A data processing method and system for video surveillance AI intelligent recognition equipment

Abstract

The present invention discloses a data processing method and system for video surveillance AI intelligent recognition equipment. The method specifically includes: obtaining real-time video data from video surveillance AI intelligent recognition equipment in all factory monitoring areas, forming an environmental feature vector based on the real-time video data; based on the environmental feature vector, dividing all factory monitoring areas into multiple environmental categories, and determining the central monitoring area of ​​each environmental category; based on the central monitoring area, training a universal recognition model for each environmental category, and using a transfer learning method to adjust the universal recognition model to obtain a specific recognition model for each factory monitoring area in each environmental category; and deploying the specific recognition model to the video surveillance AI intelligent recognition equipment in each corresponding factory monitoring area. The present invention significantly improves the intelligence level and recognition accuracy of the factory video surveillance system, while optimizing resource allocation and reducing operating costs.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a data processing method and system for video surveillance AI intelligent recognition equipment. Background Art

[0002] With the continuous advancement of artificial intelligence (AI) technology, the field of video surveillance is undergoing a profound transformation. Traditional video surveillance methods rely heavily on manual, real-time review and analysis. This model is not only inefficient but also prone to missing critical information due to human factors (such as fatigue and negligence), severely undermining surveillance effectiveness. To address this challenge, the industry is actively seeking to incorporate AI technology, aiming to build automated, intelligent video surveillance and analysis systems to significantly improve surveillance efficiency.

[0003] The importance of video surveillance is particularly prominent in factory environments. Factories are typically divided into multiple surveillance zones, each facing unique surveillance requirements and challenges due to its distinct functions. For example, production lines emphasize real-time monitoring of product quality and production efficiency, while warehouses require precise tracking of material storage and inbound and outbound movements. Secure passageways ensure personnel safety, while office areas prioritize the protection of property and information security. These diverse needs make it difficult for traditional video surveillance methods to fully meet them, highlighting the necessity and urgency of technological upgrades.

[0004] Although AI-powered video surveillance recognition equipment has been put into use, achieving initial improvements in surveillance automation, it still faces two core issues. First, the equipment generally uses a unified recognition model to process video data from different surveillance areas, ignoring the significant differences in environmental characteristics between areas. This results in inconsistent recognition results and is difficult to achieve. Second, existing equipment lacks sufficient flexibility in resource allocation strategies and is unable to dynamically adjust server resource allocation based on the real-time needs of each surveillance area. This leads to problems such as wasted resources and inefficient task completion, which in turn affects the overall effectiveness of surveillance. Summary of the Invention

[0005] The purpose of the present invention is to provide a data processing method and system for video surveillance AI intelligent recognition equipment, which significantly improves the intelligence level and recognition accuracy of the factory video surveillance system, while optimizing resource allocation and reducing operating costs to solve at least one of the above-mentioned prior art problems.

[0006] In a first aspect, the present invention provides a data processing method for a video surveillance AI intelligent recognition device, the method specifically comprising:

[0007] Obtain real-time video data from video surveillance AI intelligent recognition devices in all factory monitoring areas, extract light intensity data, shooting angle parameters, and background complexity of each factory monitoring area based on the real-time video data, and form an environmental feature vector;

[0008] Based on the environmental feature vectors, all factory monitoring areas are divided into multiple environmental categories, and a central monitoring area is determined for each environmental category. The environmental categories include production line areas, storage areas, safety passages, industrial equipment areas, and office areas.

[0009] Based on the central monitoring area, a general recognition model is trained for each environmental category, and the general recognition model is adjusted using a transfer learning method to obtain a specific recognition model for each factory monitoring area in each environmental category;

[0010] Deploy the specific recognition model to the video surveillance AI intelligent recognition device of each corresponding factory monitoring area, and determine the recognition result of each factory monitoring area using the specific recognition model;

[0011] A confidence score of each specific recognition model is determined based on the recognition result, and each specific recognition model whose confidence score is lower than a preset confidence threshold is optimized through a reinforcement learning algorithm.

[0012] In a second aspect, the present invention provides a data processing system for a video surveillance AI intelligent recognition device, the system specifically comprising:

[0013] A first data processing module is used to obtain real-time video data from the video surveillance AI intelligent recognition equipment in all factory monitoring areas, extract the light intensity data, shooting angle parameters and background complexity of each factory monitoring area based on the real-time video data, and form an environmental feature vector;

[0014] a second data processing module, configured to divide all factory monitoring areas into a plurality of environmental categories based on the environmental feature vectors, and determine a central monitoring area for each environmental category, wherein the environmental categories include production line areas, storage areas, safety passages, industrial equipment areas, and office areas;

[0015] a third data processing module, configured to train a universal recognition model for each environmental category based on the central monitoring area, and to adjust the universal recognition model using a transfer learning method to obtain a specific recognition model for each plant monitoring area in each environmental category;

[0016] a fourth data processing module, configured to deploy the specific recognition model to the video surveillance AI intelligent recognition device of each corresponding factory monitoring area, and determine the recognition result of each factory monitoring area using the specific recognition model;

[0017] The fifth data processing module is used to determine the confidence score of each specific recognition model based on the recognition result, and optimize each specific recognition model whose confidence score is lower than a preset confidence threshold through a reinforcement learning algorithm.

[0018] In a third aspect, the present invention provides a computer device comprising: a memory and a processor and a computer program stored in the memory. When the computer program is executed on the processor, it implements the data processing method of the video surveillance AI intelligent recognition device as described in any one of the above methods.

[0019] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the data processing method of the video surveillance AI intelligent recognition device as described in any one of the above methods is implemented.

[0020] Compared with the prior art, the present invention has at least one of the following technical effects:

[0021] 1. The present invention significantly improves the intelligence level and recognition accuracy of the factory video surveillance system, while optimizing resource allocation and reducing operating costs.

[0022] 2. By extracting environmental feature vectors, the present invention can fully understand the actual environmental conditions of each monitored area, providing an accurate data basis for subsequent environmental classification and model training.

[0023] 3. The present invention realizes the reasonable classification of monitoring areas through cluster analysis algorithm, and at the same time determines the central monitoring area of ​​each category, providing representative data samples for subsequent model training.

[0024] 4. The present invention obtains a universal recognition model applicable to various environmental categories through data labeling and neural network model training in the central monitoring area, thereby improving the generalization ability of the model.

[0025] 5. The present invention adopts a transfer learning method to enable the model to be fine-tuned according to the specific needs of different monitoring areas, thereby improving the accuracy and adaptability of the model.

