Digital twinning and CNN-based active safety identification system and method for crowded area

Abstract

The invention discloses an active safety identification system and method for a crowded area based on digital twinning and CNN. The system comprises a video acquisition device, a preprocessing module, a CNN model, a digital twinning platform, an early warning module and an information intercommunication module. The video acquisition equipment is a high-definition camera; the preprocessing module is used for carrying out denoising and contrast enhancement processing on video data acquired by the video acquisition equipment; the CNN model is a convolutional neural network model and is used for performing video analysis and recognition on the video data subjected to denoising and contrast enhancement processing; the digital twin platform is deployed at a cloud end and is used for training and optimizing a CNN model; the early warning module is used for operating the CNN model, performing video analysis on the video data in real time, and sending out an early warning signal when a potential safety hazard is identified; and the information intercommunication module pushes the early warning signal to family members of the user and security and public security systems of areas where the early warning signal appears. According to the invention, potential safety hazards in crowded areas are actively identified, and travel safety is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic safety, and particularly to an active safety identification system and method for crowded areas based on digital twin and CNN. Background Art

[0002] Traffic safety accidents in crowded areas occur frequently. A series of tragic cases all highlight the severity of the problem. Traffic safety is directly related to life and health, especially during the period of going to and from school around campuses, where traffic safety hazards are particularly prominent.

[0003] Existing safety management methods often rely on manual patrols and post - event playback of surveillance videos, and cannot achieve real - time monitoring and active early warning.

[0004] With the continuous development of artificial intelligence technology, especially the wide application of convolutional neural network (CNN) in image and digital twin technologies, it provides a new solution direction for safety monitoring in crowded areas.

[0005] However, there is no public disclosure of using digital twin and CNN technologies for pedestrian traffic management in the existing technology.

[0006] Therefore, how to apply digital twin and CNN technologies to pedestrian traffic management, actively identify potential safety hazards in crowded areas, and take preventive measures more quickly and accurately to improve travel safety has become a technical problem that needs to be solved urgently by those skilled in the art. Summary of the Invention

[0007] In view of the above - mentioned defects of the existing technology, the present invention provides an active safety identification system and method for crowded areas based on digital twin and CNN. The purpose is to actively identify potential safety hazards in crowded areas by collecting and analyzing video data in real time, take preventive measures more quickly and accurately, and improve travel safety.

[0008] To achieve the above - mentioned purpose, the present invention discloses an active safety identification system for crowded areas based on digital twin and CNN, including a video acquisition device, a pre - processing module, a CNN model, a digital twin platform, an early warning module, and an information inter - communication module;

[0009] The video acquisition device is a high - definition camera set on the user's wearable clothing;

[0010] The pre - processing module is used to perform denoising and contrast enhancement processing on the video data collected by the video acquisition device;

[0011] The CNN model is a convolutional neural network model used to perform video analysis and recognition on the video data that has completed denoising and contrast enhancement processing;

[0012] The digital twin platform is deployed in the cloud and is used for the training and optimization of the CNN model;

[0013] The early warning module is used to run the CNN model, perform video analysis on the video data in real time, and send out an early warning signal when a potential safety hazard is identified;

[0014] The information communication module pushes the early warning signal to the families of the users, the security and public security systems in the area where the early warning signal appears through the mobile APP and wireless communication technology.

[0015] The present invention also provides an active safety identification method, which is applied to the active safety identification system for crowded areas based on digital twin and CNN as described above, and includes the following steps:

[0016] Step 1: Acquisition and preprocessing of video data;

[0017] Step 2: Video analysis and identification based on the CNN model;

[0018] Step 3: Early warning after potential safety hazard identification;

[0019] Step 4: Information communication and prevention measures.

[0020] Preferably, Step 1 is specifically as follows:

[0021] Step 1.1: Setting of the video data acquisition device, specifically: installing a lightweight and high-definition camera with voice interaction function on the clothes worn by the user, and collecting the video data in real time through the high-definition camera;

[0022] Step 1.2: Preprocessing of the video data, specifically including: denoising processing and contrast enhancement;

[0023] Adopt Gaussian filtering or wavelet transform denoising processing algorithm to remove the noise in the video data and improve the video quality;

[0024] The contrast enhancement improves the contrast of the video data through a contrast enhancement algorithm, making the image clearer.

[0025] More preferably, the characteristics of the high-definition camera include: light weight, support for high-resolution high-definition video acquisition and voice control function.

