A campus anti-bullying edge computing system and method based on semantic recognition and image recognition technology

Abstract

The present invention discloses a campus anti-bullying edge computing system and method based on semantic recognition and image recognition technology. The system includes hardware modules, software modules and a cloud-edge collaborative integrated platform. The hardware modules include: a semantic recognition module, which collects voice data in the campus through a voice collection sensor; an image recognition module, which collects image data through a camera; an edge computing module, which recognizes and analyzes the data transmitted by the semantic recognition module and the image recognition module in real time, detects bullying behavior, and triggers an alarm module to alarm after bullying behavior is found; an alarm module, which sounds an alarm and sends the alarm location to the administrator. It can detect and prevent campus bullying behavior in real time and accurately. At the same time, the system adopts edge computing to sink the semantic recognition and image recognition algorithms into the device for real-time edge computing, and can process and analyze data locally, reduce dependence on cloud resources, and improve the real-time and reliability of the system.

Description

Technical Field

[0001] The present invention relates to the fields of natural speech processing, image recognition, and edge computing technology, and specifically to a campus anti-bullying edge computing system and method based on semantic recognition and image recognition technology. Background Art

[0002] School bullying is an increasingly serious social problem that causes great harm to students' physical and mental health. Although existing campus surveillance systems can record videos, traditional campus security surveillance systems often rely on cloud computing, which has problems such as high latency and poor real-time performance. At the same time, school bullying incidents often occur in more hidden and private locations, resulting in failure to detect and intervene in time. Summary of the invention

[0003] The purpose of the present invention is to provide a campus anti-bullying edge computing system and method based on semantic recognition and image recognition technology in response to the above-mentioned problems. The system of the present invention can detect and prevent campus bullying behavior in real time and accurately, and adopts edge computing to sink the semantic recognition and image recognition algorithms into the device for real-time edge computing. Data processing and analysis can be performed locally, reducing dependence on cloud resources and improving the real-time and reliability of the system.

[0004] The technical solution of the present invention is as follows:

[0005] A campus anti-bullying edge computing system based on semantic recognition and image recognition technology, including hardware modules, software modules and cloud-edge collaborative integrated platform. The hardware modules include:

[0006] The semantic recognition module collects voice data on campus through voice collection sensors and transmits it to the edge computing module for real-time semantic recognition and analysis;

[0007] The image recognition module collects image data through the camera and transmits it to the edge computing module for real-time image recognition and analysis;

[0008] The edge computing module recognizes and analyzes the data transmitted by the semantic recognition module and the image recognition module in real time to detect bullying behavior. Once bullying behavior is found, the alarm module is triggered to issue an alarm.

[0009] The alarm module issues an alarm and sends the alarm location to the administrator, reminding campus administrators and relevant personnel to pay attention and take corresponding actions;

[0010] Software modules are used for alarm management, monitoring management, algorithm management, user management, authority management, file management and error correction management;

[0011] The cloud-edge collaborative integrated platform is used for one-click deployment of edge recognition, edge device management, and algorithm upgrades. Through the cloud-based software modules, it controls and manages the semantic recognition module, image recognition module, edge computing module, and alarm module at the edge.

[0012] Furthermore, the cloud-edge collaborative integrated platform includes cloud services and edge services; cloud services include display layer, application layer, service layer and basic support layer; the basic support layer provides the basic support system, edge communication and edge control for cloud services; the service layer provides cloud platform services, edge computing training and algorithm training data; the application layer provides services to the user layer, including alarm management, monitoring management, algorithm management, user management, permission management, file management and error correction management.

[0013] Furthermore, the basic support layer also includes k8s cluster, data service, container image repository, communication hub, and kubeEdge; the basic layer builds a k8s cluster, mirrors the application layer programs and algorithm services, and deploys the application layer service mirror in the k8s cluster; the algorithm service mirror is deployed in the edge anti-bullying device, and combined with the kubeEdge service, the algorithm program can be updated and directly sent to the edge device.