[0026] 6. Through confidence scoring and reinforcement learning algorithms, the present invention can automatically identify and optimize models with poor performance, and continuously improve the overall recognition capability of the monitoring system.

[0027] 7. The present invention comprehensively considers real-time demand and historical data, adopts genetic algorithm to realize dynamic optimization allocation of resources, improves resource utilization efficiency, and ensures efficient completion of tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0029] Figure 1 This is a flow chart of a data processing method for a video surveillance AI intelligent recognition device provided by one embodiment of the present invention;

[0030] Figure 2 This is a structural diagram of a data processing system for a video surveillance AI intelligent recognition device provided by one embodiment of the present invention;

[0031] Figure 3 It is a structural diagram of a computer device provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0032] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0033] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0034] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0035] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0036] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0037] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0038] In the embodiments of the present application, the execution subject of the process includes a terminal device, which includes but is not limited to: a server, a computer, a smart phone, a tablet computer, and other devices capable of executing the method disclosed in the present application. Figure 1 A flow chart of a data processing method for a video surveillance AI intelligent recognition device disclosed in one embodiment of the present invention is shown, and is described in detail as follows:

[0039] S101, obtain real-time video data of the video surveillance AI intelligent recognition equipment in all factory monitoring areas, extract the light intensity data, shooting angle parameters and background complexity of each factory monitoring area based on the real-time video data, and form an environmental feature vector.

[0040] In this embodiment, multiple video surveillance AI intelligent recognition devices are installed within the factory. These devices capture and transmit video data in real time to a central processing system. The central processing system receives and stores real-time video data from all monitored areas through specific interfaces and protocols. Using image processing technology, the central processing system performs frame-by-frame analysis of the received real-time video data. By extracting pixel brightness information from video frames and calculating the average of these pixel brightness values, the system can obtain the average light intensity value for any monitored area in the factory. This value reflects the lighting conditions in the monitored area and provides key data for the subsequent construction of environmental feature vectors. To obtain shooting angle parameters, the system further analyzes the changing characteristics of object edges within the video frames. By comparing the position and direction changes of object edges across frames, the system can infer the camera's shooting angle. This step is crucial for understanding the perspective relationship and object shape of the monitored images. The system uses a Gaussian mixture model algorithm to perform background modeling on the video data. This algorithm can separate foreground objects from the static background in the video data. By calculating the color distribution and texture complexity of the static background, the system can determine the background complexity of any monitored area in the factory. Background complexity is an important indicator of the environmental complexity of the monitored area and facilitates subsequent environmental feature vector construction and model training. The extracted light intensity data, shooting angle parameters, and background complexity are integrated to form environmental feature vectors. These feature vectors comprehensively reflect the environmental conditions of the factory's monitored area and provide foundational data for subsequent environmental classification and model training.

[0041] S102, based on the environmental feature vector, all factory monitoring areas are divided into multiple environmental categories, and a central monitoring area of ​​each environmental category is determined. The environmental categories include production line area, storage area, safety passage, industrial equipment area and office area.

[0042] In this embodiment, a suitable clustering algorithm (such as K-means clustering or DBSCAN clustering) is selected to perform cluster analysis on environmental feature vectors. Based on the clustering results, the factory monitoring area is divided into multiple environmental categories, such as production line areas, storage areas, safety corridors, industrial equipment areas, and office areas. Each clustering result is assigned a corresponding environmental category label. For each environmental category, the mean or median of the environmental feature vectors of all monitoring areas within that category is calculated as the central feature vector for that category. Based on the central feature vector, the monitoring area closest to the vector is located on a map or in the monitoring system and designated as the central monitoring area for that environmental category. The preliminarily determined central monitoring area is field-verified to ensure that it represents the overall characteristics of the environmental category. Based on the verification results, the central monitoring area is adjusted and optimized as necessary. By comparing monitoring data from different environmental categories, the accuracy and practicality of the division results are evaluated, and user feedback is collected to determine whether the selected central monitoring area meets actual needs. Based on the evaluation results and user feedback, the environmental category division and the selection of the central monitoring area are optimized. Clustering algorithms and center calculation methods are regularly updated and improved to adapt to changes in the factory environment and increased monitoring needs.

[0043] S103: Based on the central monitoring area, a general recognition model is trained for each environmental category, and the general recognition model is adjusted using a transfer learning method to obtain a specific recognition model for each factory monitoring area in each environmental category.

[0044] In this embodiment, a large amount of labeled image or video data is collected from the central monitoring area. This data should cover a variety of typical scenes and objects within that environment category. The collected data is preprocessed, including image enhancement, resizing, and normalization, to ensure data quality. A deep learning model suitable for that environment category is selected as the base model, such as a convolutional neural network (CNN) or recurrent neural network (RNN). The base model is trained using the preprocessed data to produce a general recognition model capable of recognizing most objects and scenes within that environment category. Data from the central monitoring area serves as the source domain data, while data from other plant monitoring areas serves as the target domain data. Transfer learning techniques are used to transfer knowledge from the source domain data to the target domain data. Based on the general recognition model, fine-tuning is performed on the data from each plant monitoring area, including adjusting model parameters such as weights and learning rate to accommodate the data distribution and characteristics of each monitoring area. Through iterative training and optimization, a specific recognition model is developed for each plant monitoring area. These models maintain general recognition capabilities while better adapting to the characteristics of their respective monitoring areas. Use metrics such as accuracy, recall, and F1 score to evaluate general and specific recognition models. Compare the performance of different models across different monitoring areas, analyze their strengths and weaknesses, and identify areas for improvement. Based on the evaluation results, optimize the models. This may include increasing the amount of data, improving data preprocessing methods, and adjusting the model structure. Regularly update and maintain the models to adapt to changes in the factory environment and increased monitoring needs.

[0045] In this example, by training a general recognition model and fine-tuning it using transfer learning, we can generate specific recognition models for different surveillance areas. These models can more accurately identify objects and scenes in their respective surveillance areas. Transfer learning allows us to fine-tune existing models instead of training them from scratch, which significantly shortens model training time and improves development efficiency.

[0046] S104: deploy the specific recognition model to the video surveillance AI intelligent recognition device of each corresponding factory monitoring area, and determine the recognition result of each factory monitoring area through the specific recognition model.

