Computer image detection method based on big data

Through the combination of high-performance sensors and precise time synchronization, image registration technology and deep learning models, the problems of multi-source data fusion and model generalization capabilities in traditional computer image detection methods are solved, and data consistency and accuracy are achieved in real time. Improvement of monitoring and model generalization capabilities.

CN119027390APending Publication Date: 2024-11-26NANTONG INST OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411109986.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-13
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

Traditional computer image detection methods have difficulties in multi-source data fusion, real-time monitoring and model generalization capabilities, which makes it difficult to guarantee data consistency and accuracy, and models are prone to overfitting.

Method used

Using high-performance sensors and precise time synchronization, multi-source data is spatially aligned and fused through image registration technology, and a unified environmental model is generated using the iterative closest point algorithm, combining lightweight deep learning models and data augmentation technology, setting hyperparameters and Apply pruning and quantization technologies to improve model generalization capabilities and operational efficiency, and deploy real-time inference services on cloud or edge devices to monitor and optimize models in real time.

Benefits of technology

It realizes data consistency and accuracy, improves data management and call efficiency, enhances the system's real-time monitoring capabilities, improves the generalization ability of the model and prevents overfitting, and ensures that the system is in various application scenarios. efficient and stable.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119027390A_ABST
    Figure CN119027390A_ABST
Patent Text Reader

Abstract

The invention discloses a computer image detection method based on big data. By selecting and installing a camera with excellent performance and high resolution, and a radar and a high-precision laser radar sensor which are suitable for severe weather conditions, stable work in various environments can be ensured, and in order to ensure the consistency and accuracy of data, a precise time protocol is used for time synchronization, so that time synchronization is realized. The time synchronization system is regularly checked and calibrated, the time interval of data acquisition is reasonably set, the data is stored by using the distributed storage system, the data security is protected by adopting an encryption technology, and the data is regularly backed up and prevented from being lost. Through the setting, the method not only improves the generalization ability of the model and the effect of preventing overfitting, but also realizes efficient operation and cross-platform consistency. The high efficiency and the stability of the system in various practical application scenes are ensured, and a reliable image detection solution is provided for a user.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of image detection methods, and in particular to a computer image detection method based on big data. Background Art

[0002] Computer image detection methods are a technology that uses computer vision technology to analyze and process images. They are widely used in security monitoring, autonomous driving, medical image analysis, and industrial inspection. These methods extract feature information from images to identify, classify, locate, and track objects, thereby providing intelligent analysis and decision support.

[0003] However, traditional detection methods have difficulties in fusing multi-source data. Data collected by different sensors are not synchronized in time and space, which makes it difficult to ensure the consistency and accuracy of the data. Secondly, existing methods are inefficient in processing large amounts of data and achieving real-time monitoring, and it is difficult to meet the needs of efficient data management and call in dynamic environments. In addition, the generalization ability of the model and the problem of preventing overfitting are also prominent. Many image detection models are prone to overfitting during the training process, that is, the model performs well on the training data, but performs poorly in actual applications.

[0004] Therefore, those skilled in the art have proposed a computer image detection method based on big data. Summary of the invention

[0005] The invention discloses a computer image detection method based on big data, aiming to solve the technical problems in the background technology.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A computer image detection method based on big data comprises the following steps:

[0008] S1. By selecting and installing cameras with superior performance and high resolution, as well as radar and high-precision lidar sensors suitable for severe weather conditions, we ensure that they can work stably in various environments. In order to ensure the consistency and accuracy of the data, we use the Precision Time Protocol (PTP) for time synchronization, and regularly check and calibrate the time synchronization system. We set the time interval for data collection reasonably, use a distributed storage system (such as Hadoop or Amazon S3) to store data, and use encryption technology to protect data security. We also back up data regularly to prevent data loss. This process ensures the stable collection and secure storage of high-quality, multi-source data, laying a solid foundation for subsequent data processing and model training.

[0009] S2. Use image registration technology to spatially align data from different sensors, extract and fuse edge, texture, distance and shape features from each sensor data, use iterative closest point (ICP) algorithm to match data, generate a unified environmental model, store the fused data in a NoSQL database (such as MongoDB), and use Apache Kafka to process and analyze sensor data in real time. This process ensures data consistency and accuracy, improves data management and call efficiency, realizes real-time monitoring and feedback of the detection area, and provides a guarantee for the stable operation of the system.

[0010] S3. Extract representative training samples from the fused data and divide them into training set, validation set and test set in proportion, select a suitable lightweight deep learning model, set hyperparameters, and train the model. During the training process, apply data enhancement and regularization techniques to improve the generalization ability of the model and prevent overfitting. Use the test set to evaluate the model performance. Select the best model for deployment based on the key indicators of precision, recall rate and F1 score. This process ensures that the model has high reliability and effectiveness in various scenarios and conditions, providing stable support for the actual application of the system.

[0011] S4. Reduce the number of model parameters by using pruning and quantization techniques to improve operating efficiency, package the model as a Docker container to ensure cross-platform consistency, and use version control tools to manage the model's code and configuration files to ensure continuous optimization and stability of the model, deploy real-time inference services on the cloud or edge devices, receive sensor data for image detection and analysis, and use log and monitoring tools to monitor the model's operating status in real time to ensure stable operation of the system;

[0012] S5. By using environmental sensors to monitor environmental changes in real time, using online learning algorithms to continuously update models, and collecting user feedback for iterative optimization, the system can adapt to different environments and data changes to ensure the stability and accuracy of the detection model. At the same time, the system has a built-in fault detection module to monitor the running status of the model in real time, and adopts an automatic recovery mechanism to repair faults in time to ensure the continuous and stable operation of the system. In order to ensure the security of data and models, advanced encryption technology is used to protect data, and firewalls and intrusion detection systems are deployed to defend against potential network attacks;

[0013] S6. By collecting and preprocessing data from different sensors, using data mining algorithms for in-depth analysis, discovering potential rules and patterns, and using visualization tools to generate intuitive charts and reports to help users understand and make decisions, build virtual environment models for simulation verification, ensure the robustness and adaptability of the model, deploy real-time monitoring systems and design user-friendly interfaces to achieve real-time monitoring and early warning of environmental changes and detection results, and improve system stability and user experience;

[0014] S7. Design a modular architecture based on system requirements, integrate various functional modules, conduct system testing, and develop specific applications to implement image detection functions. Provide user training and technical support, collect and organize actual application cases, expand the application scope and influence of the system through publicity and promotion, improve market competitiveness, and ensure efficient collaboration, stable operation and continuous optimization of the system in all aspects.