[0014] Furthermore, the cloud service and application layer provide data interfaces to the upper layer; provide communication and control management services to the edge, and transmit data to the presentation layer through the interface for presentation; send the algorithm to the edge device with one click, and when the algorithm is updated, it can be updated to the edge in a timely manner.

[0015] Furthermore, it also includes:

[0016] Algorithm training module, including semantic recognition algorithm and image recognition algorithm. Semantic recognition extracts semantic information related to bullying behavior by analyzing and understanding text or speech;

[0017] Error correction module, which corrects false alarms at the alarm location;

[0018] The power module provides power to subsequent circuits and the main chip.

[0019] The present invention also includes a campus anti-bullying edge computing method based on semantic recognition and image recognition technology, comprising the following steps:

[0020] Anti-bullying equipment is installed on campus. The cloud sends the recognition algorithm to the edge device through kubeEdge. The edge recognition algorithm monitors the data transmitted by the semantic recognition module and the image recognition module in real time. The semantic recognition module and the image recognition module identify campus bullying through the semantic recognition algorithm and the image recognition algorithm respectively.

[0021] When bullying behavior is identified, an alarm signal is sent to the alarm module, which can be sent in various ways, such as sound, phone call, and in-app message.

[0022] After receiving the alarm, the campus security manager will immediately go to handle it to ensure the safety of the students. If a false alarm occurs, it will be corrected and the algorithm will be issued again after being updated.

[0023] Furthermore, the semantic recognition algorithm comprises the following steps:

[0024] Collect text datasets containing campus bullying and normal speech, and preprocess the text datasets to prepare them for algorithm training and testing data;

[0025] In semantic recognition, multiple features are used to represent text, including bag-of-words model, word embedding and syntactic features. The bag-of-words model represents text as a vocabulary set and constructs a feature vector by counting the frequency of each word in the text. Word embedding maps words to a low-dimensional vector space to capture the semantic information of words. Syntactic features extract the syntactic structure information of the text through syntactic analysis algorithms.

[0026] Based on the feature extraction results, the model is trained and evaluated through machine learning algorithms or deep learning models. During the training process, it is necessary to use labeled data sets for supervised learning to optimize the model parameters so that it can accurately identify campus bullying speech;

[0027] Through feature extraction and model prediction, it is determined whether the text contains bullying behavior; if the algorithm determines that the text contains bullying behavior, it can trigger the corresponding early warning mechanism to notify relevant personnel to intervene and handle the situation.

[0028] Furthermore, the text dataset includes students' social media posts, chat records, and records of the school reporting system; preprocessing includes text cleaning, word segmentation, and removal of stop words; machine learning algorithms include naive Bayes, support vector machine, and random forest; deep learning models include convolutional neural networks, recurrent neural networks, and converters.

[0029] Furthermore, the image recognition algorithm comprises the following steps:

[0030] Collect a dataset containing campus bullying images and normal images, and preprocess the images, including image denoising, resizing, and grayscale conversion, to prepare image data for algorithm training and testing;

[0031] In image recognition, a variety of features are used to represent images, including color features, texture features, and shape features;

[0032] Based on the results of feature extraction, the model is trained and evaluated through machine learning algorithms or deep learning models. During the training process, it is necessary to use labeled data sets for supervised learning to optimize the parameters of the model so that it can accurately identify school bullying images;

[0033] The image recognition algorithm inputs an image, extracts features and performs model prediction to determine whether the image contains bullying behavior.