[0026] More preferably, the high-definition camera has the function of remotely setting the opening and closing range. After the opening and closing conditions are set, when the high-definition camera enters the preset area, it can automatically power on and work, specifically including area setting and automatic triggering;

[0027] The area setting refers to setting the range within which the high-definition camera automatically powers on and operates through a mobile APP or management platform paired with the high-definition camera.

[0028] The automatic trigger means that when the high-definition camera enters the preset range where it needs to automatically power on and operate, the high-definition camera automatically turns on, and when the high-definition camera leaves the preset range where it needs to automatically power on and operate, the high-definition camera automatically turns off.

[0029] More preferably, step 2 is specifically as follows:

[0030] Step 2.1, the architecture of the CNN model is specifically: locally parsing and identifying the video data using a convolutional neural network (CNN);

[0031] The CNN automatically extracts features from the video data through convolutional layers, pooling layers, and fully connected layers. The specific architecture is as follows:

[0032] The convolutional layer is used to extract local features from the video data;

[0033] The pooling layer is used to reduce the feature dimension and computational amount;

[0034] The fully connected layer is used to classify and identify the extracted local features;

[0035] Step 2.2, the training and optimization of the CNN model is specifically: continuously training and optimizing the algorithms and key parameters in the CNN model based on digital twin technology to improve the accuracy of the algorithms. The specific steps are as follows:

[0036] Step 2.2.1, set up the digital twin platform in the cloud and improve the capabilities of the digital twin platform by continuously collecting a large number of samples;

[0037] Step 2.2.2, train the CNN model using a large amount of labeled data to improve the recognition accuracy of the CNN model;

[0038] Step 2.2.3, continuously optimize the algorithms and key parameters of the CNN model through the digital twin technology to improve the real-time performance and accuracy of the CNN model;

[0039] Step 2.3, identify potential safety hazards;

[0040] According to the recognition results of the CNN model, conduct real-time analysis of potential safety hazards in the video data and mark the specific locations and types of potential safety hazards;

[0041] The potential safety hazards include:

[0042] The vehicle collides with the guardrail, that is, it identifies and judges whether the vehicle will collide with the guardrail;

[0043] The speed is too fast after entering the crosswalk, that is, it identifies and judges whether the driving speed of the vehicle on the crosswalk is too fast;

[0044] Other potential safety hazards, that is, it identifies whether there are other potential safety hazards that may pose a threat to personnel;

[0045] The real-time analysis refers to the real-time analysis of the video data, identifying potential safety hazards and marking the hazards;

[0046] The hazard marking is to mark the specific location and type of the hazard.

[0047] More preferably, the other potential safety hazards include the vehicle entering the restricted area.

[0048] More preferably, in step 3, according to the recognition result of the safety hazard by the CNN model, when the safety hazard is recognized, a warning signal is issued.

[0049] More preferably, step 4 includes mobile phone remote collaboration;

[0050] The mobile phone remote collaboration refers to remotely controlling the camera switch, adjusting the angle, viewing the video in real time and remotely shouting, specifically including: remote control, real-time viewing and remote shouting;

[0051] The remote control is to remotely control the switch and angle adjustment of the camera of the corresponding mobile phone through the APP;

[0052] The real-time viewing refers to real-time viewing of the video data collected by the camera;

[0053] The remote shouting is to remotely shout through the APP to remind of paying attention to safety.

[0054] More preferably, step 4 further includes warning information push and public security system access;

[0055] The warning information push and public security system access refers to sending the warning information to the mobile phone end through wireless communication technology, and at the same time accessing the security room and the public security emergency command system, so as to respond to safety accidents in a timely manner, specifically including: warning push, security access and public security system access;

[0056] The warning push is to send the warning information to the paired mobile phone APP through wireless communication technology;

[0057] The security access means that the warning information is simultaneously accessed to the security, so that the security responds and takes necessary on-site management measures;

[0058] The access to the public security system refers to accessing the early warning information to the public security emergency command system and requesting the public security department to promptly dispatch police forces for intervention.

[0059] Advantages of the present invention:

[0060] The present invention has high accuracy and can continuously optimize the CNN model through digital twin technology to improve the accuracy of identifying potential safety hazards.

[0061] The present invention has real-time performance and can analyze video data in real time to promptly discover and give early warnings of potential safety hazards.

[0062] The present invention has an information intercommunication function and can, through the mobile APP and wireless communication technology, achieve real-time information intercommunication in all aspects and improve the emergency response speed.

[0063] The present invention is user-friendly. The lightweight and voice-interactive high-definition camera is comfortable to wear and easy to operate.