[0015] The computer image detection method based on big data provided by the present invention has the following advantages:

[0016] The consistency and accuracy of the data are ensured by the selection of high-performance sensors and precise time synchronization. Secondly, multi-source data fusion and real-time processing technology improve the efficiency of data management and calling, and enhance the real-time monitoring capability of the system. In addition, data enhancement and regularization techniques are used in the model training process to improve the generalization ability of the model and prevent overfitting. The use of pruning and quantization techniques to optimize the number of model parameters improves the operating efficiency, and ensures cross-platform consistency through Docker containers. Through the above settings, this method not only improves the generalization ability of the model and prevents overfitting, but also achieves efficient operation and cross-platform consistency. This ensures the efficiency and stability of the system in various practical application scenarios, providing users with a reliable image detection solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a schematic diagram of the overall process of a computer image detection method based on big data proposed in the present invention.

[0018] Figure 2 This is a schematic diagram of step S1 of a computer image detection method based on big data proposed in the present invention.

[0019] Figure 3 This is a schematic diagram of step S2 of a computer image detection method based on big data proposed in the present invention.

[0020] Figure 4 This is a schematic diagram of step S3 of a computer image detection method based on big data proposed in the present invention.

[0021] Figure 5 This is a schematic diagram of step S4 of a computer image detection method based on big data proposed in the present invention.

[0022] Figure 6 This is a schematic diagram of step S5 of a computer image detection method based on big data proposed in the present invention.

[0023] Figure 7 This is a schematic diagram of step S6 of a computer image detection method based on big data proposed in the present invention.

[0024] Figure 8 This is a schematic diagram of step S7 of a computer image detection method based on big data proposed in the present invention. DETAILED DESCRIPTION

[0025] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application usually described and marked in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0026] The invention discloses a computer image detection method based on big data.

[0027] Reference Figure 1 , Figure 2 , Figure 3 , Figure 4 ,he, Figure 5 , Figure 6 , Figure 7 and Figure 8 , a computer image detection method based on big data, comprising the following steps:

[0028] S1. By selecting and installing cameras with superior performance and high resolution, as well as radar and high-precision lidar sensors suitable for severe weather conditions, we ensure that they can work stably in various environments. In order to ensure the consistency and accuracy of the data, we use the Precision Time Protocol (PTP) for time synchronization, and regularly check and calibrate the time synchronization system. We set the time interval for data collection reasonably, use a distributed storage system (such as Hadoop or Amazon S3) to store data, and use encryption technology to protect data security. We also back up data regularly to prevent data loss. This process ensures the stable collection and secure storage of high-quality, multi-source data, laying a solid foundation for subsequent data processing and model training.

[0029] The step S1 is refined as follows:

[0030] S11. Choose a camera with superior performance and high resolution on the market to ensure image clarity and detail. Fix the camera at an appropriate location in the detection area and adjust the angle to ensure that the detection area can be covered to the maximum extent. During installation, consider camera protection measures, such as waterproof and dustproof covers, to extend the service life of the equipment and keep it working properly in harsh environments.

[0031] S12. Select radar sensors suitable for adverse weather conditions. These sensors can work stably in rain, snow, and fog. Install the radar sensors at appropriate locations within the detection area to ensure that they can transmit and receive signals without obstacles. When installing, choose the appropriate installation height and angle according to environmental conditions to ensure data accuracy and stability.

[0032] S13. Select a high-precision LiDAR sensor to obtain accurate distance and shape information. Install the LiDAR within the detection area to ensure that its field of view covers all important areas. During installation, ensure that the LiDAR is stable in both horizontal and vertical directions to avoid data errors caused by equipment movement. In addition, the LiDAR lens should be cleaned regularly to ensure data accuracy.

[0033] S14. Use the Precision Time Protocol (PTP) for time synchronization, configure the timestamp marking system of each sensor so that it can record data according to a unified time reference, and regularly check and calibrate the time synchronization system to prevent data inconsistency caused by time drift;

[0034] S15. Set reasonable time intervals to regularly collect data from each sensor and use a distributed storage system (such as Hadoop or Amazon S3) to store data to ensure data security and efficiency. During the storage process, use encryption technology to protect data to prevent unauthorized access and back up data regularly to prevent data loss.

[0035] In summary: By selecting and installing high-performance cameras, radar sensors suitable for severe weather, and high-precision lidar, we ensure that the equipment works stably in various environments. We use the precise time protocol for time synchronization and calibrate the time system regularly. We set a reasonable time interval for data collection and store the data in a distributed system. We use encryption technology to protect data security and back it up regularly. These measures ensure the stable collection and secure storage of high-quality, multi-source data, providing a reliable foundation for subsequent data processing and model training.

[0036] S2. Use image registration technology to spatially align data from different sensors, extract and fuse edge, texture, distance and shape features from each sensor data, use iterative closest point (ICP) algorithm to match data, generate a unified environmental model, store the fused data in a NoSQL database (such as MongoDB), and use Apache Kafka to process and analyze sensor data in real time. This process ensures data consistency and accuracy, improves data management and call efficiency, realizes real-time monitoring and feedback of the detection area, and provides a guarantee for the stable operation of the system.

[0037] The step S2 is refined as follows:

[0038] S21. Use image registration technology to spatially align data from different sensors. First, capture the initial data from each sensor, then use an image registration algorithm (such as phase correlation method or SIFT algorithm) to preprocess the image to find common feature points. Next, use these feature points to calculate the transformation matrix and adjust the rotation, translation and scaling parameters of the image to align them in space. At the same time, use a calibration plate (such as a checkerboard pattern) and a calibration algorithm (such as Zhang Zhengyou calibration method) to calibrate the coordinate system of each sensor to ensure the consistency and accuracy of the data.

