An integrated camera device based on big data

Abstract

This disclosure proposes an integrated camera device based on big data, comprising: a camera body, an infrared sensing unit, a face recognition unit, and a big data iteration unit. The face recognition unit is electrically connected to the camera body, the infrared sensing unit, and the big data iteration unit. The camera body is used to collect image data of personnel in real time; the infrared sensing unit is used to collect environmental perception auxiliary data; the big data iteration unit iteratively optimizes the face recognition threshold and risk assessment rules based on historical data, and sends the optimized threshold and rules to the face recognition unit; the face recognition unit fuses the image data and environmental perception auxiliary data to form recognition data, and compares the recognition data with the optimized threshold and risk assessment rules to generate a real-time security assessment strategy. This disclosure achieves dynamic and intelligent upgrading of camera security recognition through big data iterative optimization.

Description

Technical Field

[0001] This disclosure relates to the field of big data and security monitoring camera technology, and in particular to an integrated camera device based on big data. Background Technology

[0002] With the rapid development of smart cities and intelligent security technologies, video surveillance cameras have been widely used in various places such as park security, access control, and public area security. Surveillance equipment is gradually upgrading from a passive recording and storage mode to an intelligent mode with proactive intelligent identification and risk warning. Among these developments, combining big data technology with surveillance cameras to improve facial recognition accuracy and security risk assessment is one of the current development directions in the field of intelligent security.

[0003] Currently, existing smart surveillance cameras that integrate big data still have many technical shortcomings in practical applications, which restrict the accuracy and adaptability of security identification. First, most existing smart cameras rely solely on single video image data for facial recognition, lacking the fusion of multi-dimensional environmental perception data. They cannot comprehensively analyze environmental characteristics such as the duration of human stay, resulting in limited data dimensions, weak anti-interference capabilities, and susceptibility to environmental factors such as lighting, obstruction, and long-distance shooting, leading to misidentification and missed identification, and poor recognition reliability.

[0004] Secondly, the judgment mechanism of existing big data intelligent monitoring systems is severely rigid, generally adopting fixed facial recognition thresholds and risk assessment rules, which cannot be adaptively and iteratively optimized according to different application locations and different data characteristics of passersby. The personnel density, personnel characteristics, and risk assessment standards vary greatly in different security locations. Fixed judgment parameters and rules cannot adapt to the complex and ever-changing actual security situations, resulting in insufficient recognition accuracy in general scenarios and failure of risk assessment in special scenarios, with extremely poor scenario generalization ability.

[0005] Furthermore, in existing technologies, image features and environmental perception data are mostly collected and analyzed independently, lacking standardized temporal alignment and differentiated weighted fusion mechanisms. This makes it impossible to effectively link and correct multiple features such as facial similarity, personnel movement trajectory, and dwell time. The utilization rate of effective recognition features is low, and invalid interference data cannot be effectively removed, further reducing the accuracy of facial recognition and risk assessment.

[0006] In summary, current intelligent security devices that combine big data with cameras suffer from problems such as weak data fusion capabilities, fixed judgment parameters, poor scene adaptability, and lack of risk classification and control. There is an urgent need for an integrated camera device that can achieve multi-feature differentiated fusion, big data adaptive iterative optimization, and hierarchical intelligent security judgment to solve the shortcomings of existing technologies. Summary of the Invention

[0007] To address the aforementioned technical problems, this disclosure provides an integrated camera device based on big data, comprising: a main body, an infrared sensing unit, a face recognition unit, and a big data iteration unit. The infrared sensing unit and the face recognition unit are integrated on the main body, and the face recognition unit is electrically connected to the main body, the infrared sensing unit, and the big data iteration unit, respectively. The main body is used to collect image data of personnel in real time; the infrared sensing unit is used to collect environmental perception auxiliary data; the big data iteration unit iteratively optimizes the face recognition threshold and risk assessment rules based on historical data, and sends the optimized face recognition threshold and risk assessment rules to the face recognition unit; the face recognition unit is used to fuse image data and environmental perception auxiliary data to form recognition data, and compares the recognition data with the optimized face recognition threshold and risk assessment rules to generate a real-time security assessment strategy.