[0047] In this embodiment, a trained specific recognition model is deployed to the video surveillance AI intelligent recognition device in the corresponding factory monitoring area. The video surveillance AI intelligent recognition device captures video data from the factory monitoring area in real time and inputs the captured video data into the specific recognition model deployed on the device. The specific recognition model processes and analyzes the input video data, identifying objects, scenes, or behaviors in the video. The model outputs the recognition results based on preset recognition rules and thresholds. The recognition results are output to the device display interface or a designated storage location in an appropriate format (such as text, image, or sound). Based on the recognition results, corresponding alarms, linkages, or other application scenarios are triggered.

[0048] In this embodiment, by deploying specific recognition models, video surveillance AI intelligent recognition equipment can more accurately identify objects, scenes, or behaviors within the factory monitoring area, helping to reduce false positives and missed positives and improving the accuracy and reliability of the monitoring system. By deploying specific recognition models, video surveillance AI intelligent recognition equipment can implement automated recognition and alarm functions, helping to enhance the factory's intelligence level and reduce the cost and risk of manual intervention.

[0049] S105 , determining a confidence score of each specific recognition model according to the recognition result, and optimizing each specific recognition model whose confidence score is lower than a preset confidence threshold through a reinforcement learning algorithm.

[0050] In this embodiment, an appropriate confidence evaluation metric is selected based on the application scenario and requirements of a specific recognition model. For example, in an object recognition task, the accuracy, recall rate, or F1 score of the recognition results can be selected as a confidence evaluation metric. The recognition results of each specific recognition model are evaluated, and a score is calculated based on the selected confidence evaluation metric. The score is normalized to a confidence score between 0 and 1 to facilitate comparison and analysis. Based on actual needs and application scenarios, a preset confidence threshold is set to determine whether the confidence level of a specific recognition model meets the requirements. A reinforcement learning algorithm suitable for the specific recognition model optimization task is selected, such as Q-learning, Deep Q Network (DQN), or Policy Gradient Method. The specific recognition model with a confidence score below the preset confidence threshold is selected as the model to be optimized. The optimization task is defined as improving the confidence score of the model to be optimized so that it reaches or exceeds the preset confidence threshold. The recognition results of the model to be optimized and the corresponding real-world labeled data are collected as training data for reinforcement learning. A reward function is designed to provide rewards based on the confidence score of the model to be optimized and the accuracy of the recognition results. Use a reinforcement learning algorithm to train the model to be optimized. Through trial and error and adjustment of model parameters, improve the model's confidence score and recognition accuracy. During training, regularly verify the performance of the model to be optimized to ensure that the model improves confidence scores without excessively sacrificing recognition accuracy. Based on the verification results, make necessary adjustments and optimizations to the reinforcement learning algorithm and model parameters.

[0051] In this embodiment, a reinforcement learning algorithm is used to optimize specific recognition models whose confidence scores fall below a preset confidence threshold. This effectively improves the model's confidence score, helps reduce false positives and false negatives, and enhances the accuracy and reliability of the monitoring system. Through continuous trial and error and adjustment of model parameters, the reinforcement learning algorithm enables the model to better adapt to different monitoring scenarios and data distributions, enhancing its generalization capabilities and improving its performance in practical applications.

[0052] In some embodiments, in step S101, extracting the illumination intensity data, shooting angle parameters, and background complexity of each factory monitoring area based on the real-time video data specifically includes:

[0053] Extracting pixel brightness information from the real-time video data, and obtaining an average light intensity value of any factory monitoring area by calculating an average value of the pixel brightness information;

[0054] Obtaining shooting angle parameters of any factory monitoring area by analyzing the change characteristics of object edge lines in the real-time video data;

[0055] A Gaussian mixture model algorithm is used to perform background modeling on the real-time video data to separate the foreground target and static background in the real-time video data. The background complexity of any factory monitoring area is obtained by calculating the color distribution and texture complexity of the static background.

[0056] In this embodiment, pixel brightness information of each frame image in the video data is extracted; the average value of the pixel brightness information of each frame image is calculated to obtain the average light intensity value corresponding to the frame image; if the difference between the average light intensity value of the current frame and the average light intensity value of the previous frame is greater than a preset light intensity change threshold, it is determined that the lighting environment of the factory monitoring area has changed; based on the change trend of the average light intensity values ​​of multiple consecutive frames of images, the Kalman filter algorithm is used to filter the average light intensity value to remove noise interference in the light intensity data; by performing statistical analysis on the average light intensity value after filtering, the average light intensity level of the factory monitoring area over a period of time is obtained.

[0057] Perform edge detection on objects in the video to obtain edge line information of the objects; extract the change characteristics of the edge lines between different frames, and determine the shooting angle parameters of the current monitoring area based on the pre-established correspondence between the edge line change characteristics and the shooting angle.

[0058] The video data is transmitted to the background modeling module for processing. In the background modeling module, the Gaussian mixture model algorithm is used to model the real-time video data, and the probability of each pixel belonging to the foreground or background is determined by the method of probability statistics. According to the modeling results of the Gaussian mixture model algorithm, the pixels in the video data are divided into two parts: the foreground target and the static background, and the separation results of the foreground target and the static background are obtained. For the separated static background, its color distribution characteristics are calculated, and the frequency and proportion of different colors are counted to obtain the color distribution feature vector of the static background. For the separated static background, a texture feature extraction algorithm, such as the gray level co-occurrence matrix or the local binary pattern, is used to extract the texture complexity characteristics of the static background. The color distribution feature vector and the texture complexity characteristics of the static background are fused to construct a background complexity evaluation model, and the background complexity value of the factory monitoring area is obtained through model calculation.

[0059] For example, extracting pixel brightness information can help assess lighting conditions in various factory areas. For example, in the furnace area of ​​a large steel plant, calculating the average pixel brightness can yield the average light intensity for that area. Analyzing the changing characteristics of object edges can yield shooting angle parameters, which have multiple applications in factory monitoring. For example, in an automobile assembly line, analyzing the changing edges of parts on a conveyor belt can infer the camera's installation angle. Gaussian mixture models are widely used in background modeling. In a busy logistics warehouse, this algorithm can effectively separate foreground objects such as moving workers and forklifts from static background objects such as shelves and aisles. By calculating the color distribution and texture complexity of the static background, the background complexity of that area can be determined. In a chemical plant's reactor area, changes in light intensity can indicate different stages of the reaction process. By continuously monitoring the average light intensity, operators can remotely determine the reaction status and adjust parameters in a timely manner. Furthermore, obtaining shooting angle parameters can help optimize camera layout, ensuring clear capture of the reactor's interior without being affected by equipment obstruction or light reflections. On the packaging lines of food processing plants, background complexity calculations can help identify anomalies. For example, if the background complexity in a certain area suddenly increases, it may indicate a packaging material leak or product accumulation. The system can promptly issue an alarm, which not only improves production efficiency but also reduces the risk of raw material waste and quality issues. Through the combined application of these technologies, factory monitoring systems can operate more intelligently and accurately. Light intensity monitoring can ensure that lighting conditions in precision component assembly areas are always optimal; obtaining shooting angle parameters can optimize the visual system in the quality inspection process and improve the accuracy of defect detection; and background complexity analysis can help identify abnormal accumulation or human intrusion on the production line, thereby improving overall production efficiency and safety. By analyzing video data in real time, factory managers can gain more comprehensive and in-depth insights into the production process, allowing them to make more informed decisions, optimize resource allocation, improve production efficiency, and ensure worker safety and product quality.