[0039] S22. Extract edge and texture features from image data. Use the Canny edge detection algorithm to identify edges in the image. Use the Gabor filter to extract texture features in the image. At the same time, extract distance and shape features from radar and LiDAR data. For radar data, use signal processing technology (such as fast Fourier transform) to extract the distance information of the target object. For LiDAR data, use point cloud processing algorithms (such as RANSAC) to extract the shape features of the object. These features are fused to generate a multi-dimensional feature vector as input for subsequent processing.

[0040] S23, using the iterative closest point algorithm (ICP) to match the data of different sensors. First, feature points are extracted from the initial data of each sensor to construct a point cloud data set. Then, the ICP algorithm is used to calculate the optimal rigid transformation between the source point cloud and the target point cloud so that the distance between the corresponding points of the two is minimized. Through multiple iterations, the transformation parameters are continuously optimized, and finally a unified environment model is generated, which can accurately reflect the spatial position and shape information of objects in the detection area.

[0041] S24. Store the fused data in a NoSQL database, such as MongoDB. First, design an appropriate database schema to define the data structure and storage format. Then, use MongoDB's insert and update operations to store the fused data in the database. In order to achieve efficient data management and call, set index and query optimization strategies to ensure the speed and reliability of data access, back up data regularly to prevent data loss, and use encryption technology to protect data security.

[0042] S25. Use Apache Kafka to process and analyze sensor data in real time. First, configure the Kafka cluster, set topics and partitions to ensure efficient data transmission and processing. Then, develop a data stream processing application and use Kafka consumers to read sensor data from topics for real-time processing and analysis. Based on the data stream processing results, dynamically update the environmental model to achieve real-time monitoring and feedback of the detection area. To ensure the stability and reliability of the system, set up a monitoring and alarm mechanism to promptly handle possible problems.

[0043] In summary: The image registration technology is used to spatially align the data from different sensors, extract and fuse edge, texture, distance and shape features, and use the iterative closest point algorithm (ICP) to match the data to generate a unified environmental model. The fused data is stored in a NoSQL database, and Apache Kafka is used to process and analyze sensor data in real time. This process ensures data consistency and accuracy, improves data management and call efficiency, realizes real-time monitoring and feedback of the detection area, and ensures the stable operation of the system.

[0044] S3. Extract representative training samples from the fused data and divide them into training set, validation set and test set in proportion, select a suitable lightweight deep learning model, set hyperparameters, and train the model. During the training process, apply data enhancement and regularization techniques to improve the generalization ability of the model and prevent overfitting. Use the test set to evaluate the model performance. Select the best model for deployment based on the key indicators of precision, recall rate and F1 score. This process ensures that the model has high reliability and effectiveness in various scenarios and conditions, providing stable support for the actual application of the system.

[0045] The step S3 is refined as follows:

[0046] S31. Extract representative training samples from the fused data to ensure that the samples cover various scenarios and conditions. Through random sampling, select a certain proportion of data from the data set as the training set, and divide the rest into validation set and test set in proportion. The commonly used proportion is: 70% for training set, 15% for validation set, and 15% for test set. This division method can ensure the diversity and balance of the data set, thereby improving the generalization ability of the model.

[0047] S32. Select a suitable lightweight deep learning model, such as MobileNet or Tiny YOLO, based on the application scenario and computing resources. MobileNet has a smaller model size and lower computational complexity, and is suitable for embedded devices. TinyYOLO has high detection accuracy while maintaining a high detection speed. Set the model's hyperparameters, including the learning rate, optimization algorithm (such as Adam or SGD), and batch size. The learning rate determines the step size of the model parameter update, the optimization algorithm is used to minimize the loss function, and the batch size affects the training speed and memory usage.

[0048] S33. Input the training set into the model, calculate the loss value through forward propagation, use the labeled data (such as the object position and category label in the image) to guide the model learning, calculate the gradient of the loss function through the back propagation algorithm, and update the model parameters to minimize the loss. Regularly use the validation set to evaluate the model performance, calculate the precision and recall rate indicators, and adjust the hyperparameters (such as learning rate and batch size) according to the validation results to optimize the model training process. This process should be repeated until the model performance reaches the expected level.

[0049] S34. During the training process, data augmentation techniques are applied to increase the diversity of training data. For example, random cropping can simulate different perspectives, while rotation and flipping can simulate different directions and postures. These techniques can improve the generalization ability of the model and prevent overfitting. In addition, regularization methods such as Dropout and L2 regularization are used to further prevent the model from overfitting the training data during training. Dropout enhances the robustness of the model by randomly ignoring the activation values ​​of some neurons; L2 regularization limits the size of model parameters by adding parameter penalty terms in the loss function.

[0050] S35. After training is completed, the model is evaluated using the test set to calculate key performance indicators such as precision (accuracy), recall rate and F1 score. Precision indicates the proportion of correct predictions by the model, recall rate indicates the proportion of all positive samples recognized by the model, and F1 score is the harmonic mean of precision and recall rate. Based on these indicators, the model with the best performance is selected for actual deployment to ensure that the performance of the model on the test set can represent its performance in real application scenarios, further improving the reliability and effectiveness of the system.

[0051] In summary: By extracting representative training samples from the fusion data and dividing them into training set, validation set and test set in proportion, a suitable lightweight deep learning model is selected and hyperparameters are set for model training. Data enhancement and regularization techniques are applied during the training process to improve the generalization ability of the model and prevent overfitting. The model performance is evaluated using the test set, and the best model is selected for deployment based on the key indicators of precision, recall rate and F1 score to ensure that the model has high reliability and effectiveness in various scenarios and conditions, providing stable support for the actual application of the system.

[0052] S4. Reduce the number of model parameters by using pruning and quantization techniques to improve operating efficiency, package the model as a Docker container to ensure cross-platform consistency, and use version control tools to manage the model's code and configuration files to ensure continuous optimization and stability of the model, deploy real-time inference services on the cloud or edge devices, receive sensor data for image detection and analysis, and use log and monitoring tools to monitor the model's operating status in real time to ensure stable operation of the system;

[0053] The step S4 is refined as follows:

[0054] S41. Use pruning techniques (such as weight pruning and neuron pruning) to reduce the number of model parameters. Pruning techniques reduce the amount of computation and memory usage by analyzing redundant weights and neurons in the model and removing these unimportant parts. Next, use quantization techniques to convert floating-point parameters in the model into low-precision integers (such as 8-bit integers). This not only reduces computational costs, but also improves the efficiency of the model on embedded devices. During the quantization process, it is necessary to ensure that the loss of accuracy is within an acceptable range. The performance of the quantized model is further optimized through quantization-aware training methods.

