Helmet wearing detection method and system and medium
The intelligent helmet recognition system can detect in real time whether electric bike riders are wearing helmets, solving the problem of electric bike riders not wearing helmets, improving detection efficiency, reducing management costs, and providing accurate violation analysis.
Patent Information
- Application Number
- CN202510650055.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-20
- Publication Date
- 2025-10-17
AI Technical Summary
The problem of riders not wearing safety helmets is particularly prominent when using electric vehicles. Existing technologies make it difficult to detect and manage them efficiently, resulting in safety hazards and high management costs.
Using intelligent helmet recognition AI cameras and flat-panel helmet recognition terminals, combined with image recognition algorithms and machine learning models, it can detect in real time whether electric bike riders are wearing helmets. If not, it will trigger a voice alarm and close the gate. Deep learning and edge detection technology are used to improve detection accuracy.
It achieves accurate detection of electric bike riders' helmet wearing, reduces management costs, improves detection efficiency, and provides objective statistical analysis of violations.
Smart Images
Figure CN120808284A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of computer vision, in particular to a helmet wearing detection method, system and medium. BACKGROUND
[0002] With the development of society, electric vehicles as a convenient means of transportation are accepted and used by more and more people. However, these electric vehicles bring convenience, but also bring certain safety problems, and the problem of not wearing safety helmets by riders is particularly prominent. SUMMARY
[0003] The main purpose of the present application is to provide a helmet wearing detection method, system and medium, which aims to improve the helmet wearing detection efficiency and reduce the management cost.
[0004] In order to achieve the above purpose, the present application provides a helmet wearing detection method, which is applied to a helmet wearing detection system, the system includes an intelligent helmet recognition AI camera arranged in an electric vehicle riding area, and the method includes the following steps:
[0005] Step S10, acquiring target image information of the target area detected by the intelligent helmet recognition AI camera in real time, the target image information including target electric vehicle information of the target area and image information of the electric vehicle riding user, the target electric vehicle information being electric vehicles with sunshades and electric vehicles without sunshades;
[0006] If the target electric vehicle is an electric vehicle without a sunshade, directly execute step S20 to analyze the image information of the electric vehicle riding user to obtain helmet wearing information of the electric vehicle riding user, the helmet wearing information including wearing a helmet or not wearing a helmet;
[0007] Step S30, if the analysis result is that the electric vehicle riding user does not wear a helmet, closing the barrier and triggering voice alarm prompt and safety helmet wearing voice propaganda.
[0008] The further technical scheme of the present application is that the system further includes a flat helmet recognition terminal arranged at the barrier, and the step of acquiring the target image information of the target area detected by the intelligent helmet recognition AI camera in real time further includes:
[0009] If the target electric vehicle is an electric vehicle with a sunshade, the step S20 includes:
[0010] Step S101, closing the barrier and reacquiring the image information of the electric vehicle riding user through the flat helmet recognition terminal.
[0011] The further technical solution of the present application is that the step S20 of analyzing the image information of the electric vehicle riding user to obtain the helmet wearing information of the electric vehicle riding user comprises:
[0012] Step S201, analyze the color distribution in the image information of the electric vehicle riding user, and extract the pixel area corresponding to the helmet color;
[0013] Step S202, extract the edge information of the pixel area corresponding to the helmet color by an edge detection algorithm, and extract the corresponding shape feature by a Hough transform method;
[0014] Step S203, match the color distribution in the image information of the electric vehicle riding user, the pixel area corresponding to the helmet color, and the corresponding shape feature with a pre-defined helmet template or feature library, and calculate the similarity between the features of the object to be identified and the template features;
[0015] Step S204, if the similarity exceeds a certain threshold, it is considered that the helmet is detected.
[0016] The further technical solution of the present application is that the step S20 of analyzing the image information of the electric vehicle riding user to obtain the helmet wearing information of the electric vehicle riding user comprises:
[0017] Step S205, extract the texture features in the image information of the electric vehicle riding user, input the extracted texture features into a machine learning model, and judge whether there is helmet information in the image information of the electric vehicle riding user.
[0018] The further technical solution of the present application is that the machine learning model uses a support vector machine, a decision tree or a random forest machine learning algorithm for classification, takes the extracted features as input, classifies the helmet samples and non-helmet samples by finding an optimal hyperplane, and determines the parameters of the hyperplane by minimizing the structural risk in the training process.