[0008] Optionally, the big data iteration unit includes a data statistics module, an iteration module, and a parameter distribution module that are electrically connected in sequence. The data statistics module is electrically connected to the output end of the face recognition unit to receive the recognition data sent by the face recognition unit and collect it to obtain historical data. The iteration module iterates based on the historical data to obtain the optimized face recognition threshold and risk judgment rules. The parameter distribution module is connected to the parameter configuration interface of the face recognition unit to send the optimized face recognition threshold and risk judgment rules to the face recognition unit.

[0009] Optionally, the face recognition unit includes a data receiving module, a fusion processing module, and a strategy comparison output module connected in sequence. The data receiving module receives image data of the person sent by the subject and environmental perception auxiliary data sent by the infrared sensing unit, and transmits it to the fusion processing module. The fusion processing module performs feature association fusion on the image data and environmental perception auxiliary data to generate recognition data. The strategy comparison output module compares the recognition data with the optimized face recognition threshold and risk judgment rules sent by the parameter sending module to generate a real-time security judgment strategy.

[0010] Optionally, the main body includes an image acquisition module, an encoding module, and a communication module that are electrically connected in sequence. The image acquisition module is used to acquire images of people, the encoding module is used to encode the images of people to obtain image data, and the image data is transmitted to the face recognition unit through the communication module.

[0011] Optionally, the infrared sensing unit includes a detection module, a data processing module, and an output module that are electrically connected in sequence. The detection module is used to detect raw environmental data in real time, the data processing module is used to filter the raw environmental data to obtain environmental perception auxiliary data, and the output module is electrically connected to the face recognition unit to transmit the environmental perception auxiliary data to the face recognition unit.

[0012] Optionally, the integrated camera also includes an alarm unit, which is electrically connected to the face recognition unit. After the face recognition unit generates a real-time security judgment strategy, it drives the alarm unit to perform corresponding warning and alarm actions.

[0013] Optionally, the image data includes facial similarity features and motion trajectory features of people, and the environmental perception auxiliary data includes the duration of people's stay. The fusion processing module performs temporal alignment and feature fusion on the image data and environmental perception auxiliary data at the same time, removes invalid interference data, and generates recognition data including facial matching scores and effective stay duration.

[0014] Optionally, the fusion processing module uses differentiated weighting ratios to complete feature weighted fusion. The face matching score is calculated by weighting the face similarity features of the person first, and the effective dwell time is obtained by jointly correcting and weighting the movement trajectory features of the person and the dwell time data.

[0015] Optionally, historical data includes normal passage records and stranger identification records. The iteration module optimizes offline based on historical data to obtain face recognition thresholds adapted to different scenarios, as well as risk judgment thresholds as quantitative parameters for risk judgment rules. The optimized face recognition thresholds and risk judgment thresholds are then sent to the face recognition unit. The strategy comparison and output module matches the current scenario and calls the corresponding optimized face recognition thresholds and risk judgment thresholds to compare the recognition data in real time and determine the risk level of the personnel.

[0016] Optionally, the security judgment strategy includes a normal passage strategy, a mild warning strategy, and a high-risk alarm strategy. When the face matching score is not lower than the face recognition threshold and the effective stay time is less than the risk judgment threshold, the normal passage strategy is executed; when the face matching score is lower than the face recognition threshold and the effective stay time is less than the risk judgment threshold, the mild warning strategy is executed and light and voice warnings are triggered; when the face matching score is lower than the face recognition threshold and the effective stay time exceeds the risk judgment threshold, the high-risk alarm strategy is executed and sound and light alarms and background alarm uploads are triggered.