[0055] S42. Package the trained model into a Docker container, including all necessary dependency libraries and operating environment to ensure the consistency of the model on different platforms. Use Dockerfile to define the container build process and create an image. Upload the container image to the container registry (such as Docker Hub), and then deploy the container to the cloud server or edge device. Cloud deployment can use Kubernetes for container orchestration and management to ensure high availability and scalability of the model. Deployment on edge devices needs to consider the resource limitations of the device and select the appropriate device and network environment for deployment.

[0056] S43. Use Git version control tools to manage the model's code and configuration files, create different branches and tags to identify different versions of the model, submit each model update and improvement to the version control system, record changes and updates, conduct regular code review and testing to ensure the quality and stability of the model. Through Git's version control function, you can easily roll back to the previous version, which is convenient for debugging and troubleshooting. Update the model regularly, combine new data and technologies, and continuously optimize the performance and adaptability of the model;

[0057] S44. Deploy real-time reasoning services on the cloud or edge devices, and use REST API or gRPC interface to receive sensor data. The reasoning service inputs the sensor data into the model, performs image detection and analysis, and generates detection results. According to the detection results, it responds in real time and triggers corresponding control instructions, such as issuing alarms, adjusting device parameters, or performing automated operations. The reasoning service requires efficient processing power and low latency to ensure real-time performance and accuracy. Through load balancing and automatic expansion mechanisms, the stable operation of the reasoning service is guaranteed;

[0058] S45. Use logging tools (such as ELK Stack) to record the model's operation log, including input data, processing process and output results, to facilitate problem tracking and analysis. Use monitoring tools (such as Prometheus and Grafana) to monitor the model's operation status and performance indicators in real time, such as CPU usage, memory usage, response time and detection accuracy. Regularly check logs and monitoring data to identify potential problems and fix them in time. Maintain and update the model based on the monitoring results to ensure the stable operation of the model in different environments and conditions. Use automated operation and maintenance tools (such as Ansible or Terraform) to simplify the maintenance and update process and improve efficiency.

[0059] In summary: By using pruning and quantization techniques to reduce the number of model parameters and improve operating efficiency, the model is packaged as a Docker container to ensure cross-platform consistency, and version control tools are used to manage model code and configuration files to ensure continuous optimization and stability of the model. Deploy real-time inference services on the cloud or edge devices, receive sensor data for image detection and analysis, and use log and monitoring tools to monitor the operating status of the model in real time to ensure stable operation of the system.

[0060] S5. By using environmental sensors to monitor environmental changes in real time, using online learning algorithms to continuously update models, and collecting user feedback for iterative optimization, the system can adapt to different environments and data changes to ensure the stability and accuracy of the detection model. At the same time, the system has a built-in fault detection module to monitor the running status of the model in real time, and adopts an automatic recovery mechanism to repair faults in time to ensure the continuous and stable operation of the system. In order to ensure the security of data and models, advanced encryption technology is used to protect data, and firewalls and intrusion detection systems are deployed to defend against potential network attacks;

[0061] The step S5 is refined as follows:

[0062] S51. Use environmental sensors (such as light sensors, temperature sensors, and humidity sensors) to monitor environmental changes in real time and obtain current light intensity and weather condition data. These data are transmitted to the image detection system through the sensor interface. The system dynamically adjusts the parameters and detection threshold of the model according to the preset environmental change rules. For example, when the light intensity is low, the contrast enhancement parameters of the image processing can be reduced; under severe weather conditions, the fault tolerance of the detection threshold can be increased. This adaptive adjustment mechanism ensures the stability and accuracy of the model in different environments.

[0063] S52. Use online learning algorithms, such as streaming data classifiers and incremental learning models, to continuously update image detection models. After each acquisition of new data, the online learning algorithm immediately fine-tunes the model and updates the model parameters. Through regular model evaluation and performance monitoring, the system can automatically detect signs of model performance degradation and trigger a self-adjustment mechanism that can reconfigure the model structure and parameters based on real-time data to optimize the model's detection performance. The application of online learning algorithms enables the model to adapt to changing environments and data, improving the reliability and accuracy of long-term detection.

[0064] S53. Collect user feedback through the user interface, including user evaluation of the accuracy and reliability of the detection results. The system will summarize and analyze these feedback data to identify common false detections and missed detections. Based on user feedback, the development team can iteratively optimize the model, adjust model parameters or retrain the model to improve detection performance. In addition, regularly conduct user satisfaction surveys to obtain users' evaluation of the overall system experience, and use these feedback results as an important reference for model optimization. Through this feedback-driven iterative optimization, user satisfaction and trust in the system are improved;

[0065] S54. The system has a built-in fault detection module to monitor the operating status and performance indicators of the model in real time. When a model performance degradation or abnormal behavior (such as an abnormally high false positive rate or processing delay) is detected, the fault detection module will immediately record the fault information and trigger an automatic recovery mechanism. The automatic recovery mechanism includes automatically restarting the model service, switching to a backup model, or reloading model parameter operations to ensure that the system can quickly return to normal operation. The fault detection module will also generate a detailed fault report for the development team to conduct further analysis and improvement;

[0066] S55. Use advanced encryption technologies, such as AES256 encryption, to protect the security of sensor data and model parameters during transmission and storage. Encrypt sensor data before transmission to ensure that the data is not intercepted or tampered with during network transmission. Store model parameters in an encrypted database to prevent unauthorized access and data leakage. Deploy firewalls and intrusion detection systems (IDS) to monitor the system's network traffic and security events in real time, detect and defend against potential network attacks, and regularly update and review security policies to ensure that the system is always in a secure protection state.

[0067] In summary: By using environmental sensors to monitor environmental changes in real time, using online learning algorithms to continuously update models, and collecting user feedback for iterative optimization, the system can adapt to different environments and data changes to ensure the stability and accuracy of the detection model. At the same time, the system has a built-in fault detection module to monitor the running status of the model in real time, and uses an automatic recovery mechanism to repair faults in a timely manner to ensure the continuous and stable operation of the system. In order to ensure the security of data and models, advanced encryption technology is used to protect data, and firewalls and intrusion detection systems are deployed to defend against potential network attacks.

