Method and system for realizing forbidden area forbidding detection in different environments
By applying computer vision and artificial intelligence technology in the prohibited area ban system, combined with algorithms such as YOLO, SSD, Faster R-CNN, efficient and intelligent prohibited area detection in complex environments, solving the limited effect of traditional security measures The problem of the problem improves detection accuracy and real-time performance, ensuring security and privacy protection.
Patent Information
- Application Number
- CN202411057818.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-11-19
AI Technical Summary
Traditional physical security measures have limited effects in complex and changing environments, and it is difficult to meet the security needs of modern society, especially in places with large flow of people and areas that require strict control of entry. How to achieve efficient and intelligent prohibited areas. Test.
Using object detection technology based on computer vision and artificial intelligence, through data collection, preprocessing, feature extraction and matching, combined with algorithms such as YOLO, SSD, Faster R-CNN, real-time monitoring and alarming are achieved to ensure accurate identification and alarm in different environments. Prevent unauthorized entry.
It improves detection accuracy and real-time performance, reduces false alarms and missed reports, provides a variety of alarm methods, supports comprehensive security management, and ensures regional security and privacy protection.
Smart Images

Figure CN118982785A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of computers, and in particular relates to a method and system for realizing forbidden area entry detection in different environments. Background Art
[0002] With the development of society and the advancement of science and technology, public safety and privacy protection have become increasingly hot issues of global concern. In various scenarios, such as airports, stations, government agencies, commercial centers, schools, and private residences, it is particularly important to ensure the safety of certain specific areas and prevent unauthorized entry. Although traditional physical security measures (such as fences and guards) can provide protection to a certain extent, their effectiveness and flexibility are often limited in the face of complex and changing environments and potential security threats. The prohibited area entry technology based on target detection technology came into being in this context. Combined with modern computer vision and artificial intelligence technology, it provides a more intelligent and efficient solution for security protection.
[0003] Object detection technology is an important research direction in the field of computer vision, and its purpose is to identify and locate specific objects in images or videos. Early object detection methods mainly relied on manually designed features and traditional machine learning algorithms, such as the detection method combining Haar features with Adaboost classifiers. In these methods, researchers need to manually design and select image features, and then use simple classifiers for detection. Although they perform well in certain specific tasks, they are often not ideal when faced with complex backgrounds and diverse targets.
[0004] With the rise of deep learning technology, object detection algorithms based on convolutional neural networks (CNNs) have risen rapidly, significantly improving the accuracy and efficiency of detection. Typical deep learning object detection algorithms include YOLO (You Only Look Once), SSD (Single Shot Multi Box Detector), and Faster R-CNN (Region Convolutional Neural Networks). These algorithms can automatically learn deep features in images and perform object classification and positioning through end-to-end training, which greatly simplifies the process of feature design and shows strong robustness and adaptability in complex scenes. The structure of the Yolo model greatly improves the efficiency of object detection. YOLOv8 uses an improved CSPDarknet as its backbone network. This backbone network is responsible for extracting multi-scale features from the input image. The backbone network contains multiple stages, each of which consists of a convolutional layer, an activation function (such as ReLU or Mish), a pooling layer, and a CSP module. These modules work together to improve the ability and efficiency of feature extraction. The Head part of YOLOv8 uses a DecoupledHead structure to separate the classification and bounding box regression tasks, thereby improving detection accuracy. The loss function of YOLOv8 consists of three main parts: bounding box loss (bbox loss), classification loss (cls loss) and target confidence loss (objloss). These losses jointly optimize the overall performance of the model.
[0005] In practical applications, the requirements for prohibited area entry technology are extensive and diverse. For example:
[0006] Airports and stations: These places have large and complex passenger flows, and access to certain areas (such as boarding gates, baggage handling areas, control rooms, etc.) needs to be strictly controlled to ensure passenger safety and operational order.
[0007] Government agencies: Security protection of sensitive areas (such as confidential rooms, office areas, etc.) is essential to prevent unauthorized entry in order to protect state secrets and internal information.
[0008] Commercial centers and schools: Shops, warehouses, classrooms and other areas need to be protected to prevent theft, vandalism and accidental intrusion, and to ensure the safety of property and personnel.