[0034] Furthermore, the data set includes photos taken by students on campus and videos recorded by surveillance cameras; the machine learning algorithms include support vector machines, random forests and K nearest neighbors; the deep learning models include convolutional neural networks or pre-trained deep learning models, and the pre-trained deep learning models include ResNet, VGG and Inception.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] 1. A campus anti-bullying edge computing system and method based on semantic recognition and image recognition technology, which integrates cloud services and edge devices, deploys cloud services through k8s clusters to achieve high availability and load balancing of cloud services; using kubeEdge, edge algorithms can be sent and updated to edge devices with one click;

[0037] 2. A campus anti-bullying edge computing system and method based on semantic recognition and image recognition technology, which sinks the recognition algorithm to the edge device. The edge computing node will not only process data in real time, but will also gradually integrate the capabilities of machine learning and model reasoning to more efficiently and in real time identify campus bullying behaviors and ensure student safety;

[0038] 3. A campus anti-bullying edge computing system and method based on semantic recognition and image recognition technology, using a combination of software and hardware, cloud-edge collaboration, the edge device identifies campus bullying behavior and uses various methods to send an alarm, such as sound, phone calls, system messages, etc., to avoid the failure to deal with bullying behavior in a timely manner;

[0039] 4. A campus anti-bullying edge computing system and method based on semantic recognition and image recognition technology. The cloud service corrects the misidentified campus bullying behavior and trains the algorithm again to improve the accuracy of the algorithm. At the same time, the algorithm can be updated with one click to reduce manual maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic diagram of the architecture of the present invention.

[0041] Figure 2 It is a schematic diagram of the interaction of the hardware modules of the present invention.

[0042] Figure 3 It is a flow chart of the present invention.

[0043] Figure 4 1 is a circuit diagram of the system of the present invention.

[0044] Figure 5 This is a cloud architecture diagram of the present invention. DETAILED DESCRIPTION

[0045] It should be noted that relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0046] The features and performance of the present invention are further described in detail below in conjunction with the embodiments.

[0047] See also Figure 1-5 , a campus anti-bullying edge computing system based on semantic recognition and image recognition technology, including hardware modules, software modules and cloud-edge collaborative integrated platform. The hardware modules include:

[0048] The semantic recognition module collects voice data on campus through voice collection sensors and transmits it to the edge computing module for real-time semantic recognition and analysis;

[0049] The image recognition module collects image data through the camera and transmits it to the edge computing module for real-time image recognition and analysis;

[0050] The edge computing module recognizes and analyzes the data transmitted by the semantic recognition module and the image recognition module in real time to detect bullying behavior. Once bullying behavior is found, the alarm module is triggered to issue an alarm.

[0051] The alarm module issues an alarm and sends the alarm location to the administrator, reminding campus administrators and relevant personnel to pay attention and take corresponding actions;

[0052] Software modules are used for alarm management, monitoring management, algorithm management, user management, authority management, file management and error correction management;

[0053] The cloud-edge collaborative integrated platform is used for one-click deployment of edge recognition, edge device management, and algorithm upgrades. Through the cloud-based software modules, it controls and manages the semantic recognition module, image recognition module, edge computing module, and alarm module at the edge.

[0054] like Figure 1 As shown in Figure 1, the cloud-edge collaborative integrated platform includes cloud services and edge services; cloud services include display layer, application layer, service layer and basic support layer; the basic support layer provides the basic support system, edge communication and edge control of cloud services; the service layer provides cloud platform services, edge computing training and algorithm training data; Figure 5 As shown, the application layer provides services to the user layer, including alarm management, monitoring management, algorithm management, user management, authority management, file management and error correction management.

[0055] The basic support layer also includes k8s cluster, data service, container image repository, communication hub, and kubeEdge; the basic layer builds a k8s cluster, mirrors the application layer programs and algorithm services, and deploys the application layer service mirror in the k8s cluster; the algorithm service mirror is deployed in the edge anti-bullying device, and combined with the kubeEdge service, the algorithm program can be updated and directly sent to the edge device.

[0056] The cloud service and application layer provide data interfaces to the upper layer; provide communication and control management services to the edge, and transmit data to the presentation layer through the interface for presentation; send the algorithm to the edge device with one click, and when the algorithm is updated, it can be updated to the edge in time.