[0068] S6. By collecting and preprocessing data from different sensors, using data mining algorithms for in-depth analysis, discovering potential rules and patterns, and using visualization tools to generate intuitive charts and reports to help users understand and make decisions, build virtual environment models for simulation verification, ensure the robustness and adaptability of the model, deploy real-time monitoring systems and design user-friendly interfaces to achieve real-time monitoring and early warning of environmental changes and detection results, and improve system stability and user experience;

[0069] The step S6 is refined as follows:

[0070] S61. Collect data from different sensors, such as images, radar, and LiDAR, and perform preprocessing, including data cleaning, format conversion, and feature extraction. Use data mining algorithms (such as association rule mining, cluster analysis, and classification algorithms) to conduct in-depth analysis of the preprocessed data. Through analysis, potential rules and patterns in the data can be discovered, such as object detection trends in specific environments or correlations between sensor data. These analysis results can help optimize the detection model and improve the overall performance of the system.

[0071] S62. Import data analysis results into visualization tools, such as Tableau or D3.js, to generate various types of charts (such as line charts, bar charts, scatter plots, and heat maps). These charts can intuitively display the changing trends and patterns of data. Through interactive functions, users can dynamically adjust chart parameters and deeply explore data details. The generated reports contain key analysis results and chart descriptions, which are convenient for users to understand and make decisions. Visualization tools also support the export and sharing of reports, which facilitates team collaboration and discussion;

[0072] S63. Use multi-source data to build a virtual environment model, including the spatial position, shape and motion trajectory of objects. Use simulation software (such as MATLAB Simulink or Gazebo) to perform virtual simulation to simulate different scenes and environmental conditions, such as lighting changes, weather changes and object movement. Verify the robustness and adaptability of the detection model under various conditions through simulation to ensure the stability of the model in practical applications. The simulation results can be used to optimize model parameters and improve detection performance.

[0073] S64. Deploy a real-time monitoring system, connect various sensors and detection models, and obtain environmental changes and detection result data in real time. The monitoring system includes an environmental monitoring module and a detection result analysis module. The former monitors the light intensity, temperature and humidity environmental parameters, and the latter analyzes the output results of the detection model and establishes an early warning mechanism. When the environmental parameters or detection results are abnormal, the system will automatically issue an alarm and trigger the corresponding emergency handling procedures, such as adjusting the detection model parameters or starting the backup system. In this way, potential risks can be discovered and handled in advance to ensure the stable operation of the system.

[0074] S65. When designing the user interface (UI), focus on the intuitiveness and ease of use of the interface to ensure that users can easily operate and obtain feedback. The UI should include data display panel, operation control panel and alarm information panel modules. The data display panel displays real-time detection results and environmental parameters. The operation control panel provides model parameter adjustment and simulation control functions. The alarm information panel displays warning information and processing suggestions. Multi-device compatibility is achieved to ensure that the UI can run smoothly on PCs, mobile devices and VR devices. Responsive design can adapt to different screen sizes and resolutions to provide a consistent user experience.

[0075] In summary: By collecting and preprocessing data from different sensors, using data mining algorithms for in-depth analysis, discovering potential rules and patterns, and using visualization tools to generate intuitive charts and reports to help users understand and make decisions. Build a virtual environment model for simulation verification to ensure the robustness and adaptability of the model, deploy a real-time monitoring system and design a user-friendly interface to achieve real-time monitoring and early warning of environmental changes and detection results, and improve system stability and user experience.

[0076] S7. Design a modular architecture according to system requirements, integrate various functional modules, conduct system testing, and develop specific applications to implement image detection functions, provide user training and technical support, collect and organize actual application cases, expand the application scope and influence of the system through publicity and promotion, improve market competitiveness, and ensure efficient collaborative work, stable operation and continuous optimization of the system in all aspects;

[0077] The step S7 is refined as follows:

[0078] S71. Design a modular architecture based on system requirements and divide the system into several functional modules, such as data acquisition module, data processing module, model reasoning module and user interface module. To ensure the collaborative work of each module, define clear system interfaces and communication protocols. For example, use REST API or gRPC to implement communication between modules to ensure that data can be efficiently and reliably transmitted between modules. The scalability and maintainability of the system need to be considered during design to facilitate future upgrades and functional expansions.

[0079] S72. Integrate the designed modules to build a complete image detection system. During the integration process, ensure that the modules exchange data according to the defined interfaces and communication protocols. After the integration is completed, perform system testing, including unit testing, integration testing and system testing. Unit testing verifies the basic functions of each module, integration testing verifies the collaborative work between modules, and system testing comprehensively evaluates the performance and stability of the entire system. Through testing, potential problems are discovered and fixed to ensure that the system's functions and performance meet the expected requirements.

[0080] S73. Based on the system architecture and functional requirements, develop specific applications to realize all the functions of the image detection system. The application includes a front-end user interface and a back-end service. The front-end user interface is responsible for interacting with users, displaying detection results and receiving user instructions; the back-end service is responsible for processing sensor data, running the detection model and returning results. After development is completed, deploy the application to the production environment to ensure the stable operation of the system in actual use. During deployment, configure load balancing and automatic expansion mechanisms to improve the reliability and response speed of the system.

[0081] S74. Provide detailed training for system users, including basic system operations, solutions to common problems and precautions. The training can be in the form of face-to-face classroom teaching, online courses or operation manuals. Through training, ensure that users can use the system proficiently and maximize the functions of the system. Establish a technical support team to provide immediate technical support and problem solving to help users solve problems encountered during use. The technical support team should also collect user feedback to continuously improve and optimize the system.

[0082] S75. Collect and organize cases of the system in actual application, record in detail the system's application scenarios, problems solved and results achieved, write case reports, demonstrate the system's advantages and actual application results, and publicize and promote it through online and offline channels.

[0083] In summary: By designing a modular architecture according to system requirements, integrating various functional modules, conducting system testing, and developing specific applications to implement image detection functions. Provide user training and technical support, collect and organize actual application cases, and expand the application scope and influence of the system through publicity and promotion, and improve market competitiveness. Ensure efficient collaboration, stable operation, and continuous optimization of the system in all aspects.