[0019] The further technical solution of the present application is that the method further comprises model training and evaluation: dividing the labeled helmet image and non-helmet image data set into a training set, a validation set and a test set;
[0020] The training set is used to train the model, and the parameters of the model are adjusted so that the model can accurately classify the helmet and the non-helmet on the training set;
[0021] The validation set is used to select the optimal model parameters and adjust the model structure to prevent overfitting;
[0022] Finally, the performance of the model is evaluated using the test set, and evaluation indexes including accuracy and recall rate are calculated.
[0023] The further technical solution of the present application is that the step S20 of analyzing the image information of the electric vehicle riding user to obtain the helmet wearing information of the electric vehicle riding user comprises:
[0024] In step S206, deep features in the image information of the electric vehicle riding user are extracted based on a deep learning algorithm and a convolutional neural network architecture, wherein the convolutional neural network architecture uses VGGNet and ResNet as backbone networks.
[0025] In step S207, based on the deep features, a detection head is added to the backbone network for helmet detection and positioning. The further technical solution of the present application is that the method further comprises training the convolutional neural network model, wherein in order to enhance the diversity of data and prevent model overfitting, a data enhancement operation is performed on the training data, the data enhancement operation includes random cropping, flipping, rotation, scaling, and color jittering, and in the process of training the convolutional neural network model, a classification loss function and a regression loss function are used.
[0027] To achieve the above-mentioned purpose, the present application further provides a helmet wearing detection system, which comprises a memory, a processor, and a helmet wearing detection program stored in the processor, and the helmet wearing detection program performs the steps of the above-mentioned method when executed by the processor.
[0028] To achieve the above-mentioned purpose, the present application further provides a computer readable storage medium, which stores a helmet wearing detection program, and the helmet wearing detection program performs the steps of the above-mentioned method when executed by the processor.
[0029] The helmet wearing detection method, system and medium of the present application adopt the above-mentioned technical solution to obtain target image information of a target region detected by an intelligent helmet recognition AI camera in real time, the target image information comprises target electric vehicle information of the target region and image information of an electric vehicle riding user, the target electric vehicle information is electric vehicles with and without sunshades, if the target electric vehicle is an electric vehicle without a sunshade, the image information of the electric vehicle riding user is analyzed to obtain helmet wearing information of the electric vehicle riding user, the helmet wearing information comprises wearing a helmet or not wearing a helmet, if the analysis result is that the electric vehicle riding user does not wear a helmet, a barrier gate is closed, a voice alarm prompt and a safety helmet wearing voice propaganda are triggered, the helmet wearing detection efficiency is improved, and the management cost is reduced. BRIEF DESCRIPTION OF DRAWINGS
[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only show some of the embodiments of the present application, and other drawings can be obtained by those of ordinary skill in the art without any creative effort based on the drawings shown.
[0031] Figure 1 is a flowchart of a preferred embodiment of the helmet wearing detection method of the present application.
[0032] The implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION
[0033] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without any creative effort belong to the scope of protection of the present application.
[0034] The present application provides a helmet wearing detection method, which mainly detects whether an electric vehicle rider wears a helmet through AI. Advanced image recognition algorithms and helmet detection technology are adopted to monitor and identify whether the rider wears a helmet in real time through an electric vehicle helmet AI recognition linkage barrier gate platform system, intelligent cameras, flat helmet identification terminals and helmet identification barriers. Once the situation of not wearing a helmet is detected, voice warning and safety helmet wearing voice propaganda are triggered immediately, the barrier gate is automatically closed, and the rider can pass through after wearing a helmet. The two-wheeled electric vehicle (not including bicycles), head and helmet target pictures are accurately identified, and the highest accuracy of industry helmet recognition rate is achieved.
[0035] Please refer to Figure 1 The helmet wearing detection method provided by the present application is applied to a helmet wearing detection system, which comprises an intelligent helmet recognition AI camera arranged in an electric vehicle riding area. The preferred embodiment of the helmet wearing detection method provided by the present application comprises the following steps:
[0036] Step S10: acquiring target image information of a target area detected by the intelligent helmet recognition AI camera in real time, wherein the target image information comprises target electric vehicle information of the target area and image information of an electric vehicle riding user, and the target electric vehicle information is for electric vehicles with sunshades and electric vehicles without sunshades.
[0037] If the target electric vehicle is an electric vehicle without a sunshade, directly execute step S20 to analyze the image information of the electric vehicle riding user to obtain the helmet wearing information of the electric vehicle riding user, the helmet wearing information including wearing a helmet or not wearing a helmet.
[0038] Step S30, if the analysis result is that the electric vehicle riding user does not wear a helmet, close the barrier gate and trigger voice alarm prompt and safety helmet wearing voice propaganda.
