality according to the current ambient lighting conditions. In some embodiments, the terminal image enhancer 218 is an artificial intelligence image signal processor; in others, it is an infrared fill light; in still others, it is a high dynamic range imager; and in yet another embodiment, it is a combination of two or three of the above devices. Therefore, by setting the terminal image processor 217 and the terminal image enhancer 218 on the terminal device side, both image enhancement and identity category determination can be completed locally on the terminal, improving recognition accuracy in low-light, no-light, and backlight environments, and allowing identity category determination to be completed before voice interaction begins.

[0082] In one exemplary embodiment, the voice collector 212 is a terminal voice collector 2121, and the security system also includes a cloud voice processor 220; The cloud-based voice processor 220 is used to analyze the sentiment and semantics of voice data to obtain demand information; the sentiment analysis results include at least one of urgency, politeness, and threat; the semantic analysis results include at least one of door opening requests, package retrieval requests, and person finding requests. The processor 214 is specifically used to select a target template that matches the combination from multiple preset templates based on the combination of identity information and demand information.

[0083] In this embodiment, the voice collector 212 is a terminal voice collector 2121, meaning that voice collection is completed locally on the terminal.

[0084] In some embodiments, the terminal voice collector 2121 is a microphone mounted on a doorbell, which collects audio data in real time. The cloud-based voice processor 220 is located in a cloud device and is used to analyze the emotional tendency and semantics in the voice data to obtain demand information, thereby placing the computationally intensive semantic and emotional analysis in the cloud. Placing the semantic and emotional analysis of the voice in the cloud reduces the consumption of terminal resources such as flash memory, and the cloud model has better performance and is easier to switch on demand. The processor 214 is specifically used to select a target template that matches the combination of identity information and demand information from multiple preset templates. Based on this, the security system uses a collaborative approach of terminal voice collection and cloud semantic analysis, which reduces the burden on terminal resources while improving the accuracy of demand analysis, and selects a matching target template based on the combination of identity and demand.

[0085] In an exemplary embodiment, the processor 214 is further configured to acquire current time information and geographical location information of the terminal voice collector 2121, and select a matching target template from multiple preset templates based on a combination of current time information, geographical location information, identity information and demand information.

[0086] In this embodiment, the processor 214 also obtains the current time information and the geographical location information of the terminal voice collector 2121, and combines the current time information, geographical location information, identity information and demand information together to select a matching target template from multiple preset templates.

[0087] In some embodiments, for the same identity and the same needs, the processor 214 selects different target templates at night and during the day, or in different geographical locations. Based on this, the processor 214 further considers time and geographical location factors when selecting target templates, so that the response output by the security system can adapt to the differences in time of day and location.

[0088] In one exemplary embodiment, the processor 214 is further configured to receive a custom configuration submitted by an authorized user through a terminal application, and generate a user-defined template as a new preset template.

[0089] In some embodiments, authorized users can set their preferred response tone, dialogue content, or actions through a terminal application; in other embodiments, authorized users can record response voices for specific scenarios through a terminal application. The processor 214 generates a user-defined template based on the received custom configuration and stores it as a new preset template for later use when selecting a target template. Based on this, authorized users can customize templates according to their own preferences, ensuring that the responses output by the security system match the user's personalized needs, thereby improving the personalization of the interactive experience and user satisfaction.

[0090] In an exemplary embodiment, the at least one terminal actuator 216 includes at least one of an access control lock, a lighting device, a camera tracking mechanism, and a display screen.

[0091] In this embodiment, the terminal actuator 216 may be an access control lock, configured to perform a locking or unlocking action; the terminal actuator 216 may also be a lighting device, configured to perform an action of adjusting color and brightness; the terminal actuator 216 may also be a camera tracking mechanism, configured to perform an action of adjusting the camera tracking direction; the terminal actuator 216 may also be a display screen, configured to perform an action of adjusting displayed content. The security system includes two or more of the above terminal actuators 216 at the same time.

[0092] As an optional embodiment, when a visitor such as a friend arrives, the processor 214 may link the display screen or speaker serving as the terminal actuator 216 in the system to play corresponding audio and video to alert the authorized user that someone is visiting. On this basis, by configuring the terminal actuators 216 such as the access control lock, lighting device, camera tracking mechanism and display screen, the security system can implement differentiated interaction behaviors on multiple terminals, realizing the linkage of responses at the physical level and the display level.

[0093] In an exemplary embodiment, the memory 213 is further configured to cache historical interaction records, where the historical interaction records include the historical target identity category of each target object and the corresponding historical selection template; The processor 214 is further configured to, when the same target object visits again, preferentially load the historical selection template, and perform incremental update on the historical selection template in combination with the current recognition result.

[0094] In this embodiment, the memory 213 also caches historical interaction records, in which the historical target identity category of each target object and its corresponding historical selection template are recorded. When the same target object visits again, the processor 214 preferentially loads the historical selection template of the target object, so that there is no need to perform template selection from scratch during repeated visits. The processor 214 also performs incremental update on the historical selection template in combination with the current recognition result, for example, when there is a difference between the current recognition result and the historical record, the historical selection template is adjusted or supplemented accordingly.