[0017] By adopting the above technical solution, the present invention has at least the following beneficial effects: This disclosure proposes an integrated camera device based on big data. By fusing image data and environmental perception auxiliary data for multi-feature recognition, it improves the anti-interference capability and recognition accuracy. At the same time, by using big data and historical data, it continuously iterates and optimizes the face recognition threshold and risk judgment rules offline. It can adaptively update and match the optimal judgment parameters and rules according to different situations, overcoming the problems of traditional smart cameras such as single recognition dimension, fixed parameters, rigid rules, and poor scene adaptability. It improves the accuracy and universality of personnel risk level judgment and realizes the dynamic and intelligent upgrade of camera security recognition. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of an integrated camera device according to an embodiment of the present disclosure; Figure 2 This is a flowchart of a face recognition threshold and risk assessment threshold according to an embodiment of the present disclosure. Detailed Implementation

[0020] The embodiments of this disclosure will be further described in detail below with reference to the accompanying drawings and examples. The detailed description of the embodiments and the accompanying drawings are used to illustrate the principles of this disclosure by way of example, but should not be used to limit the scope of this disclosure. This disclosure can be implemented in many different forms and is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

[0021] These embodiments are provided to make the disclosure thorough and complete, and to fully express the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specifically stated, the relative arrangement of components and steps, material composition, numerical expressions, and values ​​set forth in these embodiments should be interpreted as exemplary only and not as limiting.

[0022] It should be noted that, in the description of this disclosure, unless otherwise stated, "a plurality of" means two or more; the terms "upper," "lower," "left," "right," "inner," and "outer," etc., indicating orientation or positional relationship, are only for the convenience of describing this disclosure and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this disclosure. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0023] Furthermore, the terms "first," "second," and similar terms used in this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different parts. "Vertical" is not strictly vertical, but within the permissible margin of error. "Parallel" is not strictly parallel, but within the permissible margin of error. Terms such as "including" or "contains" mean that the element preceding the word encompasses the element listed after the word, and do not exclude the possibility of encompassing other elements as well.

[0024] It should also be noted that, in the description of this disclosure, unless otherwise expressly specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this disclosure depending on the specific circumstances. When a particular device is described as being located between a first device and a second device, an intermediary device may or may not be present between the particular device and the first or second device.

[0025] All terms used in this disclosure have the same meaning as understood by one of ordinary skill in the art to which this disclosure pertains, unless otherwise specifically defined. It should also be understood that terms defined in general dictionaries should be interpreted as having meanings consistent with their meanings in the context of the relevant art, and not as idealized or highly formalized, unless expressly defined herein.

[0026] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, they should be considered part of the specification.

[0027] As mentioned in the background section, most existing smart cameras rely solely on single image data for facial recognition. This results in a limited recognition dimension, susceptibility to environmental factors such as lighting changes, human obstruction, and image interference, weak anti-interference capabilities, and limited recognition accuracy, making them unsuitable for complex on-site security environments. Therefore, this application proposes an integrated camera device based on big data to address these technical problems.

[0028] To address the aforementioned technical problems, this invention provides an integrated camera device based on big data, such as... Figure 1 As shown, the system includes: a main body, an infrared sensing unit, a face recognition unit, and a big data iteration unit. The infrared sensing unit and the face recognition unit are integrated on the main body. The face recognition unit is electrically connected to the main body, the infrared sensing unit, and the big data iteration unit. The main body is used to collect image data of personnel in real time; the infrared sensing unit is used to collect environmental perception auxiliary data; the big data iteration unit iteratively optimizes the face recognition threshold and risk judgment rules based on historical data, and sends the optimized face recognition threshold and risk judgment rules to the face recognition unit; the face recognition unit is used to fuse image data and environmental perception auxiliary data to form recognition data, and compares the recognition data with the optimized face recognition threshold and risk judgment rules to generate a real-time security judgment strategy.

[0029] Specifically, the main unit, acting as the image acquisition carrier, is responsible for capturing real-time images of the monitored area, collecting personnel images, and generating corresponding image data. The infrared sensing unit, independent of visual image acquisition, is responsible for collecting environmental perception auxiliary data to compensate for the recognition deficiencies of single image data. The big data iteration unit is the core of the device's intelligent optimization, capable of continuously collecting historical data generated during device operation. Based on this extensive historical data, it continuously iterates and optimizes facial recognition thresholds and risk assessment rules, and then sends the optimized, precise parameters and rules to the facial recognition unit. The facial recognition unit, as the core judgment module, receives image data and environmental perception auxiliary data, performs multi-feature fusion, generates standardized recognition data, and simultaneously matches the optimized thresholds and judgment rules issued by the big data iteration unit to generate corresponding security judgment strategies in real time, achieving automated and intelligent security management.