[0060] In some embodiments, in step S102, dividing all factory monitoring areas into multiple environmental categories based on the environmental feature vector and determining the central monitoring area of ​​each environmental category specifically includes:

[0061] Based on the distance measurement between the environmental feature vectors of each factory monitoring area, a cluster analysis algorithm is used to cluster all industrial monitoring areas, and industrial monitoring areas with similar distances and characteristics are divided into the same environmental category;

[0062] For each environmental category, calculating the mean vector of the environmental feature vectors of all industrial monitoring areas within the environmental category, and determining the mean vector as the central feature vector of the environmental category;

[0063] In each environment category, the industrial monitoring area closest to the central feature vector is determined as the central monitoring area.

[0064] In this embodiment, a K-means clustering algorithm is used to cluster all industrial monitoring areas based on the distance metric between environmental feature vectors. Industrial monitoring areas with similar characteristics and close proximity are classified into the same environmental category. For each environmental category, the mean vector of the environmental feature vectors of all industrial monitoring areas within that category is calculated, and the mean vector is determined as the central feature vector of that environmental category. Within each environmental category, the distance between the environmental feature vectors of each industrial monitoring area and the central feature vector of that category is calculated, and the industrial monitoring area with the closest distance is determined as the central monitoring area of ​​that environmental category. Real-time environmental monitoring data for each central monitoring area is obtained, environmental feature parameters are extracted, and the central feature vector of the corresponding environmental category is updated. The distance between the real-time environmental feature vector of each industrial monitoring area and the central feature vector of the environmental category to which it belongs is determined. If the distance exceeds a preset threshold, an early warning message is triggered, indicating that the environment in the monitoring area is abnormal. Based on the historical environmental monitoring data of each industrial monitoring area, a support vector machine algorithm is used to establish a mapping model between environmental categories and environmental risk levels, and the environmental risk level of each environmental category is assessed in real time. Obtain high-risk environmental categories, adjust environmental parameter threshold settings for each industrial monitoring area, increase monitoring frequency, improve early warning sensitivity, and simultaneously control related environmental conditioning equipment to adjust and optimize environmental parameters and reduce environmental risks.

[0065] For example, consider a large, integrated manufacturing plant encompassing diverse workshops such as electronic component production, machining, and chemical processing. First, an environmental feature vector must be established for each monitored area. This vector may include data on multiple dimensions, such as light intensity, temperature, humidity, noise level, and air quality. For the electronic component production workshop, the environmental feature vector may exhibit low noise, constant temperature and humidity, and high cleanliness. In contrast, the feature vector for the machining workshop may show high noise, large temperature fluctuations, and high metal dust levels. The chemical processing workshop may exhibit high concentrations of special gases, high temperatures, and strong corrosiveness. Next, cluster analysis algorithms, such as K-means or hierarchical clustering, are used to classify these feature vectors. Euclidean or Manhattan distance can be used as the distance metric. This method may yield clustering results such as: precision manufacturing environments (including electronic component production and precision instrument assembly), heavy industrial environments (including machining and metal smelting), and chemical processing environments. For each environmental category, the central feature vector is calculated. For example, the central feature vector for the precision manufacturing environment category might show an average temperature of 22°C, relative humidity of 45%, a noise level of 45 decibels, and a dust concentration of less than 0.1 μg / m³. This central feature vector represents the typical environmental characteristics of this category and can be used as a reference standard for that category. After determining the central feature vector, the actual monitoring area that most closely matches it can be identified and designated as the central monitoring area. This area serves as a representative of the environmental category, allowing for a quick assessment of the overall status of the entire category. For example, within the precision manufacturing environment category, the integrated circuit packaging workshop, with its highest cleanliness and most stable temperature and humidity control, might be selected as the central monitoring area. This clustering and central area identification method has multiple implications. First, it helps managers quickly understand the overall status of different environmental types within the factory without having to review all areas individually. Second, it provides a benchmark for anomaly detection: if the environmental characteristics of an area differ significantly from the central feature vector of its corresponding category, it may indicate an anomaly requiring further investigation. Finally, this method can guide resource allocation: equipment maintenance and upgrades in the central monitoring areas can be prioritized, as the status of these areas has the greatest impact on the monitoring effectiveness of the entire category. Through this systematic environmental analysis and monitoring method, factories can manage complex production environments more efficiently and improve production efficiency, while ensuring that environmental parameters in each area are always maintained at optimal levels, thereby ensuring product quality and employee safety.

[0066] In some embodiments, in step S103, training a general recognition model for each environment category based on the central monitoring area specifically includes:

[0067] Determine a monitoring recognition task for each central monitoring area, and annotate the first video data of the central monitoring area according to the monitoring recognition task to form a monitoring recognition data set. The monitoring recognition tasks include product defect recognition, worker operation behavior recognition, cargo classification and location recognition, industrial equipment anomaly recognition, and outsider recognition;

[0068] The monitoring identification data set is input into a neural network model for modeling training to obtain a universal identification model of the environmental category corresponding to each central monitoring area.

[0069] In this embodiment, first video data from each central monitoring area is acquired and annotated according to preset monitoring recognition tasks to generate a monitoring recognition dataset. These tasks include product defect recognition, worker operation behavior recognition, cargo classification and location recognition, industrial equipment anomaly recognition, and outsider recognition. The monitoring recognition dataset is input into a neural network model, which is then trained using a supervised learning algorithm to generate a pre-trained neural network model. Environmental category information for the central monitoring area is acquired, and the pre-trained neural network model is fine-tuned based on this information to generate a general recognition model for the corresponding environmental category in each central monitoring area. For each central monitoring area, real-time video data is collected as test data and input into the corresponding general recognition model for predictions to determine whether anomalies such as product defects, improper worker operation, incorrect cargo placement, abnormal industrial equipment operation, and the presence of outsiders exist. If the general recognition model predicts an anomaly, an early warning signal is triggered, prompting relevant personnel through a human-computer interface to promptly address the anomaly. The processing results of the anomaly are acquired and added to the monitoring recognition dataset along with the corresponding test data to form new training data. Use new training data to retrain the general recognition model, continuously improve the model's generalization ability and recognition accuracy, and achieve adaptive optimization of monitoring and recognition tasks.