[0084] The above description is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. The replacement may be a replacement of a part of the structure, device, method step, or a complete technical solution. Any replacement or change according to the technical solution and the inventive concept of the present invention shall be included in the protection scope of the present invention.

Claims

1. A computer image detection method based on big data, characterized in that: The following steps are involved: S1. By selecting and installing cameras with superior performance and high resolution, as well as radars and high-precision lidar sensors suitable for severe weather conditions, we ensure that they can work stably in various environments. In order to ensure the consistency and accuracy of the data, we use the precise time protocol for time synchronization, and regularly check and calibrate the time synchronization system. We set the time interval for data collection reasonably, use a distributed storage system to store data, and use encryption technology to protect data security. We also back up data regularly to prevent data loss. This process ensures the stable collection and safe storage of high-quality, multi-source data, laying the foundation for subsequent data processing and model training. S2. Use image registration technology to spatially align data from different sensors, extract and fuse edge, texture, distance and shape features from each sensor data, use iterative closest point algorithm to match data, generate a unified environmental model, store the fused data in a NoSQL database, and use Apache Kafka to process and analyze sensor data in real time. This process ensures data consistency and accuracy, improves data management and call efficiency, realizes real-time monitoring and feedback of the detection area, and provides a guarantee for the stable operation of the system. S3. Extract representative training samples from the fused data and divide them into training set, validation set and test set in proportion, select a suitable lightweight deep learning model, set hyperparameters, and train the model. During the training process, apply data enhancement and regularization techniques to improve the generalization ability of the model and prevent overfitting. Use the test set to evaluate the model performance. Select the best model for deployment based on the key indicators of precision, recall rate and F1 score. This process ensures that the model has high reliability and effectiveness in various scenarios and conditions, providing stable support for the actual application of the system. S4. Reduce the number of model parameters by using pruning and quantization techniques to improve operating efficiency, package the model as a Docker container to ensure cross-platform consistency, and use version control tools to manage the model's code and configuration files to ensure continuous optimization and stability of the model, deploy real-time inference services on the cloud or edge devices, receive sensor data for image detection and analysis, and use log and monitoring tools to monitor the model's operating status in real time to ensure stable operation of the system; S5. By using environmental sensors to monitor environmental changes in real time, using online learning algorithms to continuously update models, and collecting user feedback for iterative optimization, the system can adapt to different environments and data changes to ensure the stability and accuracy of the detection model. At the same time, the system has a built-in fault detection module to monitor the running status of the model in real time, and adopts an automatic recovery mechanism to repair faults in time to ensure the continuous and stable operation of the system. In order to ensure the security of data and models, advanced encryption technology is used to protect data, and firewalls and intrusion detection systems are deployed to defend against potential network attacks; S6. By collecting and preprocessing data from different sensors, using data mining algorithms for in-depth analysis, discovering potential rules and patterns, and using visualization tools to generate intuitive charts and reports to help users understand and make decisions, build virtual environment models for simulation verification, ensure the robustness and adaptability of the model, deploy real-time monitoring systems and design user-friendly interfaces to achieve real-time monitoring and early warning of environmental changes and detection results, and improve system stability and user experience; S7. Design a modular architecture based on system requirements, integrate various functional modules, conduct system testing, and develop specific applications to implement image detection functions. Provide user training and technical support, collect and organize actual application cases, expand the application scope and influence of the system through publicity and promotion, improve market competitiveness, and ensure efficient collaboration, stable operation and continuous optimization of the system in all aspects.

2. A computer image detection method based on big data according to claim 1, characterized in that: The step S1 is refined as follows: S11. Choose a camera with superior performance and high resolution on the market to ensure the clarity and detail of the image. Fix the camera in an appropriate position in the detection area and adjust the angle to ensure that the detection area can be covered to the maximum extent. During the installation process, the camera protection measures should be considered, such as waterproof and dustproof covers, to extend the service life of the equipment and keep it working normally in harsh environments. S12. Select radar sensors suitable for adverse weather conditions. These sensors can work stably in rain, snow, and fog. Install the radar sensors at appropriate locations within the detection area to ensure that they can transmit and receive signals without obstacles. When installing, choose the appropriate installation height and angle according to environmental conditions to ensure data accuracy and stability. S13. Select a high-precision LiDAR sensor to obtain accurate distance and shape information. Install the LiDAR within the detection area to ensure that its field of view covers all important areas. During installation, ensure that the LiDAR is stable in both horizontal and vertical directions to avoid data errors caused by equipment movement. In addition, the LiDAR lens should be cleaned regularly to ensure data accuracy. S14. Use the precise time protocol for time synchronization, configure the timestamp marking system of each sensor so that it can record data according to a unified time reference, and regularly check and calibrate the time synchronization system to prevent data inconsistency caused by time drift; S15. Set reasonable time intervals to regularly collect data from each sensor and use a distributed storage system to store data to ensure data security and efficiency. During the storage process, use encryption technology to protect data to prevent unauthorized access and back up data regularly to prevent data loss.

3. The computer image detection method based on big data according to claim 1, characterized in that: The step S2 is refined as follows: S21. Use image registration technology to spatially align data from different sensors. First, capture the initial data from each sensor, then use the image registration algorithm to preprocess the image to find common feature points. Next, use these feature points to calculate the transformation matrix and adjust the rotation, translation and scaling parameters of the image to align them in space. At the same time, use the calibration plate and calibration algorithm to calibrate the coordinate system of each sensor to ensure the consistency and accuracy of the data. S22. Extract edge and texture features from image data. Use Canny edge detection algorithm to identify edges in images. Use Gabor filter to extract texture features in images. At the same time, extract distance and shape features from radar and LiDAR data. For radar data, use signal processing technology to extract distance information of target objects. For LiDAR data, use point cloud processing algorithm to extract shape features of objects. These features are fused to generate multi-dimensional feature vectors as input for subsequent processing. S23, using the iterative closest point algorithm to match the data of different sensors. First, extract the feature points from the initial data of each sensor to construct a point cloud data set. Then, use the ICP algorithm to calculate the optimal rigid transformation between the source point cloud and the target point cloud so that the distance between the corresponding points of the two is minimized. Through multiple iterations, the transformation parameters are continuously optimized to finally generate a unified environment model, which can accurately reflect the spatial position and shape information of objects in the detection area. S24. Store the fused data in a NoSQL database. First, design an appropriate database schema to define the data structure and storage format. Then, use MongoDB insert and update operations to store the fused data in the database. In order to achieve efficient data management and call, set index and query optimization strategies to ensure the speed and reliability of data access, back up data regularly to prevent data loss, and use encryption technology to protect data security. S25. Use Apache Kafka to process and analyze sensor data in real time. First, configure the Kafka cluster, set topics and partitions to ensure efficient data transmission and processing. Then, develop a data stream processing application and use Kafka consumers to read sensor data from topics for real-time processing and analysis. Based on the data stream processing results, dynamically update the environmental model to achieve real-time monitoring and feedback of the detection area. To ensure the stability and reliability of the system, set up a monitoring and alarm mechanism to promptly handle possible problems.