[0009] Private residences: Security in residential areas is equally important to prevent strangers or uninvited guests from entering private property and to protect the privacy and safety of residents.
[0010] Although traditional physical isolation measures provide basic security, they are powerless in the face of the complex and ever-changing security needs in modern society. For example, fences and door locks can be breached, and guards and patrols also have omissions and blind spots. The prohibited area entry system based on target detection technology can monitor, automatically identify and alarm in real time in a large range and complex environment through intelligent means, greatly improving the efficiency and reliability of security protection.
[0011] The prohibited area entry prohibition system based on target detection technology is usually composed of the following core modules:
[0012] Image acquisition module: Use cameras or other sensors to continuously monitor the target area and collect high-quality image or video data. These cameras can be fixed or movable to cover a wider monitoring range.
[0013] Data preprocessing module: preprocess the collected images or videos, including filtering, contrast enhancement, brightness adjustment and other steps to improve image quality and enhance target features.
[0014] Object detection module: Use advanced object detection algorithms (such as YOLO, SSD, Faster R-CNN, etc.) to identify and locate objects in preprocessed images. This module not only requires high accuracy, but also real-time processing capabilities to ensure timely response when abnormal situations are detected.
[0015] Feature matching module: compares and matches the detected target features with the predefined prohibited area pattern library to determine whether the target is within the prohibited area. The accuracy of feature matching is directly related to the false alarm rate and missed alarm rate of the system.
[0016] Alarm and response module: When illegal entry is detected, the system automatically triggers an alarm and sends a notification to relevant managers. At the same time, the system can also be linked with other security protection measures (such as access control systems, monitoring centers) to take further countermeasures.
[0017] The prohibited area entry prohibition system based on target detection technology has shown broad application prospects in many fields. Its intelligent and automated features enable the system to provide efficient and reliable security protection in various complex environments. However, in practical applications, the system also faces some challenges:
[0018] Environmental changes: Lighting conditions, weather changes, background complexity, etc. will affect the accuracy of target detection. The system needs to have good adaptability and be able to operate stably under various environmental conditions.
[0019] Computing resources: Real-time processing of large amounts of high-resolution images or video data requires powerful computing resources, especially on mobile devices or in resource-constrained scenarios. Algorithms and system architectures need to be optimized to balance performance and resource consumption.
[0020] Privacy protection: When deploying cameras and surveillance systems in public places, how to balance security needs and personal privacy protection is an issue that needs to be carefully considered. System design needs to comply with relevant laws and regulations, and take appropriate technical measures to protect personal privacy.
[0021] The prohibited area entry prohibition system based on target detection technology combines modern computer vision and artificial intelligence technology to provide advanced solutions for public safety and privacy protection. With the continuous development and improvement of technology, such systems will gradually be applied in more fields and become an important part of intelligent security protection. Summary of the invention
[0022] The purpose of the present invention is to combine computer vision, image processing and pattern matching technologies to provide an efficient, all-weather automatic solution for prohibited area target detection, which has the advantages of low cost and high efficiency compared with the traditional manual coverage method. The purpose of the present invention is achieved through the following technical solutions:
[0023] A method for realizing prohibited area entry detection under different environments, characterized in that it comprises the following steps:
[0024] S1: Data collection;
[0025] S2: data preprocessing;
[0026] S3: feature extraction;
[0027] S4: feature processing and matching;
[0028] S5: Real-time monitoring and alarm
[0029] Further, the data collection in step S1 includes environmental scanning and prohibited area marking; wherein, environmental scanning is to use a camera or other sensor to scan the area to be monitored and collect environmental images or video data; prohibited area marking is to mark the location and range of the prohibited area in the collected image or video to establish a reference data set. In the prohibited area entry detection system, environmental scanning is a basic step. This process requires the deployment of cameras or sensors to continuously monitor specific areas. According to the characteristics of the environment, select appropriate hardware equipment to ensure that it can cover all key areas; in addition, the system must have the ability to handle different lighting conditions (such as daytime, night, shadows, etc.) to ensure effective monitoring around the clock. The goal of environmental scanning is to collect clear, stable and continuous image or video data to provide a reliable basis for subsequent processing steps. After collecting the environmental data, the prohibited area needs to be marked next. This step is crucial because it directly affects the detection accuracy of the system. The marking process can be completed manually or by automated tools. Manual marking requires experts or administrators to accurately mark the location and range of the prohibited area in the image or video frame, and these marking information will serve as reference data for the system. The automated annotation tool uses predefined rules and algorithms to identify and mark potential prohibited areas. Whether it is manual or automatic annotation, the results need to be verified and adjusted to ensure the accuracy and comprehensiveness of the annotation. The annotation information not only includes the position and range on the two-dimensional plane, but also the height and depth information in the three-dimensional space, especially in scenes where different floors or three-dimensional structures need to be distinguished.