[0064] The present invention has high flexibility. The opening and closing areas of the camera can be set remotely to flexibly meet the requirements of different scenarios.

[0065] The concept, specific structure and technical effects of the present invention will be further described below in conjunction with the drawings to fully understand the purpose, features and effects of the present invention. Description of the Drawings

[0066] Figure 1 A working schematic diagram showing an embodiment of the present invention is shown. Detailed Embodiments

[0067] Embodiment

[0068] As Figure 1 shown, the active safety recognition system for crowded areas based on digital twin and CNN includes a video acquisition device, a preprocessing module, a CNN model, a digital twin platform, an early warning module and an information intercommunication module;

[0069] The video acquisition device is a high-definition camera disposed on the clothing worn by the user;

[0070] The preprocessing module is used to perform denoising and contrast enhancement processing on the video data acquired by the video acquisition device;

[0071] The CNN model is a convolutional neural network model used to perform video parsing and recognition on the video data that has completed denoising and contrast enhancement processing;

[0072] The digital twin platform is deployed in the cloud and is used for the training and optimization of the CNN model;

[0073] The early warning module is used to run the CNN model, perform video analysis on video data in real time, and issue an early warning signal when a potential safety hazard is identified.

[0074] The information sharing module uses the mobile APP and wireless communication technology to push the early warning signal to the families of the users, as well as the security and public security systems in the area where the early warning signal appears.

[0075] Through real-time video analysis and proactive early warning, the present invention takes preventive measures in advance, effectively improves the level of traffic safety management, enhances the sense of security during travel, improves the efficiency of safety management, strengthens the supervision of the public security department, helps to reduce casualties and property losses caused by vicious collision incidents, and has broad application prospects and important social value.

[0076] The present invention also provides an active safety recognition method, which applies the active safety recognition system for crowded areas based on digital twin and CNN as described above, and includes the following steps:

[0077] Step 1: Collection and preprocessing of video data;

[0078] Step 2: Video analysis and recognition based on the CNN model;

[0079] Step 3: Early warning after potential safety hazard recognition;

[0080] Step 4: Information sharing and prevention measures.

[0081] In some embodiments, Step 1 is specifically as follows:

[0082] Step 1.1: Setting of the video data collection device, specifically: installing a lightweight and high-definition camera with voice interaction function on the clothing worn by the user, and collecting video data in real time through the high-definition camera;

[0083] Step 1.2: Preprocessing of the video data, specifically including: denoising processing and contrast enhancement;

[0084] Adopt Gaussian filtering or wavelet transform denoising processing algorithm to remove the noise in the video data and improve the video quality;

[0085] Contrast enhancement uses the contrast enhancement algorithm to increase the contrast of the video data and make the image clearer.

[0086] In practical applications, denoising processing and contrast enhancement can improve the accuracy of subsequent video analysis.

[0087] In some embodiments, the characteristics of the high-definition camera include: light weight, support for high-resolution high-definition video collection and voice control function.

[0088] In practical applications, the lightweight design of the high-definition camera makes it light in weight, comfortable to wear, does not affect the daily activities of the user, supports high-resolution high-definition video acquisition, can ensure clear images, and the voice control function enables the user to easily turn on or off the camera when needed.

[0089] In some embodiments, the high-definition camera has the function of remotely setting the on and off range. After setting the on and off conditions, when the high-definition camera enters the preset area, it can automatically power on and work, specifically including area setting and automatic triggering;

[0090] The area setting refers to setting the range where the high-definition camera automatically powers on and works through the mobile APP or management platform paired with the high-definition camera;

[0091] Automatic triggering means that when the high-definition camera enters the preset range where it needs to automatically power on and work, the high-definition camera automatically turns on, and when the high-definition camera leaves the preset range where it needs to automatically power on and work, the high-definition camera automatically turns off.