[0070] For example, in a precision manufacturing environment, the central monitoring area might be an integrated circuit packaging workshop. The main monitoring and recognition tasks in this workshop include product defect recognition, worker behavior recognition, and outsider identification. Product defect recognition involves detecting various defects that may occur during the packaging process, such as poor soldering and package cracks. To build an effective recognition model, a large number of packaged products must be captured and annotated. During the annotation process, professionals carefully review each video frame, marking the location, type, and severity of the defect. This meticulous annotation process provides a high-quality data foundation for subsequent model training. Worker behavior recognition aims to ensure that workers adhere to standard operating procedures and improve production efficiency and safety. During the annotation process, various worker actions, such as material removal, placement, and debugging, must be classified and labeled. This includes not only correct operating behaviors but also common errors, allowing the model to identify and issue timely warnings. Outsider identification is crucial for maintaining safety and confidentiality in the workshop. The annotation process requires distinguishing between normal workers and outsiders, including their clothing characteristics and behavioral patterns. This helps the system quickly identify unauthorized access to sensitive areas. By training these labeled datasets into a neural network model, a universal recognition model suitable for precision manufacturing environments can be developed. This model can handle multiple recognition tasks simultaneously, improving the efficiency and flexibility of the monitoring system. For example, a convolutional neural network (CNN) combined with an object detection algorithm can effectively identify product defects, while a recurrent neural network (RNN) is more suitable for analyzing worker action sequences. Models trained using this approach have strong generalization capabilities and can be applied not only to the central monitoring area but also to other areas within the same environmental category. This significantly improves the model's practicality and cost-effectiveness. For example, a model trained in an electronic component production workshop can be well-suited for precision instrument assembly workshops, as these two environments belong to the same environmental category and share similar environmental characteristics and monitoring requirements. This universal recognition model approach based on environmental categories not only enhances the intelligence of monitoring systems but also provides a powerful tool for factory management. It can promptly detect product quality issues, standardize worker operations, and ensure workshop safety, thereby comprehensively improving production efficiency and product quality. Furthermore, this approach is highly scalable. With the continuous accumulation of data and continuous model optimization, its recognition accuracy and applicability will continue to improve, providing strong technical support for the intelligent transformation of factories.

[0071] In some embodiments, in step S103, the general recognition model is adjusted using a transfer learning method to obtain a specific recognition model for each factory monitoring area in each environmental category, specifically including:

[0072] Determine the attention weight distribution parameters for each factory monitoring area in different monitoring and recognition tasks;

[0073] Based on the attention weight distribution parameters, a transfer learning method is used to adjust the general recognition model of the environmental category corresponding to each factory monitoring area to obtain a specific recognition model for each factory monitoring area.

[0074] In this embodiment, surveillance video data from each factory monitoring area is acquired, preprocessed, and key frames and regions of interest are extracted. Corresponding attention weight distribution parameters are determined for different monitoring and recognition tasks in each factory monitoring area, such as personnel recognition and vehicle recognition. Based on the attention weight distribution parameters, feature extraction is performed on the preprocessed video data to obtain feature vectors representing the different monitoring and recognition tasks. A transfer learning approach is employed, based on a pre-trained general recognition model. The feature vectors are input into the model for fine-tuning training to obtain a recognition model adapted to the specific environment of each factory monitoring area. The adjusted specific recognition model is used to process the real-time video stream of each factory monitoring area to achieve real-time recognition and tracking of different monitored objects. Based on the recognition results, a determination is made as to whether an abnormality has occurred. If an abnormality has occurred, an alarm mechanism is triggered to notify relevant personnel to handle the situation. The specific recognition model for each factory monitoring area is continuously optimized, and newly collected data is added to the training set through incremental learning to continuously improve the model's generalization ability and recognition accuracy.

[0075] For example, in a precision manufacturing environment, an integrated circuit packaging workshop may face multiple tasks, including product defect recognition, worker behavior recognition, and outsider identification. Depending on the specific circumstances of the workshop, these tasks can be assigned different weights. Assuming that the workshop has significant product quality issues, the weights for product defect recognition can be set to 0.5, worker behavior recognition to 0.3, and outsider identification to 0.2. This weight distribution reflects the workshop's current primary focus and helps the model allocate computing resources more efficiently. After determining the parameters for the attention weight distribution, transfer learning can be used to adapt the general recognition model to the needs of a specific monitoring area. The core idea of ​​transfer learning is to transfer knowledge learned from one task to another related task, thereby reducing the need for large amounts of labeled data for the new task. In this example, we can start with a pretrained general recognition model that has been trained on a large amount of precision manufacturing environment data. We can then fine-tune the model using a small amount of labeled data specific to the workshop. Specifically, we can adopt a fine-tuning strategy that leaves most of the pretrained model's parameters unchanged and adjusts only the last few layers. This approach preserves the model's understanding of general features while adapting to the unique characteristics of a specific workshop. For example, for product defect identification tasks, a workshop-specific product sample can be used for fine-tuning, enabling the model to identify defect types unique to that workshop. During fine-tuning, an attention mechanism can also be introduced to adjust the model's focus on different tasks based on previously determined weight distribution parameters. For example, for product defect identification tasks with higher weights, the model's attention can be enhanced in the relevant feature extraction layer, making it more sensitive to subtle changes in the product surface. This approach yields a model that retains general recognition capabilities while adapting to the needs of specific monitoring areas. This specialized recognition model can more accurately identify product defects in that workshop and better understand worker behavior, while also not neglecting the safety-related task of identifying outsiders. This approach, based on attention weights and transfer learning, offers significant technical advantages. First, it significantly reduces the need for large amounts of annotated data, lowering the cost and time of model training. Second, by adjusting the attention weights, the model can quickly adapt to the workshop's priority tasks at different times, increasing the system's flexibility. Finally, this approach ensures that the model maintains general versatility while accurately responding to specific environments, improving overall monitoring efficiency.