[0095] In some embodiments, the historical interaction records further include records of visitor feedback, for feeding back to the authorized user. On this basis, by caching and preferentially loading the historical selection template, the security system can quickly give a response consistent with the previous one when the same object visits repeatedly, and use incremental update to make the template continuously fit the reality as interactions accumulate.

[0096] In an exemplary embodiment, the security system further includes an optical character recognizer 219; Optical character recognition device 219 is used to perform optical character recognition on image data and extract express waybill information or package information from the image data; The processor 214 is also used to generate corresponding voice prompts based on the recognition results of the optical character recognizer 219, and output the voice prompts through the voice outputter 215.

[0097] The optical character recognition (OCR) device 219 performs optical character recognition on image data, extracting express waybill or package information from the image data. In some embodiments, it extracts waybill information such as waybill number and recipient information; in other embodiments, it extracts package information from the outer packaging of the package. The processor 214 generates corresponding voice prompts based on the recognition results of the OCR device 219 and outputs them through the voice output device 215. Based on this, the security system sets up the OCR device 219 at the hardware level, enabling it to automatically recognize waybill or package text and generate and play targeted voice prompts in scenarios such as express delivery.

[0098] Embodiments of this application also provide an electronic device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the interaction method of the security system as described in any of the embodiments above.

[0099] The electronic device includes a processor, a memory, a communication interface, and a data bus. The processor, memory, and communication interface are interconnected via the data bus to enable communication between them. The memory stores computer programs, and the processor executes the computer programs stored in the memory to implement the interaction method of the security system described below. In some embodiments, the processor is a central processing unit (CPU). In other embodiments, the processor is a graphics processor, a digital signal processor, an application-specific integrated circuit (ASIC), or a field-programmable gate array (FPGA), or a combination of two or more of the above devices. In some embodiments, the memory includes read-only memory (ROM) and random access memory (RAM), and provides instructions and data to the processor. In other embodiments, the memory also includes non-volatile memory such as flash memory for caching facial feature databases and historical interaction records. In some embodiments, the electronic device is a terminal device such as a smart doorbell, access control machine, service robot, or smart speaker. In other embodiments, the electronic device is a server that works in conjunction with the terminal device, i.e., the aforementioned cloud device. Thus, by using the cooperation of the processor and memory to execute computer programs, the electronic device can implement the interaction method of the security system described in any of the above embodiments, and can be implemented on a general-purpose processor without relying on specific dedicated circuits, facilitating deployment and upgrades on existing terminal products.

[0100] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0101] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An interaction method for a security system, characterized in that, include: Acquire image data of the target object, and determine the identity information of the target object based on the image data; Acquire the voice data of the target object, and parse the demand information of the target object based on the voice data, wherein at least a portion of the voice data parsing is performed by a cloud device; Based on the identity information and the requirement information, a target template is determined from multiple preset templates; According to the target template, the corresponding interactive operation is executed, which includes outputting the corresponding voice response and / or driving at least one terminal actuator to perform the corresponding action.

2. The interaction method of the security system according to claim 1, characterized in that, The identity information includes the target identity category; Acquiring image data of a target object and determining the identity information of the target object based on the image data includes: The image data is obtained by acquiring images of the target object through a terminal device; Based on the image recognition model, the facial features and morphological features of the target object in the image data are analyzed; The target identity category of the target object is determined using the facial features and / or the morphological features; wherein, multiple preset identity categories include authorized identity categories and unauthorized identity categories; the authorized identity categories include at least two of the following: owner, family member, acquaintance, courier, and visitor; wherein, the unauthorized identity categories include user-defined dangerous persons.

3. The interaction method of the security system according to claim 2, characterized in that, Determining the target identity category of the target object using the facial features and / or the morphological features includes: Using the facial features and / or the morphological features, determine the target identity category that matches the target object from a plurality of preset identity categories; and / or, When the facial feature fails to match the corresponding authorized identity category in the first feature library, the morphological feature is matched with each reference morphological feature in the second feature library, which includes the correspondence between the authorized identity category and the reference morphological feature. The target identity category of the target object is determined based on the matching results.

4. The interaction method of the security system according to claim 2, characterized in that, The interaction method of the security system also includes: When the target identity category of the target object is determined to be the unauthorized identity category, the terminal device is controlled to enter an automatic enhanced monitoring mode. This automatic enhanced monitoring mode includes at least one of the following: increasing the image acquisition frame rate, initiating multi-angle camera tracking, and increasing the environmental sensor sampling frequency; and / or, When the target identity category of the target object is determined to be the unauthorized identity category, the behavior trajectory of the target object is obtained, and the behavior trajectory is used to determine whether the target object has loitering behavior. An alarm is triggered when the target object exhibits loitering behavior.