[0092] In some embodiments, step 2 is specifically as follows:

[0093] Step 2.1, the architecture of the CNN model, specifically: using a convolutional neural network (CNN) to parse and identify video data;

[0094] The CNN automatically extracts features in the video data through convolutional layers, pooling layers, and fully connected layers. The specific architecture is as follows:

[0095] The convolutional layer is used to extract local features in the video data;

[0096] The pooling layer is used to reduce the feature dimension and reduce the amount of calculation;

[0097] The fully connected layer is used to classify and identify the extracted local features;

[0098] Step 2.2, the training and optimization of the CNN model, specifically: continuously training and optimizing the algorithms and key parameters in the CNN model based on digital twin technology to improve the accuracy of the algorithms. The specific steps are as follows:

[0099] Step 2.2.1, set the digital twin platform in the cloud, and continuously collect a large number of samples to enhance the capabilities of the digital twin platform;

[0100] Step 2.2.2, use a large amount of labeled data to train the CNN model to improve the recognition accuracy of the CNN model;

[0101] Step 2.2.3, through digital twin technology, continuously optimize the algorithms and key parameters of the CNN model to improve the real-time performance and accuracy of the CNN model;

[0102] Step 2.3, Safety hazard identification;

[0103] Based on the recognition results of the CNN model, conduct real-time analysis of safety hazards in the video data, and mark the specific locations and types of safety hazards;

[0104] Safety hazards include:

[0105] A vehicle ramming into the guardrail, that is, identifying and judging whether the vehicle will ram into the guardrail;

[0106] Excessive speed after driving into the crosswalk, that is, identifying and judging whether the driving speed of the vehicle on the crosswalk is too fast;

[0107] Other safety hazards, that is, identifying whether there are other safety hazards that may pose a threat to personnel;

[0108] Real-time analysis means real-time analysis of video data, identifying potential safety hazards, and marking the hazards;

[0109] Hazard marking means marking the specific locations and types of hazards.

[0110] In practical applications, marking the specific locations and types of hazards can facilitate relevant personnel to take measures in a timely manner.

[0111] In some embodiments, other safety hazards include a vehicle driving into a restricted area.

[0112] In some embodiments, in step 3, according to the recognition results of the CNN model for safety hazards, when a safety hazard is recognized, a warning signal is issued.

[0113] In some embodiments, step 4 includes mobile phone remote collaboration;

[0114] Mobile phone remote collaboration means remotely controlling the camera switch, adjusting the angle, viewing the video in real time, and remotely shouting through the remote control, specifically including: remote control, real-time viewing, and remote shouting;

[0115] Remote control is to remotely control the switch and angle adjustment of the camera of the corresponding mobile phone through the APP;

[0116] Real-time viewing means real-time viewing of the video data collected by the camera;

[0117] Remote shouting means remotely shouting through the APP to remind to pay attention to safety.

[0118] In some embodiments, step 4 further includes warning information push and public security system access;

[0119] Early warning information push and public security system access refer to sending early warning information to the mobile phone terminal through wireless communication technology, and at the same time accessing the security room and the public security emergency command system to respond to safety accidents in a timely manner. Specifically, it includes: early warning push, security access, and public security system access;

[0120] Early warning push means sending early warning information to the paired mobile phone APP through wireless communication technology;

[0121] Security access means that the physical early warning information is simultaneously accessed to the security system, enabling security personnel to respond in real time and take necessary on-site management measures;

[0122] Public security system access means accessing early warning information to the public security emergency command system and requesting the public security department to dispatch police forces in a timely manner for intervention.

[0123] The preferred specific embodiments of the present invention have been described in detail above. It should be understood that those of ordinary skill in the art can make many modifications and variations based on the concept of the present invention without creative work. Therefore, all technical solutions that can be obtained by those skilled in the art in the technical field of the present invention through logical analysis, reasoning, or limited experiments based on the concept of the present invention on the basis of the prior art shall fall within the protection scope determined by the claims.

Claims

1. Active safety identification system for crowded areas based on digital twins and CNN; characterized by: It includes video acquisition equipment, preprocessing module, CNN model, digital twin platform, early warning module and information exchange module; The video acquisition device is a high-definition camera installed on the clothes worn by the user; The preprocessing module is used to perform denoising and contrast enhancement processing on the video data collected by the video acquisition device; The CNN model is a convolutional neural network model used to perform video analysis and recognition on the video data that has undergone denoising and contrast enhancement processing; The digital twin platform is deployed in the cloud for training and optimization of the CNN model; The early warning module is used to run the CNN model, perform video analysis on the video data in real time, and issue an early warning signal when a safety hazard is identified; The information exchange module pushes the warning signal to the user's family, the security and public security systems in the area where the warning signal appears through the mobile phone APP and wireless communication technology.

2. Active safety identification method, characterized in that: The application of the active safety identification system for crowded areas based on digital twins and CNN as claimed in claim 1 comprises the following steps: Step 1: Video data collection and preprocessing; Step 2: Video parsing and recognition based on CNN model; Step 3: Early warning after safety hazard identification; Step 4: Information exchange and prevention and control measures.