[0076] In some embodiments, in step S105, determining the confidence score of each specific recognition model based on the recognition result, and optimizing each specific recognition model whose confidence score is lower than a preset confidence threshold by a reinforcement learning algorithm specifically includes:

[0077] Based on the recognition results, determining the accuracy of each specific recognition model on different monitoring recognition tasks;

[0078] Determining a confidence score for each specific recognition model based on the accuracy rates on different monitoring recognition tasks and the attention weight distribution parameters;

[0079] Comparing the confidence score of each specific recognition model with a preset confidence threshold, and determining a number of first specific recognition models having confidence scores lower than the preset confidence threshold score;

[0080] The model parameter adjustment method of the first specific recognition model is used as the action space, the model parameter information of the first specific recognition model is used as the state space, and the confidence score of the first specific recognition model is used as the reward function;

[0081] Based on the action space, the state space and the reward function, each first specific recognition model is optimized using a Q-Learning algorithm.

[0082] In this embodiment, the recognition results of each specific recognition model on different monitoring recognition tasks are obtained; for each specific recognition model's recognition results, its accuracy on each monitoring recognition task is calculated; a pre-established attention weight distribution parameter is obtained; the accuracy of each specific recognition model on different monitoring recognition tasks is weighted and summed with the corresponding attention weight distribution parameter to obtain a comprehensive performance indicator of the model, namely, a confidence score. The confidence score is compared with a preset confidence threshold to determine whether the confidence score is lower than the preset confidence threshold. If the confidence score is lower than the preset confidence threshold, the specific recognition model is determined as a first specific recognition model. For each first specific recognition model, its model parameter adjustment method is obtained, and the model parameter adjustment method is used as the action space of reinforcement learning. For each first specific recognition model, its model parameter information is obtained, and the model parameter information is used as the state space of reinforcement learning. For each first specific recognition model, its confidence score is obtained, and the confidence score is used as the reward function of reinforcement learning. Based on the obtained action space, state space, and reward function, the Q-Learning algorithm is used to optimize each first specific recognition model to obtain an optimized first specific recognition model. For each optimized first specific recognition model, its confidence score is obtained, and the confidence score of the optimized first specific recognition model is compared with a preset confidence threshold to determine whether the confidence score of the optimized first specific recognition model is higher than the preset confidence threshold. If the confidence score of the optimized first specific recognition model is higher than the preset confidence threshold, the optimized first specific recognition model is determined as the target specific recognition model. All target specific recognition models are obtained, all target specific recognition models are combined into an optimized specific recognition model group, and the optimized specific recognition model group is output.

[0083] For example, in an integrated circuit packaging workshop, assume that accuracy rates of 85%, 78%, and 92% were achieved for the three tasks of product defect recognition, worker behavior recognition, and outsider recognition, respectively. These accuracy rates reflect the model's performance on each task and provide a basis for subsequent optimization. Combining the accuracy rate and the attention weight distribution parameters, a confidence score for each specific recognition model can be calculated. For example, if the weight for product defect recognition is 0.5, for worker behavior recognition is 0.3, and for outsider recognition is 0.2, the overall confidence score for the model can be calculated by weighted average. Assume the calculated confidence score is 83%. Comparing the calculated confidence score with a preset threshold can identify the model requiring optimization. If the preset threshold is 85%, the model is marked as the first specific recognition model requiring optimization. This approach effectively screens out underperforming models and focuses optimization resources. For models requiring optimization, the Q-Learning algorithm from reinforcement learning can be used for optimization. In this process, the model parameter adjustment methods are defined as an action space, which can include adjusting the learning rate and changing the network structure. Model parameter information constitutes a state space, reflecting the model's current state. The confidence score serves as a reward function, guiding optimization. In practice, a series of discrete parameter adjustments can be designed, such as adjusting the learning rate to 0.8 times or 1.2 times the current value. The state space can include key parameters such as the current number of network layers and the number of neurons per layer. The reward function can be designed as the difference between the old and new confidence scores, encouraging adjustments that improve model performance. Through multiple iterations, the Q-Learning algorithm gradually finds the optimal parameter adjustment strategy. For example, the algorithm may discover that increasing the number of convolutional layers significantly improves product defect recognition accuracy, thereby increasing the overall confidence score. The advantage of this approach is that it automatically explores a large number of possible optimization paths to identify the most effective strategy. This reinforcement learning-based optimization method not only improves model performance but also adapts to the dynamically changing factory environment. As production conditions change, the model can continuously optimize and maintain high recognition capabilities. Furthermore, this approach provides factory managers with valuable insights into the key factors influencing model performance, enabling broader improvements to production processes and quality control.

[0084] In some embodiments, in steps S101 to S105 above, the method further includes:

[0085] Obtain real-time video traffic, model inference demand information, and historical resource usage data for each factory monitoring area;

[0086] With the optimization goals of minimizing resource utilization and maximizing task completion rate, a fitness function is set in combination with the real-time video traffic, the model reasoning requirement information, and the historical resource usage data;

[0087] A genetic algorithm is used to randomly generate a set of initial resource allocation schemes for each factory monitoring area. Based on the fitness function, the initial resource allocation schemes are subjected to selection, crossover and mutation operations until the convergence condition of the fitness function or the preset number of iterations is reached, and a target resource allocation scheme is generated. The target resource allocation scheme includes the allocation ratio of computing resources and the allocation ratio of storage resources.

[0088] In this embodiment, real-time video traffic data is obtained for each plant monitoring area. The required computing and storage resources are determined based on the video traffic volume and model inference requirements. Historical resource usage data is obtained for each plant monitoring area. By analyzing this historical data, the relationship between resource utilization and task completion rate is determined. A fitness function is constructed with the optimization goal of minimizing resource utilization and maximizing task completion rate. A genetic algorithm is used to randomly generate a set of initial resource allocation plans for each plant monitoring area. Each plan includes the allocation ratio of computing and storage resources. Based on the fitness function, the initial resource allocation plans are evaluated and a fitness value is calculated for each plan. A higher fitness value indicates a better plan. A selection operation is performed on the initial resource allocation plans, with plans with higher fitness values ​​being more likely to be selected. A crossover operation is performed to exchange some genes of the selected high-quality plans to generate new resource allocation plans. A mutation operation is performed on the newly generated resource allocation plans, introducing random perturbations to the plan's genes to maintain population diversity and prevent the algorithm from falling into a local optimum. Determine whether the convergence condition of the fitness function or the preset number of iterations has been met. If not, return to step 4 to continue iterative optimization. If so, output the resource allocation plan with the highest fitness value as the target plan, including the computing resource and storage resource allocation ratios for each factory monitoring area.