[0030] Furthermore, the big data iteration unit includes a data statistics module, an iteration module, and a parameter distribution module that are electrically connected in sequence. The data statistics module is electrically connected to the output end of the face recognition unit to receive the recognition data sent by the face recognition unit and collect it to obtain historical data. The iteration module iterates based on the historical data to obtain the optimized face recognition threshold and risk judgment rules. The parameter distribution module is connected to the parameter configuration interface of the face recognition unit to send the optimized face recognition threshold and risk judgment rules to the face recognition unit.

[0031] Specifically, the data statistics module continuously connects to the output of the face recognition unit, receiving various recognition data generated during the device's daily operation in real time. It categorizes, collects, and stores data such as normal passage and unfamiliar person identification, compiling a dedicated historical data sample library to support subsequent iterative optimization. The iteration module retrieves the collected historical data and performs offline iterative training and optimization based on big data algorithms, continuously correcting and updating the face recognition thresholds and risk assessment rules to adapt to different scenarios, abandoning the traditional fixed parameter mode. The parameter distribution module, acting as the parameter transmission carrier, interfaces with the face recognition unit's parameter configuration interface, distributing the latest iteratively optimized face recognition thresholds and risk assessment rules in real time, achieving adaptive updates of device parameters.

[0032] Furthermore, the face recognition unit includes a data receiving module, a fusion processing module, and a strategy comparison output module connected in sequence. The data receiving module receives image data of the person sent by the host and environmental perception auxiliary data sent by the infrared sensing unit, and transmits it to the fusion processing module. The fusion processing module performs feature association fusion on the image data and environmental perception auxiliary data to generate recognition data. The strategy comparison output module compares the recognition data with the optimized face recognition threshold and risk judgment rules sent by the parameter sending module to generate a real-time security judgment strategy.

[0033] Specifically, the data receiving module monitors the data from the acquisition end in real time, synchronously receiving personnel image data transmitted by the main body and environmental perception auxiliary data transmitted by the infrared sensing unit, and uniformly transmits the two types of heterogeneous data to the fusion processing module. The fusion processing module performs time-series alignment processing on the image data and environmental perception auxiliary data, eliminating the time difference between the acquisition of the two types of data, completing multi-dimensional feature association fusion, removing redundant information such as image interference and invalid acquisition data, and generating accurate and effective standardized recognition data. The strategy comparison and output module retrieves the optimized face recognition threshold and risk judgment rules issued by the big data iteration unit, accurately compares the real-time generated recognition data with the standard parameters and rules, and outputs the corresponding real-time security judgment strategy in combination with the preset judgment logic to complete the intelligent judgment of personnel risk.

[0034] Furthermore, the main body includes an image acquisition module, an encoding module, and a communication module that are electrically connected in sequence. The image acquisition module is used to acquire images of people, the encoding module is used to encode the images of people to obtain image data, and the communication module transmits the image data to the face recognition unit.

[0035] Specifically, the image acquisition module uses a high-definition lens to acquire real-time dynamic images of people within the monitoring area, capturing visual features such as faces and movement trajectories; the encoding module compresses, encodes, and converts the acquired raw images, removing invalid pixels and noise from the raw images to generate standardized, identifiable image data, ensuring the stability of data transmission and processing; the communication module has a stable data transmission function, transmitting the encoded image data to the face recognition unit in real time and without interruption, providing raw image data support for subsequent multi-feature fusion recognition, and ensuring the continuity of monitoring and acquisition throughout the day.

[0036] Furthermore, the infrared sensing unit includes a detection module, a data processing module, and an output module that are electrically connected in sequence. The detection module is used to detect raw environmental data in real time, the data processing module is used to filter the raw environmental data to obtain environmental perception auxiliary data, and the output module is electrically connected to the face recognition unit to transmit the environmental perception auxiliary data to the face recognition unit.