[0030] Furthermore, the data preprocessing in step S2 includes image conversion, filtering, contrast enhancement and brightness adjustment; wherein, the image conversion converts the original image or video into a format suitable for processing. The original data may come from different types of cameras or sensors, and their formats, resolutions, color spaces, etc. may be different. Therefore, image conversion is an important step to unify data formats and improve processing efficiency. Common conversion operations include adjusting resolution, changing color space (such as converting from RGB to grayscale image or YUV format), and image cropping and scaling. These conversions not only improve the compatibility of subsequent processing algorithms, but also reduce the amount of calculation and enhance the real-time processing capability of the system. The filtering is to apply filtering technology to remove noise and improve image quality. Noise may come from the camera's photosensitive element, changes in ambient light, interference during transmission, etc. Common filtering methods include mean filtering, median filtering, Gaussian filtering, etc. Mean filtering smoothes the image by taking the average value of surrounding pixels, but it may blur the edges; median filtering removes noise by taking the median value of surrounding pixels, which has a better effect on retaining edges; Gaussian filtering uses Gaussian function to perform weighted averaging on pixels, which can smooth noise while maintaining the overall structure of the image. When selecting a filtering method, it is necessary to optimize it according to the specific application scenario and noise type to strike a balance between noise removal and detail retention. The contrast enhancement and brightness adjustment are to adjust the image contrast and brightness to make the features more obvious.
[0031] Furthermore, the feature extraction in step S3 includes image segmentation and target detection; wherein, the image segmentation is to divide the image into several regions for further processing. Commonly used image segmentation methods include threshold segmentation, edge detection, region growing and watershed algorithm. Threshold segmentation is based on the threshold of pixel gray value for division, which is suitable for images with obvious contrast between background and foreground; edge detection divides the image into different parts by identifying edge pixels in the image; region growing starts from the seed point and gradually expands the region according to pixel similarity; the watershed algorithm uses the gradient information of the image to divide the image into regions with significant boundaries. Through image segmentation, the target of interest can be separated from the background, laying the foundation for subsequent target detection and feature extraction.
[0032] The target detection is to use target detection algorithms (such as YOLO, SSD, Faster R-CNN, etc.) to identify objects in the image and extract features such as its contour, position and size. Common target detection algorithms include YOLO (You Only Look Once), SSD (Single Shot MultiBox Detector) and Faster R-CNN (Region Convolutional Neural Networks). These algorithms process images through convolutional neural networks (CNN), extract features and identify targets in the image. YOLO is a real-time target detection algorithm that divides the image into several grids and predicts the category and position of the target in each grid; SSD is also a real-time detection algorithm that detects through multi-scale feature maps and can process targets of different sizes at the same time; Faster R-CNN generates candidate regions through a region proposal network (RPN) and performs target detection in these regions, with high detection accuracy.
[0033] Furthermore, the feature processing and matching in step S4 includes feature fitting and feature matching; wherein, the feature fitting is to fit and process the extracted target features to eliminate noise and errors. Common feature fitting methods include least squares method, polynomial fitting and curve fitting. The least squares method fits data points by minimizing the sum of squared errors, and is applicable to linear and nonlinear models; polynomial fitting fits data by polynomial functions, and can describe complex feature changes; curve fitting fits data according to a specific curve model, and is applicable to features with regular changes. Through feature fitting, more accurate and stable feature parameters can be obtained, reducing deviations caused by detection errors and noise, and providing reliable data support for feature matching.