[0089] For example, first, obtaining real-time video traffic, model inference requirements, and historical resource usage data for each monitoring area forms the foundation for optimization. For example, in a large automobile manufacturing plant, the assembly line, paint shop, and quality inspection area may have different monitoring requirements. The assembly line may require high-frame-rate video to capture fast-moving parts, while the quality inspection area requires high-resolution images to detect subtle defects. Setting the fitness function is key to the optimization process. This function should comprehensively consider resource utilization and task completion rate. In practical applications, resource utilization can be defined as the ratio of allocated resources to total available resources, while task completion rate can be the ratio of the number of successfully processed video frames to the total number of frames. The fitness function can be a weighted sum of these two metrics, with the weights adjustable based on the specific needs of the factory. The application of genetic algorithms makes optimizing resource allocation schemes efficient and feasible. The generation of initial schemes can be based on historical data or completely random. For example, each zone can be randomly allocated 10% to 50% of available computing and storage resources. The selection process can use a roulette wheel method to increase the probability of selecting a scheme with high fitness. The crossover operation swaps resource allocation ratios between different scenarios, for example, swapping the compute resource allocation ratio for the assembly line in Plan A with the corresponding ratio in Plan B. The mutation operation randomly adjusts resource allocation within a specific area, such as increasing or decreasing the storage resources in the paint shop by 5%. This optimization approach offers the advantage of adapting to complex and changing factory environments. For example, when a factory introduces a new production line or upgrades equipment, the system can quickly adjust resource allocation to meet the new demand. Furthermore, by analyzing the optimization results, factory managers can gain insight into resource demand patterns across different areas, providing a basis for future equipment investment and system upgrades. In practice, certain areas may experience sudden high demand during specific periods. For example, the quality inspection area may require additional computing resources to handle a large number of inspection tasks during a production batch change. By incorporating the time dimension into the optimization model, the system can predict and proactively adjust resource allocation to ensure good performance even during peak periods. The resulting target resource allocation plan not only includes static allocation ratios but also dynamic adjustment strategies. For example, if a sudden increase in video traffic on an assembly line is detected, the system can automatically redeploy resources from other temporarily unused areas. This flexible resource allocation mechanism further improves the efficiency and reliability of the entire monitoring system. Through this genetic algorithm-based optimization method, factories can maximize the effectiveness of the monitoring system under limited resource conditions and improve production efficiency and product quality.

[0090] Reference Figure 2 An embodiment of the present invention provides a data processing system 2 for a video surveillance AI intelligent recognition device, the system 2 specifically comprising:

[0091] The first data processing module 201 is used to obtain real-time video data from the video surveillance AI intelligent recognition equipment in all factory monitoring areas, extract the light intensity data, shooting angle parameters and background complexity of each factory monitoring area based on the real-time video data, and form an environmental feature vector;

[0092] A second data processing module 202 is configured to divide all factory monitoring areas into multiple environmental categories based on the environmental feature vectors, and determine a central monitoring area for each environmental category, wherein the environmental categories include production line areas, storage areas, safety passages, industrial equipment areas, and office areas;

[0093] The third data processing module 203 is used to train a universal recognition model for each environmental category based on the central monitoring area, and adjust the universal recognition model using a transfer learning method to obtain a specific recognition model for each factory monitoring area in each environmental category;

[0094] A fourth data processing module 204 is configured to deploy the specific recognition model to the video surveillance AI intelligent recognition device of each corresponding factory monitoring area, and determine the recognition result of each factory monitoring area using the specific recognition model;

[0095] The fifth data processing module 205 is used to determine the confidence score of each specific recognition model based on the recognition result, and optimize each specific recognition model whose confidence score is lower than a preset confidence threshold through a reinforcement learning algorithm.

[0096] It is understandable that if Figure 1 The contents of the data processing method embodiment of the video surveillance AI intelligent recognition device shown in the figure are applicable to the data processing system embodiment of the video surveillance AI intelligent recognition device. The functions specifically implemented by the data processing system embodiment of the video surveillance AI intelligent recognition device are similar to those in the embodiment shown in the figure. Figure 1 The data processing method embodiment of the video surveillance AI intelligent recognition device shown is the same as that of the embodiment shown in FIG. Figure 1 The beneficial effects achieved by the data processing method embodiment of the video surveillance AI intelligent recognition device shown are also the same.

[0097] It should be noted that the information interaction, execution process and other contents between the above-mentioned systems are based on the same concept as the embodiment of the method of the present invention. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0098] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0099] Reference Figure 3 , an embodiment of the present invention further provides a computer device 3, comprising: a memory 302 and a processor 301 and a computer program 303 stored in the memory 302. When the computer program 303 is executed on the processor 301, the data processing method of the video surveillance AI intelligent recognition device as described in any one of the above methods is implemented.

[0100] The computer device 3 may be a desktop computer, a notebook computer, a PDA, a cloud server or other computing devices. The computer device 3 may include, but is not limited to, a processor 301 and a memory 302. Those skilled in the art will understand that Figure 3 This is merely an example of the computer device 3 and does not constitute a limitation on the computer device 3 . The computer device 3 may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the computer device 3 may also include input and output devices, network access devices, etc.

[0101] The processor 301 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor may be a microprocessor or any conventional processor.

[0102] In some embodiments, the memory 302 may be an internal storage unit of the computer device 3, such as a hard drive or memory of the computer device 3. In other embodiments, the memory 302 may also be an external storage device of the computer device 3, such as a plug-in hard drive, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. equipped on the computer device 3. Furthermore, the memory 302 may include both an internal storage unit of the computer device 3 and an external storage device. The memory 302 is used to store an operating system, application programs, a boot loader, data, and other programs, such as the program code of the computer program. The memory 302 may also be used to temporarily store data that has been output or is about to be output.

[0103] An embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the data processing method of the video surveillance AI intelligent recognition device as described in any of the above methods is implemented.

[0104] In this embodiment, if the integrated unit is implemented as a software functional unit and sold or used as a standalone product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the process steps in the above-mentioned method embodiments by using a computer program to instruct the relevant hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a camera / terminal device, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signals, telecommunication signals, and software distribution media. Examples include USB flash drives, removable hard drives, magnetic disks, or optical disks. In some jurisdictions, based on legislation and patent practice, computer-readable media cannot be electric carrier signals or telecommunication signals.