[0037] Specifically, the detection module uses infrared sensing to detect the dynamic environmental data of people in the monitoring area in real time; the data processing module filters and denoises the collected raw infrared data, removing invalid data such as ambient temperature and minor debris interference, and purifies it to obtain accurate environmental perception auxiliary data; the output module is electrically connected to the face recognition unit to transmit the processed environmental perception auxiliary data in real time, complementing the face image data to achieve multi-dimensional data collaborative recognition.

[0038] Furthermore, the integrated camera also includes an alarm unit electrically connected to the face recognition unit. After the face recognition unit generates a real-time security judgment strategy, it drives the alarm unit to execute corresponding warning and alarm actions. Specifically, after generating the real-time security judgment strategy, the face recognition unit outputs corresponding control commands to the alarm unit according to the strategy type, driving the alarm unit to execute differentiated actions. For different judgment results—normal passage, minor warning, and high-risk alarm—the alarm unit can correspondingly perform operations such as silent passage, light and voice warnings, sound and light alarms, and background data upload.

[0039] Furthermore, the image data includes facial similarity features and movement trajectory features of people, while the environmental perception auxiliary data includes the duration of people's stay. The fusion processing module performs temporal alignment and feature fusion on the image data and environmental perception auxiliary data at the same time, removes invalid interference data, and generates recognition data including facial matching scores and effective stay duration.

[0040] Furthermore, the fusion processing module uses differentiated weighting ratios to complete feature weighted fusion. The face matching score is calculated by weighting the facial similarity features of the person first, and the effective dwell time is obtained by jointly correcting and weighting the person's movement trajectory features and dwell time data.

[0041] In one embodiment, the face matching score is calculated by weighting the facial similarity features of individuals, and the specific process is as follows: (1) The face similarity features include the original face feature vector. After extracting facial features, texture and contour features from the ontology, the standardized original face feature vector A is output. 原始 =[a1, a2, a3, a4, a5], where a1, a2, a3, a4, and a5 are respectively: the overall contour feature quantization value of the face, the relative position structure feature of the facial features, the facial texture detail feature quantization value, the key topological point feature quantization value of the face, and the steady-state feature quantization value of the face's light and shadow.

[0042] (2) Set the priority weight coefficient W for facial features, and perform feature weighting enhancement and noise suppression on the original facial feature vector to obtain a weighted enhanced facial feature vector. The original facial feature vector is then weighted and corrected dimension by dimension to strengthen the effective facial feature components and suppress the invalid noise components introduced by slight image blurring and illumination fluctuations. The weighting formula is: A = A 原始 ×W+A 均值 ×(1-W)=[A1,A2,A3,A4,A5],A 均值 = (a1 + a2 + a3 + a4 + a5) ÷ 5, where: A is the weighted enhanced facial feature vector, A 均值 The mean of the original face feature vector is used to preserve the overall feature distribution and avoid weighted distortion.

[0043] (3) The face matching score is calculated using the general cosine similarity calculation formula for face recognition based on the weighted enhanced face feature vector and the standard feature vector of the personnel database.

[0044] In one embodiment, the effective dwell time is obtained by a weighted average of the personnel movement trajectory characteristics and the personnel dwell time data after joint correction. (1) The infrared sensing unit collects the personnel's stay status and outputs the original stay duration T. 原始 Simultaneously, the core features of the personnel's movement trajectory are extracted, and the trajectory confidence quantization value S is output. This value is obtained by comprehensively judging the three-dimensional features of trajectory continuity quantization value, movement speed quantization value, and position steady-state offset quantization value. The value range is 0 to 1, which is used to characterize the credibility of the personnel's actual stationing movement state and determine whether the personnel are stationed at a fixed point or passing by momentarily.