5. The interaction method of the security system according to claim 2, characterized in that, Before acquiring an image of the target object via the terminal device, the process further includes: based on the current ambient lighting conditions, activating at least one of the following methods for image quality enhancement: artificial intelligence image signal processor enhancement, infrared illumination, and high dynamic range imaging; and / or, After determining the target identity category of the target object using the facial features and / or the morphological features, the method further includes: outputting guiding voice prompts based on the target identity category; acquiring voice data of the target object based on the guiding voice prompts, and analyzing the emotional tendency and semantics of the voice data through a cloud device to obtain demand information; the analysis result of the emotional tendency includes at least one of urgency, politeness, and threat; the analysis result of the semantics includes at least one of door opening request, package pickup request, and person finding request; and / or, The interaction method of the security system further includes: performing optical character recognition on the image data, extracting express waybill information or package information from the image data, and generating corresponding voice prompts based on the recognition results.

6. The interaction method of the security system according to claim 5, characterized in that, Determining a target template from multiple preset templates based on the identity information and the requirement information includes: Based on the combination of the identity information and the demand information, a target template that matches the combination is selected from the plurality of preset templates; The step of selecting a target template that matches the combination of the identity information and the demand information from the plurality of preset templates further includes: obtaining current time information and the geographical location information of the terminal device, and selecting a matching target template from the plurality of preset templates based on the combination of the current time information, the geographical location information, the demand information and the identity information; Among the multiple preset templates is a serious warning template; when the analysis result of the emotional tendency is a threat, the serious warning template is determined as the target template; according to the serious warning template, the door lock is controlled to lock and an alarm notification is generated.

7. The interaction method of the security system according to any one of claims 1 to 6, characterized in that, The preset template includes at least response tone parameters, semantic content library parameters, and behavioral action parameters; wherein: The tone type defined by the response tone parameter includes at least one of the following: gentle and friendly, formal and polite, serious and warning, and cold and rejecting. Different tone types correspond to different speech synthesis timbre and speech rate. The behavioral actions defined by the behavioral action parameters include at least one of the following: adjusting the locked or unlocked state of the door lock, adjusting the color and brightness of the lighting equipment, controlling the tracking direction of the camera, and adjusting the display content of the screen.

8. A security system, characterized in that, It includes an image acquisition unit (211), a voice acquisition unit (212), a memory (213), a processor (214), a voice output unit (215), and at least one terminal actuator (216). The image acquisition device (211) includes an imaging sensor and an optical lens, which are oriented toward the area to be acquired to generate image data; The voice collector (212) is used to collect voice data from the target object. The memory (213) is used to store multiple preset templates; The processor (214) is used to acquire the identity information of the target object determined by analyzing the image data, acquire the demand information of the target object obtained by parsing the voice data, and determine the target template from the plurality of preset templates according to the identity information and the demand information; The voice output device (215) is used to output a corresponding voice response according to the target template; At least one of the terminal executors (216) is used to perform corresponding behavioral actions according to the target template.

9. The security system according to claim 8, characterized in that, The image acquisition device (211) is a terminal image acquisition device (2111), and the security system also includes a terminal image processor (217) and a terminal image intensifier (218); wherein: The terminal image processor (217) is used to analyze the facial features and morphological features of the target object in the image data based on the image recognition model, and to determine the target identity category of the target object using the facial features and / or the morphological features; The terminal image enhancer (218) is used to perform image quality enhancement processing according to the current ambient lighting conditions. The terminal image enhancer (218) includes at least one of an artificial intelligence image signal processor, an infrared fill light, and a high dynamic range imager; and / or, The security system also includes: The cloud-based voice processor (220) is used to analyze the emotional tendencies and semantics in the voice data to obtain the demand information; the analysis results of the emotional tendencies include at least one of urgency, politeness, and threat; the analysis results of the semantics include at least one of door opening request, package pickup request, and person search request; the processor (214) is specifically used to select a target template that matches the combination from the plurality of preset templates based on the combination of the identity information and the demand information. The processor (214) is also configured to acquire current time information and the geographical location information of the terminal voice collector (2121), and select a matching target template from the plurality of preset templates based on the combination of the current time information, the geographical location information, the identity information, and the demand information; and / or, the processor (214) is also configured to receive a custom configuration submitted by an authorized user through a terminal application, and generate a user-defined template as a new preset template; and / or, At least one of the terminal actuators (216) includes at least one of an access control lock, a lighting device, a camera tracking mechanism, and a display screen; and / or, The memory (213) is also used to cache historical interaction records, which include the historical target identity category and the corresponding historical selection template for each target object; The processor (214) is further configured to, when the same target object is visited again, prioritize loading the historical selection template and incrementally update the historical selection template in conjunction with the current identification result; and / or, The security system also includes an optical character recognition device (219); the optical character recognition device (219) is used to perform optical character recognition on the image data and extract express waybill information or package information from the image data; the processor (214) is also used to generate corresponding voice prompts based on the recognition results of the optical character recognition device (219) and output the voice prompts through the voice output device (215).

10. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the interaction method of the security system as described in any one of claims 1 to 7.

Images