[0057] Also includes:

[0058] Algorithm training module, including semantic recognition algorithm and image recognition algorithm. Semantic recognition extracts semantic information related to bullying behavior by analyzing and understanding text or speech;

[0059] Error correction module, which corrects false alarms at the alarm location;

[0060] The power module supplies power to the subsequent circuits and the main chip. The 220V AC is stepped down to the rated voltage of the transformer. The low-voltage AC is converted into DC through a bridge rectifier capacitor filter circuit; the voltage is stepped down to 5V through a three-stage integrated voltage regulator, and then the power is supplied after filtering through electrolytic capacitors and CBB capacitors.

[0061] The semantic module sensor senses the incoming sound from the surroundings, collects the sound through the sensor, converts the digital signal of the voice into an analog signal and transmits it to the main chip IO port through the circuit connection. In the main control chip, the transmitted analog signal is processed through the program and algorithm, and then converted into a digital signal and transmitted to the alarm module and peripheral circuit.

[0062] The signals collected by the image recognition module camera are converted into analog signals and transmitted to our main control chip. Secondly, the main control chip will unify the collected information and then convert it into digital signals by our internal core circuit and transmit it to the alarm circuit.

[0063] The alarm circuit is mainly connected to the main control chip, and sends the processed signal transmitted back from the front-end network to the alarm circuit. When the signal is detected to be greater than the original set value, our buzzer will emit a beeping sound alarm, and at the same time transmit the alarm data to the cloud service through cloud-edge communication for processing.

[0064] The present invention also includes a campus anti-bullying edge computing method based on semantic recognition and image recognition technology, such as Figure 3 As shown, the following steps are included:

[0065] Anti-bullying equipment is installed on campus. The cloud sends the recognition algorithm to the edge device through kubeEdge. The edge recognition algorithm monitors the data transmitted by the semantic recognition module and the image recognition module in real time. The semantic recognition module and the image recognition module identify campus bullying through the semantic recognition algorithm and the image recognition algorithm respectively.

[0066] When bullying behavior is identified, an alarm signal is sent to the alarm module, which can be sent in various ways, such as sound, phone call, and in-app message.

[0067] After receiving the alarm, the campus security manager will immediately go to handle it to ensure the safety of the students. If a false alarm occurs, it will be corrected and the algorithm will be issued again after being updated.

[0068] The semantic recognition algorithm comprises the following steps:

[0069] Collect text datasets containing campus bullying and normal speech, and preprocess the text datasets to prepare them for algorithm training and testing data;

[0070] In semantic recognition, multiple features are used to represent text, including bag-of-words model, word embedding and syntactic features. The bag-of-words model represents text as a vocabulary set and constructs a feature vector by counting the frequency of each word in the text. Word embedding maps words to a low-dimensional vector space to capture the semantic information of words. Syntactic features extract the syntactic structure information of the text through syntactic analysis algorithms.

[0071] Based on the feature extraction results, the model is trained and evaluated through machine learning algorithms or deep learning models. During the training process, it is necessary to use labeled data sets for supervised learning to optimize the model parameters so that it can accurately identify campus bullying speech;

[0072] Through feature extraction and model prediction, it is determined whether the text contains bullying behavior; if the algorithm determines that the text contains bullying behavior, it can trigger the corresponding early warning mechanism to notify relevant personnel to intervene and handle the situation.

[0073] The text dataset includes students' social media posts, chat records, and records of the school's reporting system; preprocessing includes text cleaning, word segmentation, and removal of stop words; machine learning algorithms include naive Bayes, support vector machine, and random forest; deep learning models include convolutional neural networks, recurrent neural networks, and transformers.

[0074] The image recognition algorithm comprises the following steps:

[0075] Collect a dataset containing campus bullying images and normal images, and preprocess the images, including image denoising, resizing, and grayscale conversion, to prepare image data for algorithm training and testing;

[0076] In image recognition, a variety of features are used to represent images, including color features, texture features, and shape features;

[0077] Based on the results of feature extraction, the model is trained and evaluated through machine learning algorithms or deep learning models. During the training process, it is necessary to use labeled data sets for supervised learning to optimize the parameters of the model so that it can accurately identify school bullying images;

[0078] The image recognition algorithm inputs an image, extracts features and performs model prediction to determine whether the image contains bullying behavior.