[0105] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0106] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0107] In the embodiments disclosed in the present application, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are merely schematic. For example, the division of the modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0108] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

Claims

1. A data processing method for video surveillance AI intelligent recognition equipment, characterized in that: The method specifically includes: Obtain real-time video data from video surveillance AI intelligent recognition devices in all factory monitoring areas, extract light intensity data, shooting angle parameters, and background complexity of each factory monitoring area based on the real-time video data, and form an environmental feature vector; Based on the environmental feature vectors, all factory monitoring areas are divided into multiple environmental categories, and a central monitoring area is determined for each environmental category. The environmental categories include production line areas, storage areas, safety passages, industrial equipment areas, and office areas. Based on the central monitoring area, a general recognition model is trained for each environmental category, and the general recognition model is adjusted using a transfer learning method to obtain a specific recognition model for each factory monitoring area in each environmental category; Deploy the specific recognition model to the video surveillance AI intelligent recognition device of each corresponding factory monitoring area, and determine the recognition result of each factory monitoring area using the specific recognition model; A confidence score of each specific recognition model is determined based on the recognition result, and each specific recognition model whose confidence score is lower than a preset confidence threshold is optimized through a reinforcement learning algorithm.

2. The method according to claim 1, characterized in that Extracting the illumination intensity data, shooting angle parameters, and background complexity of each factory monitoring area based on the real-time video data specifically includes: Extracting pixel brightness information from the real-time video data, and obtaining an average light intensity value of any factory monitoring area by calculating an average value of the pixel brightness information; Obtaining shooting angle parameters of any factory monitoring area by analyzing the change characteristics of object edge lines in the real-time video data; A Gaussian mixture model algorithm is used to perform background modeling on the real-time video data to separate the foreground target and static background in the real-time video data. The background complexity of any factory monitoring area is obtained by calculating the color distribution and texture complexity of the static background.

3. The method according to claim 1, characterized in that Based on the environmental feature vector, all factory monitoring areas are divided into multiple environmental categories, and the central monitoring area of ​​each environmental category is determined, specifically including: Based on the distance measurement between the environmental feature vectors of each factory monitoring area, a cluster analysis algorithm is used to cluster all factory monitoring areas, and factory monitoring areas with similar distances and characteristics are divided into the same environmental category; For each environmental category, calculating the mean vector of the environmental feature vectors of all factory monitoring areas within the environmental category, and determining the mean vector as the central feature vector of the environmental category; In each environment category, the plant monitoring area closest to the central feature vector is determined as the central monitoring area.

4. The method according to claim 1, wherein The method of training a general recognition model for each environment category based on the central monitoring area specifically includes: Determine a monitoring recognition task for each central monitoring area, and annotate the first video data of the central monitoring area according to the monitoring recognition task to form a monitoring recognition data set. The monitoring recognition tasks include product defect recognition, worker operation behavior recognition, cargo classification and location recognition, industrial equipment anomaly recognition, and outsider recognition; The monitoring identification data set is input into a neural network model for modeling training to obtain a universal identification model of the environmental category corresponding to each central monitoring area.

5. The method according to claim 4, characterized in that The transfer learning method is used to adjust the general recognition model to obtain a specific recognition model for each factory monitoring area in each environmental category, specifically including: Determine the attention weight distribution parameters for each factory monitoring area in different monitoring and recognition tasks; Based on the attention weight distribution parameters, a transfer learning method is used to adjust the general recognition model of the environmental category corresponding to each factory monitoring area to obtain a specific recognition model for each factory monitoring area.

6. The method according to claim 5, characterized in that Determining the confidence score of each specific recognition model based on the recognition result, and optimizing each specific recognition model whose confidence score is lower than a preset confidence threshold through a reinforcement learning algorithm, specifically includes: Based on the recognition results, determining the accuracy of each specific recognition model on different monitoring recognition tasks; Determining a confidence score for each specific recognition model based on the accuracy rates on different monitoring recognition tasks and the attention weight distribution parameters; Comparing the confidence score of each specific recognition model with a preset confidence threshold, and determining a number of first specific recognition models having confidence scores lower than the preset confidence threshold score; The model parameter adjustment method of the first specific recognition model is used as the action space, the model parameter information of the first specific recognition model is used as the state space, and the confidence score of the first specific recognition model is used as the reward function; Based on the action space, the state space and the reward function, each first specific recognition model is optimized using a Q-Learning algorithm.

7. The method according to any one of claims 1 to 6, characterized in that The method further comprises: Obtain real-time video traffic, model inference demand information, and historical resource usage data for each factory monitoring area; With the optimization goals of minimizing resource utilization and maximizing task completion rate, a fitness function is set in combination with the real-time video traffic, the model reasoning requirement information, and the historical resource usage data; A genetic algorithm is used to randomly generate a set of initial resource allocation schemes for each factory monitoring area. Based on the fitness function, the initial resource allocation schemes are subjected to selection, crossover and mutation operations until the convergence condition of the fitness function or the preset number of iterations is reached, and a target resource allocation scheme is generated. The target resource allocation scheme includes the allocation ratio of computing resources and the allocation ratio of storage resources.

8. A data processing system for video surveillance AI intelligent recognition equipment, characterized in that: The system specifically includes: A first data processing module is used to obtain real-time video data from the video surveillance AI intelligent recognition equipment in all factory monitoring areas, extract the light intensity data, shooting angle parameters and background complexity of each factory monitoring area based on the real-time video data, and form an environmental feature vector; a second data processing module, configured to divide all factory monitoring areas into a plurality of environmental categories based on the environmental feature vectors, and determine a central monitoring area for each environmental category, wherein the environmental categories include production line areas, storage areas, safety passages, industrial equipment areas, and office areas; a third data processing module, configured to train a universal recognition model for each environmental category based on the central monitoring area, and to adjust the universal recognition model using a transfer learning method to obtain a specific recognition model for each plant monitoring area in each environmental category; a fourth data processing module, configured to deploy the specific recognition model to the video surveillance AI intelligent recognition device of each corresponding factory monitoring area, and determine the recognition result of each factory monitoring area using the specific recognition model; The fifth data processing module is used to determine the confidence score of each specific recognition model based on the recognition result, and optimize each specific recognition model whose confidence score is lower than a preset confidence threshold through a reinforcement learning algorithm.

9. A computer device, characterized in that: include: A memory, a processor, and a computer program stored in the memory. When the computer program is executed on the processor, the data processing method of the video surveillance AI intelligent recognition device as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by the processor, the data processing method of the video surveillance AI intelligent recognition device as described in any one of claims 1 to 7 is implemented.

Images