[0045] (2) Regarding the original stay duration T 原始 The effective dwell time is obtained by fusing and weighting the motion trajectory confidence value S. The specific fusing and weighting formula is as follows: T 有效 =T 原始 ×W t +S×W s In the formula: T 有效 For the effective stay duration, T 原始 W represents the original duration of stay. t W is the baseline weight for the original stay duration. s As the weight for motion trajectory correction, and satisfying W t +W s =1, the confidence quantization value of the S-trajectory.

[0046] Original stay duration T 原始 There is inherent false triggering noise, and relying solely on this duration data is insufficient to distinguish between genuine dwell time and transient interference. This solution employs a dual-weighted joint fusion approach, with W... t Stabilize the basic duration data, avoid distortion caused by duration reset, and combine motion trajectory confidence S with W s Dynamic duration correction: When personnel are actually stationary, the confidence level S of the motion trajectory approaches 1, and the complete and valid duration is retained after weighting; when personnel pass by momentarily or sensors are falsely triggered, the trajectory confidence level S approaches 0, and false duration components are automatically eliminated, achieving the effect of suppressing invalid duration noise. This invention sets two sets of independent fixed weights for duration correction scenarios, which are respectively matched with infrared duration reference data and motion trajectory correction data. Through complementary fusion of dual weights, the component of actual stationary duration is strengthened, and invalid duration noise caused by momentary passing, heat source jitter, and sensor false triggering is suppressed.

[0047] Furthermore, historical data includes normal passage records and stranger identification records. The iterative module optimizes offline based on historical data to obtain face recognition thresholds adapted to different scenarios, as well as risk assessment thresholds as quantitative parameters for risk assessment rules. The optimized face recognition thresholds and risk assessment thresholds are then sent to the face recognition unit. The strategy comparison and output module matches the current scenario and calls the corresponding optimized face recognition thresholds and risk assessment thresholds to compare the recognition data in real time and determine the risk level of the personnel.

[0048] In one embodiment, such as Figure 2 As shown, the iterative module optimizes offline based on historical data to obtain face recognition thresholds suitable for different scenarios, as well as risk judgment thresholds as quantification parameters for risk judgment rules. The specific process is as follows: (1) Iterative sample data collection and classification The iterative module pre-collects historical data generated by the equipment during long-term operation. This historical data is mainly divided into two categories of valid iterative samples: the first category is normal access records, including facial matching scores, effective dwell time, and normal access records for authorized personnel such as community residents and permanent residents; the second category is stranger identification records, including facial matching scores, dwell time, and risk identification records for unfamiliar visitors, passersby, and individuals loitering abnormally. The iterative module classifies and archives these two types of historical data, performs time-series cleaning, removes invalid and corrupted data generated by equipment malfunctions and collection anomalies, and retains valid identification samples from real-world scenarios.

[0049] (2) Offline statistical sample characteristic distribution The iterative module employs an offline batch statistical method to perform feature analysis on the categorized historical sample data: for normal passage records, it analyzes the distribution patterns of authorized personnel's normal face matching scores and effective stay duration to extract the feature baseline range for normal passage; for stranger identification records, it analyzes the feature ranges of unfamiliar personnel's low face matching scores and abnormal stay duration to extract the feature boundaries of abnormal passage and risky loitering. By comparing the feature distributions of the two types of samples, the module distinguishes the differentiated boundaries between legitimate passage features and risk identification features.

[0050] (3) Threshold Iterative Optimization Based on the feature distribution boundaries of two types of historical samples, the iterative module uses an incremental mean iterative algorithm to complete the offline iterative update of the face recognition threshold and risk judgment threshold, and iteratively generates exclusive thresholds adapted to different scenarios.

[0051] The iterative formula for the face recognition threshold is as follows: Th = α × Th1 + (1 - α) × S v In the formula: Th is the face recognition threshold after iterative update, Th1 is the old threshold completed in the previous iteration, α is the iterative smoothing coefficient, and S v This represents the average face matching score of authorized personnel in the current scenario. For example, during the day when lighting is sufficient and facial images are clear, the overall face matching score of authorized personnel is relatively high. After iteration, the face recognition threshold is automatically and appropriately increased to improve the strictness of recognition and reduce the false recognition rate. At night when lighting is low and there is a lot of noise in the image, the overall face feature score is relatively low. The iteration mechanism automatically and appropriately lowers the threshold based on the historical sample average at night to avoid legitimate owners being unable to pass or missing recognition issues.