4. The computer image detection method based on big data according to claim 1, characterized in that: The step S3 is refined as follows: S31. Extract representative training samples from the fused data to ensure that the samples cover various scenarios and conditions. Through random sampling, select a certain proportion of data from the data set as the training set, and divide the rest into validation set and test set in proportion. The commonly used proportion is: 70% for training set, 15% for validation set, and 15% for test set. This division method can ensure the diversity and balance of the data set, thereby improving the generalization ability of the model. S32. According to the application scenario and computing resources, select a suitable lightweight deep learning model, MobileNet or TinyYOLO. MobileNet has a smaller model size and lower computational complexity, which is suitable for embedded devices. Tiny YOLO has high detection accuracy while maintaining high detection speed. Set the model's hyperparameters, including learning rate, optimization algorithm, and batch size. The learning rate determines the step size of the model parameter update. The optimization algorithm is used to minimize the loss function. The batch size affects the training speed and memory usage. S33. Input the training set into the model, calculate the loss value through forward propagation, use the labeled data to guide model learning, calculate the gradient of the loss function through the back-propagation algorithm, and update the model parameters to minimize the loss. Regularly use the validation set to evaluate the model performance, calculate the precision and recall rate indicators, and adjust the hyperparameters according to the validation results to optimize the model training process. This process should be repeated until the model performance reaches the expected level. S34. During the training process, data enhancement techniques are applied to increase the diversity of training data. Random cropping can simulate different perspectives, and rotation and flipping can simulate different directions and postures. These techniques can improve the generalization ability of the model and prevent overfitting. In addition, regularization methods, Dropout and L2 regularization, are used to further prevent the model from overfitting the training data during training. Dropout enhances the robustness of the model by randomly ignoring the activation values ​​of some neurons; L2 regularization limits the size of model parameters by adding parameter penalty terms in the loss function; S35. After training, the model is evaluated using the test set to calculate key performance indicators, including precision, recall, and F1 score. Precision indicates the proportion of correct predictions by the model, recall indicates the proportion of all positive samples recognized by the model, and F1 score is the harmonic mean of precision and recall. Based on these indicators, the model with the best performance is selected for actual deployment to ensure that the performance of the model on the test set can represent its performance in real application scenarios, further improving the reliability and effectiveness of the system.

5. The computer image detection method based on big data according to claim 1, characterized in that: The step S4 is refined as follows: S41. Use pruning technology to reduce the number of model parameters. Pruning technology reduces the amount of computation and memory usage by analyzing redundant weights and neurons in the model and removing these unimportant parts. Next, use quantization technology to convert floating-point parameters in the model into low-precision integers, which can not only reduce computing costs but also improve the running efficiency of the model on embedded devices. During the quantization process, it is necessary to ensure that the loss of accuracy is within an acceptable range. The performance of the quantized model is further optimized through quantization-aware training methods. S42. Package the trained model into a Docker container, including all necessary dependency libraries and operating environment to ensure the consistency of the model on different platforms. Use Dockerfile to define the container build process and create an image. Upload the container image to the container registry, and then deploy the container to the cloud server or edge device. Cloud deployment can use Kubernetes for container orchestration and management to ensure high availability and scalability of the model. Deployment on edge devices needs to consider the resource limitations of the device and select the appropriate device and network environment for deployment. S43. Use Git version control tools to manage the model's code and configuration files, create different branches and tags to identify different versions of the model, submit each model update and improvement to the version control system, record changes and updates, conduct regular code review and testing to ensure the quality and stability of the model. Through Git's version control function, you can easily roll back to the previous version, which is convenient for debugging and troubleshooting. Update the model regularly, combine new data and technologies, and continuously optimize the performance and adaptability of the model; S44. Deploy real-time reasoning services on the cloud or edge devices, use REST API or gRPC interface to receive sensor data, the reasoning service inputs the sensor data into the model, performs image detection and analysis, and generates detection results. According to the detection results, respond in real time and trigger corresponding control instructions, issue alarms, adjust device parameters or perform automated operations. The reasoning service requires efficient processing power and low latency to ensure real-time and accuracy. Through load balancing and automatic expansion mechanism, the stable operation of the reasoning service is guaranteed; S45. Use log tools to record the model's operation log, including input data, processing process and output results, to facilitate problem tracking and analysis. Use monitoring tools to monitor the model's operation status and performance indicators, CPU usage, memory usage, response time and detection accuracy in real time. Regularly check logs and monitoring data to identify potential problems and fix them in time. Maintain and update the model based on the monitoring results to ensure the model's stable operation in different environments and conditions. Use automated operation and maintenance tools to simplify the maintenance and update process and improve efficiency.