[0079] The data set includes photos taken by students on campus and videos recorded by surveillance cameras; the machine learning algorithms include support vector machines, random forests and K nearest neighbors; the deep learning models include convolutional neural networks or pre-trained deep learning models, and the pre-trained deep learning models include ResNet, VGG and Inception.

[0080] The above-mentioned embodiments only express the specific implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the protection scope of the present application. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the technical solution concept of the present application, and these all belong to the protection scope of the present application.

Claims

1. A campus anti-bullying edge computing system based on semantic recognition and image recognition technology, characterized in that: It includes hardware modules, software modules and cloud-edge collaborative integrated platform. The hardware modules include: The semantic recognition module collects voice data on campus through voice collection sensors and transmits it to the edge computing module for real-time semantic recognition and analysis; The image recognition module collects image data through the camera and transmits it to the edge computing module for real-time image recognition and analysis; The edge computing module recognizes and analyzes the data transmitted by the semantic recognition module and the image recognition module in real time to detect bullying behavior. Once bullying behavior is found, the alarm module is triggered to issue an alarm. The alarm module issues an alarm and sends the alarm location to the administrator, reminding campus administrators and relevant personnel to pay attention and take corresponding actions; Software modules are used for alarm management, monitoring management, algorithm management, user management, authority management, file management and error correction management; The cloud-edge collaborative integrated platform is used for one-click deployment of edge recognition, edge device management, and algorithm upgrades. Through the cloud software module, it controls and manages the semantic recognition module, image recognition module, edge computing module, and alarm module at the edge. The cloud-edge collaborative integrated platform includes cloud services and edge services; cloud services include presentation layer, application layer, service layer and basic support layer; the basic support layer also includes k8s cluster, data service, container image repository, communication hub and kubeEdge; The cloud sends the recognition algorithm to the edge computing module through kubeEdge. The algorithm used by the semantic recognition module includes the following steps: Collect text datasets containing campus bullying and normal speech, and preprocess the text datasets to prepare them for algorithm training and testing data; In semantic recognition, multiple features are used to represent text, including bag-of-words model, word embedding and syntactic features. The bag-of-words model represents text as a vocabulary set and constructs a feature vector by counting the frequency of each word in the text. Word embedding maps words to a low-dimensional vector space to capture the semantic information of words. Syntactic features extract the syntactic structure information of the text through syntactic analysis algorithms. Based on the feature extraction results, the model is trained and evaluated through machine learning algorithms or deep learning models. During the training process, it is necessary to use labeled data sets for supervised learning to optimize the model parameters so that it can accurately identify campus bullying speech; Through feature extraction and model prediction, it is determined whether the text contains bullying behavior. If the algorithm determines that the text contains bullying behavior, it can trigger the corresponding early warning mechanism to notify relevant personnel to intervene and deal with it. The text dataset includes students' social media posts, chat records, and records of the school's reporting system; preprocessing includes text cleaning, word segmentation, and removal of stop words; machine learning algorithms include naive Bayes, support vector machine, and random forest; deep learning models include convolutional neural networks, recurrent neural networks, and transformers.

2. According to claim 1, a campus anti-bullying edge computing system based on semantic recognition and image recognition technology is characterized in that: The basic support layer provides the basic support system, edge communication and edge control for cloud services; the service layer provides cloud platform services, edge computing training and algorithm training data; The application layer provides services to the user layer, including alarm management, monitoring management, algorithm management, user management, permission management, file management and error correction management.

3. According to claim 2, a campus anti-bullying edge computing system based on semantic recognition and image recognition technology is characterized in that: The basic support layer builds a k8s cluster, mirrors the application layer programs and algorithm services, and deploys the application layer service mirror in the k8s cluster; The algorithm service image is deployed in the edge anti-bullying device. Combined with the kubeEdge service, the algorithm program can be updated and sent directly to the edge device.