[0052] The iterative formula for the risk assessment threshold is as follows: R = α × R1 + (1 - α) × T v In the formula: R is the risk judgment threshold after iterative update, R1 is the old threshold completed in the previous iteration, α is the iterative smoothing coefficient (the same as the iterative smoothing coefficient in the face recognition threshold iteration formula to ensure consistent iteration rhythm of the dual thresholds), T v This represents the average duration of abnormal stays by strangers in the current scenario. For example, during the day, when there are many visitors and dense crowds in the community, short stays are normal. The iteration automatically raises the risk assessment threshold based on historical data to avoid frequent false alarms. At night and on holidays, when there are few people in the community, even a short stay by a stranger poses a potential risk. The iteration automatically lowers the risk threshold based on historical risk data for the corresponding time period to improve the sensitivity of nighttime security.

[0053] (4) Threshold update After the iteration module completes offline iteration optimization, it synchronously writes the updated face recognition threshold and risk assessment threshold into the device's recognition program, replacing the original fixed thresholds. During subsequent operation, the device uses the iteratively optimized adaptive thresholds to perform face identity matching and personnel risk assessment, while continuously accumulating new access data and recognition records to provide sample support for the next round of offline iteration.

[0054] Furthermore, the security judgment strategy includes a normal passage strategy, a mild warning strategy, and a high-risk alarm strategy. When the face matching score is not lower than the face recognition threshold and the effective stay time is less than the risk judgment threshold, the normal passage strategy is executed; when the face matching score is lower than the face recognition threshold and the effective stay time is less than the risk judgment threshold, the mild warning strategy is executed and light and voice warnings are triggered; when the face matching score is lower than the face recognition threshold and the effective stay time exceeds the risk judgment threshold, the high-risk alarm strategy is executed and sound and light alarms and background alarm uploads are triggered.

[0055] In summary, compared with existing technologies, this disclosure provides an integrated camera device based on big data. By fusing image data and environmental perception auxiliary data for multi-feature recognition, it improves the anti-interference capability and recognition accuracy. At the same time, by using big data and historical data, it continuously iterates and optimizes the face recognition threshold and risk judgment rules offline. It can adaptively update and match the optimal judgment parameters and rules according to different situations, overcoming the problems of traditional smart cameras such as single recognition dimension, fixed parameters, rigid rules, and poor scene adaptability. It improves the accuracy and universality of personnel risk level judgment and realizes the dynamic and intelligent upgrade of camera security recognition.

[0056] The embodiments of this disclosure have now been described in detail. To avoid obscuring the concept of this disclosure, some details known in the art have not been described. Those skilled in the art can fully understand how to implement the technical solutions disclosed herein based on the above description.

[0057] While specific embodiments of this disclosure have been described in detail by way of examples, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of this disclosure. Those skilled in the art should understand that modifications can be made to the above embodiments or equivalent substitutions can be made to some technical features without departing from the scope and spirit of this disclosure. In particular, as long as there is no structural conflict, the technical features mentioned in the various embodiments can be combined in any manner.

Claims

1. An integrated camera device based on big data, characterized in that, include: The system comprises a body, an infrared sensing unit, a face recognition unit, and a big data iteration unit. The infrared sensing unit and the face recognition unit are integrated on the body. The face recognition unit is electrically connected to the body, the infrared sensing unit, and the big data iteration unit, respectively. The body is used to collect image data of people in real time; The infrared sensing unit is used to collect environmental perception auxiliary data; The big data iteration unit iteratively optimizes the face recognition threshold and risk assessment rules based on historical data, and sends the optimized face recognition threshold and risk assessment rules to the face recognition unit. The face recognition unit is used to fuse the image data and the environmental perception auxiliary data to form recognition data, and to compare the recognition data with the optimized face recognition threshold and the risk judgment rule to generate a real-time security judgment strategy.