6. The computer image detection method based on big data according to claim 1, characterized in that: The step S5 is refined as follows: S51. Use environmental sensors to monitor environmental changes in real time and obtain current light intensity and weather condition data. These data are transmitted to the image detection system through the sensor interface. The system dynamically adjusts the model parameters and detection thresholds according to the preset environmental change rules. When the light intensity is low, the contrast enhancement parameters of the image processing can be reduced; under severe weather conditions, the fault tolerance of the detection threshold can be increased. This adaptive adjustment mechanism ensures the stability and accuracy of the model in different environments. S52. Use online learning algorithms, streaming data classifiers, and incremental learning models to continuously update image detection models. After each acquisition of new data, the online learning algorithm immediately fine-tunes the model and updates the model parameters. Through regular model evaluation and performance monitoring, the system can automatically detect signs of model performance degradation and trigger a self-adjustment mechanism that can reconfigure the model structure and parameters based on real-time data to optimize the model's detection performance. The application of online learning algorithms enables the model to adapt to changing environments and data, improving the reliability and accuracy of long-term detection. S53. Collect user feedback through the user interface, including user evaluation of the accuracy and reliability of the detection results. The system will summarize and analyze these feedback data to identify common false detections and missed detections. Based on user feedback, the development team can iteratively optimize the model, adjust model parameters or retrain the model to improve detection performance. In addition, regularly conduct user satisfaction surveys to obtain users' evaluation of the overall system experience, and use these feedback results as an important reference for model optimization. Through this feedback-driven iterative optimization, user satisfaction and trust in the system are improved; S54. The system has a built-in fault detection module to monitor the operating status and performance indicators of the model in real time. When a model performance degradation or abnormal behavior is detected, the fault detection module will immediately record the fault information and trigger the automatic recovery mechanism. The automatic recovery mechanism includes automatically restarting the model service, switching to a backup model, or reloading model parameters to ensure that the system can quickly return to normal operation. The fault detection module will also generate a detailed fault report for the development team to further analyze and improve. S55. Use advanced encryption technology, AES256 encryption, to protect the security of sensor data and model parameters during transmission and storage. Sensor data is encrypted before transmission to ensure that the data is not intercepted or tampered with during network transmission. Model parameters are stored in an encrypted database to prevent unauthorized access and data leakage. Deploy firewalls and intrusion detection systems to monitor the system's network traffic and security events in real time, detect and defend against potential network attacks, and regularly update and review security policies to ensure that the system is always in a secure protection state.

7. The computer image detection method based on big data according to claim 1, characterized in that: The step S6 is refined as follows: S61. Collect data from different sensors, including images, radar, and LiDAR, and perform preprocessing, including data cleaning, format conversion, and feature extraction. Use data mining algorithms to conduct in-depth analysis of the preprocessed data to discover potential rules and patterns in the data, object detection trends in specific environments, or correlations between sensor data. These analysis results can help optimize the detection model and improve the overall performance of the system. S62. Import data analysis results into visualization tools, such as Tableau or D3.js, to generate various types of charts that can intuitively display the changing trends and patterns of data. Through interactive functions, users can dynamically adjust chart parameters and deeply explore data details. The generated reports contain key analysis results and chart descriptions, which are convenient for users to understand and make decisions. Visualization tools also support the export and sharing of reports, which facilitates team collaboration and discussion; S63. Use multi-source data to build a virtual environment model, including the spatial position, shape and motion trajectory of objects. Use simulation software to perform virtual simulation to simulate different scenes and environmental conditions, lighting changes, weather changes and object movement. Verify the robustness and adaptability of the detection model under various conditions through simulation to ensure the stability of the model in practical applications. The simulation results can be used to optimize model parameters and improve detection performance. S64. Deploy a real-time monitoring system, connect various sensors and detection models, and obtain environmental changes and detection result data in real time. The monitoring system includes an environmental monitoring module and a detection result analysis module. The former monitors the light intensity, temperature and humidity environmental parameters, and the latter analyzes the output results of the detection model and establishes an early warning mechanism. When the environmental parameters or detection results are abnormal, the system will automatically issue an alarm and trigger the corresponding emergency handling procedures, adjust the detection model parameters or start the backup system. In this way, potential risks can be discovered and handled in advance to ensure the stable operation of the system. S65. When designing the user interface, focus on the intuitiveness and ease of use of the interface to ensure that users can easily operate and obtain feedback. The designed user interface should include data display panel, operation control panel and alarm information panel modules. The data display panel displays real-time detection results and environmental parameters. The operation control panel provides model parameter adjustment and simulation control functions. The alarm information panel displays warning information and processing suggestions to achieve multi-device compatibility and ensure that the UI can run smoothly on PCs, mobile devices and VR devices. Responsive design can adapt to different screen sizes and resolutions to provide a consistent user experience.

8. The computer image detection method based on big data according to claim 1, characterized in that: The step S7 is refined as follows: S71. Design a modular architecture based on system requirements and divide the system into several functional modules, including data acquisition module, data processing module, model reasoning module and user interface module. To ensure the collaborative work of each module, define clear system interfaces and communication protocols, use REST API or gRPC to implement communication between modules, and ensure that data can be efficiently and reliably transmitted between modules. The scalability and maintainability of the system need to be considered during design to facilitate future upgrades and functional expansions. S72. Integrate the designed modules to build a complete image detection system. During the integration process, ensure that the modules exchange data according to the defined interfaces and communication protocols. After the integration is completed, perform system testing, including unit testing, integration testing and system testing. Unit testing verifies the basic functions of each module, integration testing verifies the collaborative work between modules, and system testing comprehensively evaluates the performance and stability of the entire system. Through testing, potential problems are discovered and fixed to ensure that the system's functions and performance meet the expected requirements. S73. Based on the system architecture and functional requirements, develop specific applications to realize all the functions of the image detection system. The application includes a front-end user interface and a back-end service. The front-end user interface is responsible for interacting with users, displaying detection results and receiving user instructions; the back-end service is responsible for processing sensor data, running the detection model and returning results. After development is completed, deploy the application to the production environment to ensure the stable operation of the system in actual use. During deployment, configure load balancing and automatic expansion mechanisms to improve the reliability and response speed of the system. S74. Provide detailed training for system users, including basic system operations, solutions to common problems and precautions. The training can be in the form of face-to-face classroom teaching, online courses or operation manuals. Through training, ensure that users can use the system proficiently and maximize the functions of the system. Establish a technical support team to provide immediate technical support and problem solving to help users solve problems encountered during use. The technical support team should also collect user feedback to continuously improve and optimize the system. S75. Collect and organize cases of the system in actual application, record in detail the system's application scenarios, problems solved and results achieved, write case reports, demonstrate the system's advantages and actual application results, and publicize and promote it through online and offline channels.