[0034] Feature matching is to compare and match the processed features with the predefined prohibited area pattern library to determine whether the target is in the prohibited area. Common methods include template matching, feature point matching and machine learning algorithms. Template matching compares the target features with the predefined template one by one and calculates the similarity to determine the degree of matching; feature point matching extracts the feature points of the target and the template and uses descriptors for matching, which is suitable for complex backgrounds and deformed targets; machine learning algorithms train classifiers to determine whether the target belongs to the prohibited area based on the input features. The results of feature matching will be used to trigger the alarm system and further processing steps to ensure the accuracy and reliability of detection.
[0035] Furthermore, the step S5 real-time monitoring and alarm includes a real-time processing system and an alarm system; wherein the real-time processing system includes continuously monitoring the input image or video stream, and performing the processing and detection of the above steps in real time; the alarm system triggers the alarm system when the target enters the prohibited area, and sends a notification or alarm to the management personnel. Real-time processing is one of the core functions of the system, which aims to continuously monitor the input image or video stream, and perform data preprocessing, feature extraction and feature matching in real time. Real-time processing requires efficient algorithms and powerful computing power to ensure that the system can complete all processing tasks within milliseconds. In order to achieve real-time performance, parallel computing, hardware acceleration (such as GPU acceleration) and optimization algorithms can be used. At the same time, real-time processing also needs to have the ability to respond quickly to environmental changes, such as changes in illumination, moving targets, etc., to ensure stable operation under various conditions. The results of real-time processing will be used for subsequent alarm triggering and user interface display, providing timely monitoring information and response measures. The alarm system is an important part of real-time monitoring. When the target is detected to enter the prohibited area, the system will trigger an alarm and send a notification or alarm signal to the management personnel. Alarm systems can take many forms, including audible alarms, visual alarms (such as flashing lights), text messages or push notifications, etc. The design of the alarm system needs to take into account response speed, reliability and user experience to ensure that the alarm is issued as soon as the target enters the prohibited area. In addition, the alarm system can also be integrated with other security systems (such as access control systems, monitoring centers) to achieve comprehensive security management. With timely alarms, managers can take quick action to prevent unauthorized entry and ensure regional security.
[0036] In addition, the present application also includes a computer device for implementing prohibited area entry detection in different environments, characterized in that it includes a memory, a processor and a computer program stored on the memory, and the processor executes the computer program to implement a prohibited area entry detection method in different environments.
[0037] In addition, the present application also includes a computer program product for implementing prohibited area entry detection in different environments, characterized in that it includes a computer program / instructions, which, when executed by a processor, implement a prohibited area entry detection method in different environments.
[0038] A computer-readable storage medium, characterized in that when the computer program / instructions are executed by a processor, a prohibited area entry detection method is implemented under different environments.
[0039] The beneficial effects of the present invention are:
[0040] 1. Improve detection accuracy: Combine image preprocessing, feature extraction, and feature matching to effectively reduce noise and errors and improve the detection accuracy of the system.
[0041] 2. Real-time and efficiency: Through efficient image preprocessing, feature extraction and matching algorithms, the system can continuously monitor the input images or video streams, and perform detection and alarm triggering in real time to ensure timely response.
[0042] 3. Multiple alarm forms: The system has multiple alarm triggering methods, including sound alarms, visual alarms (such as flashing lights), SMS or push notifications, etc., to ensure that managers are notified in time when the target enters the prohibited area.
[0043] 4. Comprehensive security management: Through integration with other security systems such as access control systems and monitoring centers, comprehensive security management is achieved. When the system detects that a target enters a prohibited area, it can quickly issue an alarm and notify the management personnel.
[0044] 5. Reduce false positives and negatives: Combine advanced image segmentation and target detection algorithms (such as YOLO, SSD, FasterR-CNN, etc.) to accurately identify targets and extract their features. Through feature fitting and matching, reduce the deviation caused by detection errors and noise, and effectively reduce the probability of false positives and negatives. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] Figure 1 A technical flow chart of the present invention;
[0046] Figure 2 This is a schematic diagram of the results of detecting the forbidden area according to the present invention;
[0047] Figure 3 A schematic diagram of the result of pedestrian detection according to the present invention;
[0048] Figure 4 It is a schematic diagram of the result of detecting entry into a prohibited area according to the present invention. DETAILED DESCRIPTION
[0049] In order to better understand the technical content of the present invention, specific embodiments are given and described as follows in conjunction with the accompanying drawings.