2. The integrated camera device based on big data according to claim 1, characterized in that, The big data iteration unit includes a data statistics module, an iteration module, and a parameter distribution module that are electrically connected in sequence. The data statistics module is electrically connected to the output end of the face recognition unit to receive the recognition data sent by the face recognition unit and collect it to obtain the historical data. The iteration module iterates based on the historical data to obtain the optimized face recognition threshold and the risk determination rule. The parameter distribution module is connected to the parameter configuration interface of the face recognition unit to send the optimized face recognition threshold and the risk determination rule to the face recognition unit.

3. The integrated camera device based on big data according to claim 2, characterized in that, The face recognition unit includes a data receiving module, a fusion processing module, and a strategy comparison output module connected in sequence. The data receiving module receives image data of the person sent by the host and environmental perception auxiliary data sent by the infrared sensing unit, and transmits it to the fusion processing module. The fusion processing module performs feature association fusion on the image data and the environmental perception auxiliary data to generate recognition data. The strategy comparison output module compares the recognition data with the optimized face recognition threshold and risk judgment rule sent by the parameter sending module to generate a real-time security judgment strategy.

4. The integrated camera device based on big data according to claim 1, characterized in that, The main body includes an image acquisition module, an encoding module, and a communication module that are electrically connected in sequence. The image acquisition module is used to acquire images of people, the encoding module is used to encode the images of people to obtain image data, and the communication module transmits the image data to the face recognition unit.

5. The integrated camera device based on big data according to claim 1, characterized in that, The infrared sensing unit includes a detection module, a data processing module, and an output module that are electrically connected in sequence. The detection module is used to detect raw environmental data in real time. The data processing module is used to filter the raw environmental data to obtain environmental perception auxiliary data. The output module is electrically connected to the face recognition unit to transmit the environmental perception auxiliary data to the face recognition unit.

6. The integrated camera device based on big data according to claim 1, characterized in that, The integrated camera also includes an alarm unit, which is electrically connected to the face recognition unit. After the face recognition unit generates a real-time security judgment strategy, it drives the alarm unit to perform corresponding warning and alarm actions.

7. The integrated camera device based on big data according to claim 3, characterized in that, The image data includes facial similarity features and movement trajectory features of people. The environmental perception auxiliary data includes the duration of people's stay. The fusion processing module performs temporal alignment and feature fusion on the image data and environmental perception auxiliary data at the same time, removes invalid interference data, and generates the recognition data including facial matching scores and effective stay duration.

8. The integrated camera device based on big data according to claim 7, characterized in that, The fusion processing module uses differentiated weighting ratios to complete feature weighted fusion. The face matching score is calculated by weighting the face similarity features of the person first, and the effective dwell time is obtained by jointly correcting and weighting the person's movement trajectory features and the dwell time data.

9. The integrated camera device based on big data according to claim 2, characterized in that, The historical data includes normal passage records and stranger identification records. The iterative module optimizes offline based on the historical data to obtain face recognition thresholds adapted to different scenarios and risk judgment thresholds as quantitative parameters of the risk judgment rules. The optimized face recognition thresholds and risk judgment thresholds are then sent to the face recognition unit. The strategy comparison and output module matches the current scenario, calls the corresponding optimized face recognition threshold and risk judgment threshold to compare the recognition data in real time, and determines the risk level of the personnel.

10. The integrated camera device based on big data according to claim 9, characterized in that, The security judgment strategy includes a normal passage strategy, a mild warning strategy, and a high-risk alarm strategy. When the face matching score is not lower than the face recognition threshold and the effective stay time is less than the risk judgment threshold, the normal passage strategy is executed. When the face matching score is lower than the face recognition threshold and the effective stay time is less than the risk judgment threshold, the mild warning strategy is executed and light and voice warnings are triggered. When the face matching score is lower than the face recognition threshold and the effective stay time exceeds the risk judgment threshold, the high-risk alarm strategy is executed and sound and light alarms and background alarm uploads are triggered.

Images