3. The active safety identification method according to claim 2, characterized in that: Step 1 is as follows: Step 1.1, setting the video data acquisition device, specifically: installing a lightweight high-definition camera with voice interaction function on the user's wearable clothing, and collecting the video data in real time through the high-definition camera; Step 1.2: preprocessing the video data, specifically including: denoising and contrast enhancement; Using Gaussian filtering or wavelet transform denoising processing algorithm to remove noise in the video data and improve video quality; The contrast enhancement improves the contrast of the video data through a contrast enhancement algorithm to make the image clearer.

4. The active safety identification method according to claim 3, characterized in that: The high-definition camera has the following features: light weight, support for high-resolution high-definition video acquisition and voice control functions.

5. The active safety identification method according to claim 3, characterized in that: The high-definition camera has the function of remotely setting the opening and closing range. After the opening and closing conditions are set, the high-definition camera can automatically start working when entering a preset area, specifically including area setting and automatic triggering; The area setting refers to setting the range of automatic startup of the high-definition camera through the mobile phone APP or management platform paired with the high-definition camera; The automatic trigger means that when the high-definition camera enters a preset range where automatic power-on is required, the high-definition camera automatically turns on, and when the high-definition camera leaves the preset range where automatic power-on is required, the high-definition camera automatically turns off.

6. The active safety identification method according to claim 5, characterized in that: Step 2 is as follows: Step 2.1, the architecture of the CNN model, specifically: using a convolutional neural network (CNN) to parse and identify the video data; The CNN automatically extracts features from the video data through convolutional layers, pooling layers, and fully connected layers. The specific architecture is as follows: The convolution layer is used to extract local features in the video data; The pooling layer is used to reduce the feature dimension and reduce the amount of calculation; The fully connected layer is used to classify and identify the extracted local features; Step 2.2, training and optimization of the CNN model, specifically: continuously training and optimizing the algorithm and key parameters in the CNN model based on digital twin technology to improve the accuracy of the algorithm. The specific steps are as follows: Step 2.2.1, setting up a digital twin platform in the cloud, and improving the capabilities of the digital twin platform by continuously collecting a large number of samples; Step 2.2.2, using a large amount of labeled data to train the CNN model to improve the recognition accuracy of the CNN model; Step 2.2.3, continuously optimizing the algorithm and key parameters of the CNN model through the digital twin technology to improve the real-time performance and accuracy of the CNN model; Step 2.3: Identify safety hazards; According to the recognition result of the CNN model, the potential safety hazards in the video data are analyzed in real time, and the specific location and type of the potential safety hazards are marked; The safety hazards include: Vehicle collision with guardrail, i.e. identifying and judging whether the vehicle will collide with the guardrail; Driving too fast after entering the crosswalk, that is, identifying and judging whether the vehicle is driving too fast on the crosswalk; Other safety hazards, i.e. identifying whether there are other safety hazards that may pose a threat to personnel; The real-time analysis refers to analyzing the video data in real time, identifying potential safety hazards, and marking them; The hidden danger mark refers to marking the specific location and type of the hidden danger.

7. The active safety identification method according to claim 6, characterized in that: The other safety hazards mentioned include vehicles entering prohibited areas.

8. The active safety identification method according to claim 6, characterized in that: In step 3, based on the recognition result of the safety hazard by the CNN model, a warning signal is issued when the safety hazard is identified.

9. The active safety identification method according to claim 8, characterized in that: The step 4 includes remote collaboration on the mobile phone side; The mobile phone remote collaboration refers to remotely controlling the camera switch, adjusting the angle, viewing the video in real time and remote shouting, specifically including: remote control, real-time viewing and remote shouting; The remote control is to remotely control the switch and angle adjustment of the camera of the corresponding mobile phone through the APP; The real-time viewing refers to real-time viewing of video data collected by the camera; The remote shouting refers to remote shouting through the APP to remind people to pay attention to safety.

10. The active safety identification method according to claim 8, characterized in that: The step 4 also includes early warning information push and public security system access; The warning information push and public security system access refer to sending the warning information to the mobile phone through wireless communication technology, and accessing the security room and public security emergency command system at the same time, so as to respond to security incidents in a timely manner, specifically including: warning push, security access and public security system access; The warning push refers to sending the warning information to the paired mobile phone APP through wireless communication technology; The security access entity simultaneously accesses the warning information to the security, so that the security responds and takes necessary on-site management measures; The public security system access refers to accessing the early warning information to the public security emergency command system, requesting the public security department to dispatch police forces to intervene in a timely manner.

Images