[0050] like Figure 1 As shown, the method and system for implementing prohibited area entry detection in different environments provided by the present invention include the following steps:
[0051] S1: Data collection, including environmental scanning: Use cameras or other sensors to scan the area to be monitored and collect environmental images or video data. Prohibited area annotation: Annotate the location and range of prohibited areas in the collected images or videos to establish a reference data set.
[0052]
[0053] S2: Data preprocessing, including image conversion: converting raw images or videos into a format suitable for processing. Filtering: applying filtering techniques to remove noise and improve image quality. Contrast enhancement and brightness adjustment: adjusting image contrast and brightness to make features more obvious.
[0054]
[0055] S3: Feature extraction, including image segmentation: Divide the image into several regions for further processing. Object detection: Use object detection algorithms (such as YOLO, SSD, Faster R-CNN, etc.) to identify objects in the image and extract their features such as outline, position and size.
[0056]
[0057] S4: Feature processing and matching, including feature fitting: fitting and processing the extracted target features to eliminate noise and errors. Feature matching: comparing and matching the processed features with the predefined prohibited area pattern library to determine whether the target is located in the prohibited area.
[0058]
[0059] S5: Real-time monitoring and alarm, including real-time processing: Continuously monitor the input image or video stream, and perform the above steps of processing and detection in real time. Alarm system: When the target is detected entering the prohibited area, the alarm system is triggered to send a notification or alarm to the management personnel.
[0060]
[0061] Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. A person with ordinary knowledge in the technical field to which the present invention belongs may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be determined by the definition of the claims.
Claims
1. A method for realizing prohibited area entry detection in different environments, characterized in that: The following steps are involved: S1: Data collection; S2: data preprocessing; S3: feature extraction; S4: feature processing and matching; S5: Real-time monitoring and alarm.
2. The method for realizing prohibited area entry detection under different environments according to claim 1, characterized in that: The data collection in step S1 includes environmental scanning and prohibited area marking; wherein, environmental scanning is to use a camera or other sensor to scan the area to be monitored and collect environmental images or video data; prohibited area marking is to mark the location and range of the prohibited area in the collected image or video to establish a reference data set.
3. The method for realizing prohibited area entry detection under different environments according to claim 1, characterized in that: The data preprocessing in step S2 includes image conversion, filtering, contrast enhancement and brightness adjustment; wherein the image conversion is to convert the original image or video into a format suitable for processing; the filtering is to apply filtering technology to remove noise and improve image quality; the contrast enhancement and brightness adjustment are to adjust the image contrast and brightness to make the features more obvious.
4. The method for realizing prohibited area entry detection under different environments according to claim 1, characterized in that: The feature extraction in step S3 includes image segmentation and target detection; wherein the image segmentation is to divide the image into several regions for further processing; the target detection is to use a target detection algorithm (such as YOLO, SSD, Faster R-CNN, etc.) to identify the object in the image and extract its contour, position, size and other features.
5. The method for realizing prohibited area entry detection under different environments according to claim 1, characterized in that: The feature processing and matching in step S4 includes feature fitting and feature matching; wherein, the feature fitting is to fit and process the extracted target features to eliminate noise and errors; the feature matching is to compare and match the processed features with a predefined prohibited area pattern library to determine whether the target is located in the prohibited area.
6. The method for realizing prohibited area entry detection under different environments according to claim 1, characterized in that: The step S5 real-time monitoring and alarm includes a real-time processing system and an alarm system; wherein the real-time processing system includes continuously monitoring the input image or video stream, and performing the processing and detection of the above steps in real time; the alarm system is to trigger the alarm system when it detects that the target enters the prohibited area, and send a notification or alarm to the management personnel.
7. A computer device for realizing forbidden area entry detection in different environments, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the detection method according to claims 1 to 6.
8. A computer program product for implementing prohibited area entry detection in different environments, characterized in that: The invention comprises a computer program / instruction, which implements the detection method according to claims 1 to 6 when executed by a processor.
9. A computer-readable storage medium, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.