4. According to claim 2, a campus anti-bullying edge computing system based on semantic recognition and image recognition technology is characterized in that: The cloud service and application layer provide data interfaces to the upper layer; Provide communication and control management services to the edge, and pass the data to the presentation layer through the interface for display; The algorithm can be sent to the edge device with one click, and when the algorithm is updated, it can be updated to the edge in time.

5. According to claim 1, a campus anti-bullying edge computing system based on semantic recognition and image recognition technology is characterized in that: Also includes: Algorithm training module, including semantic recognition algorithm and image recognition algorithm. Semantic recognition extracts semantic information related to bullying behavior by analyzing and understanding text or speech; Error correction module, which corrects false alarms at the alarm location; The power module provides power to subsequent circuits and the main chip.

6. A campus anti-bullying edge computing method based on semantic recognition and image recognition technology, characterized in that: The following steps are involved: Anti-bullying equipment is installed on campus. The cloud sends the recognition algorithm to the edge device through kubeEdge. The edge recognition algorithm monitors the data transmitted by the semantic recognition module and the image recognition module in real time. The semantic recognition module and the image recognition module identify campus bullying through the semantic recognition algorithm and the image recognition algorithm respectively. When bullying behavior is identified, an alarm signal is sent to the alarm module, which can be sent in various ways, such as sound, phone call, and in-app message. After receiving the alarm, the campus security manager will immediately go to handle it to ensure the safety of the students. If there is a false alarm, the error will be corrected and the algorithm will be issued again after updating the algorithm. The semantic recognition algorithm comprises the following steps: Collect text datasets containing campus bullying and normal speech, and preprocess the text datasets to prepare them for algorithm training and testing data; In semantic recognition, multiple features are used to represent text, including bag-of-words model, word embedding and syntactic features. The bag-of-words model represents text as a vocabulary set and constructs a feature vector by counting the frequency of each word in the text. Word embedding maps words to a low-dimensional vector space to capture the semantic information of words. Syntactic features extract the syntactic structure information of the text through syntactic analysis algorithms. Based on the feature extraction results, the model is trained and evaluated through machine learning algorithms or deep learning models. During the training process, it is necessary to use labeled data sets for supervised learning to optimize the model parameters so that it can accurately identify campus bullying speech; Through feature extraction and model prediction, it is determined whether the text contains bullying behavior. If the algorithm determines that the text contains bullying behavior, it can trigger the corresponding early warning mechanism to notify relevant personnel to intervene and deal with it. The text dataset includes students' social media posts, chat records, and records of the school's reporting system; preprocessing includes text cleaning, word segmentation, and removal of stop words; machine learning algorithms include naive Bayes, support vector machine, and random forest; deep learning models include convolutional neural networks, recurrent neural networks, and transformers.

7. According to claim 6, a campus anti-bullying edge computing method based on semantic recognition and image recognition technology is characterized in that: The image recognition algorithm comprises the following steps: Collect a dataset containing campus bullying images and normal images, and preprocess the images, including image denoising, resizing, and grayscale conversion, to prepare image data for algorithm training and testing; In image recognition, a variety of features are used to represent images, including color features, texture features, and shape features; Based on the results of feature extraction, the model is trained and evaluated through machine learning algorithms or deep learning models. During the training process, it is necessary to use labeled data sets for supervised learning to optimize the parameters of the model so that it can accurately identify school bullying images; The image recognition algorithm inputs an image, extracts features and predicts the model to determine whether the image contains bullying behavior.

8. According to claim 7, a campus anti-bullying edge computing method based on semantic recognition and image recognition technology is characterized in that: The data set includes photos taken by students on campus and videos recorded by surveillance cameras; the machine learning algorithms include support vector machines, random forests and K nearest neighbors; the deep learning models include convolutional neural networks or pre-trained deep learning models, and the pre-trained deep learning models include ResNet, VGG and Inception.

Images