[0039] In this embodiment, the system further includes a flat helmet recognition terminal arranged at the barrier gate, and after the step of acquiring the target image information of the target region detected by the intelligent helmet recognition AI camera, the system further includes:
[0040] If the target electric vehicle is an electric vehicle with a sunshade, the step S20 includes:
[0041] Step S101, close the barrier gate and reacquire the image information of the electric vehicle riding user through the flat helmet recognition terminal.
[0042] Specifically, as an implementation, in this embodiment, the step S20 of analyzing the image information of the electric vehicle riding user to obtain the helmet wearing information of the electric vehicle riding user includes:
[0043] Step S201, analyze the color distribution in the image information of the electric vehicle riding user to extract the pixel region corresponding to the helmet color;
[0044] Step S202, extract the edge information of the pixel region corresponding to the helmet color through an edge detection algorithm, and extract the corresponding shape feature by using a Hough transform method;
[0045] Step S203, match the color distribution in the image information of the electric vehicle riding user, the pixel region corresponding to the helmet color, and the corresponding shape feature with a pre-defined helmet template or feature library, and calculate the similarity between the features of the object to be recognized and the template features;
[0046] Step S204, if the similarity exceeds a certain threshold, it is considered that a helmet is detected.
[0047] The helmet color is usually bright and unique, and has color features such as common red, yellow, blue, etc. The color distribution of the pixels in the image can be analyzed in the RGB or HSV color space to extract the pixel region corresponding to the helmet color. For example, in the HSV space, a specific hue, saturation and lightness range is set to filter the pixels that may belong to the helmet.
[0048] In addition, the helmet generally has an elliptical or hemispherical shape and a certain size ratio, and has its unique shape characteristics. The edge information in the image can be extracted by an edge detection algorithm such as the Canny algorithm, and then an ellipse or a circle shape can be detected by using a Hough transform method or the like, and whether it is a helmet can be determined according to the size and ratio of the shape.
[0049] After the color features and shape features are extracted, classification and recognition can be performed in this embodiment: according to the extracted color, shape and other features, a helmet template or feature library defined in advance is matched. The similarity between the features of the object to be recognized and the template features is calculated, such as using the Euclidean distance, cosine similarity and the like. If the similarity exceeds a certain threshold, it is considered that a helmet is detected.
[0050] As another embodiment, in this embodiment, the step S20 can use a machine learning algorithm to determine whether there is helmet information in the image information of the electric vehicle riding user.
[0051] Specifically, in this embodiment, the step of analyzing the image information of the electric vehicle riding user to obtain the helmet wearing information of the electric vehicle riding user includes:
[0052] In step S205, texture features in the image information of the electric vehicle riding user are extracted, and the extracted texture features are input into a machine learning model to determine whether there is helmet information in the image information of the electric vehicle riding user.
[0053] In addition to color and shape features, texture features can also be extracted from the helmet, such as using a gray level co-occurrence matrix to calculate texture parameters such as energy, contrast and correlation. In addition, geometric features such as the area, perimeter and aspect ratio of the circumscribed rectangle of the object can also be extracted.
[0054] In this embodiment, the machine learning model uses machine learning algorithms such as support vector machines (SVM), decision trees and random forests for classification. Taking SVM as an example, the extracted features are input, and an optimal hyperplane is found to classify the helmet samples and non-helmet samples. In the training process, the parameters of the hyperplane are determined by minimizing the structural risk.
[0055] In this embodiment, the method further includes model training and evaluation: the labeled helmet image and non-helmet image data set is divided into a training set, a validation set and a test set;
[0056] The training set is used to train the model, and the parameters of the model are adjusted so that the model can accurately classify the helmet and non-helmet on the training set;
[0057] The validation set is used to select the optimal model parameters and adjust the model structure to prevent overfitting;
[0058] Finally, the performance of the model is evaluated using the test set, and evaluation metrics including accuracy and recall are calculated.
[0059] As an embodiment, in this embodiment, the algorithm based on deep learning can also be used to determine whether there is helmet information in the image information of the electric vehicle riding user.
[0060] Specifically, in this embodiment, the step S20 of analyzing the image information of the electric vehicle riding user to obtain the helmet wearing information of the electric vehicle riding user includes:
[0061] Step S206, based on the deep learning algorithm, a convolutional neural network architecture is used to extract deep features in the image information of the electric vehicle riding user, wherein the convolutional neural network architecture uses VGGNet and ResNet as backbone network.
[0062] Step S207, based on the deep features, a detection head is added to the backbone network for detection and positioning of the helmet.
[0063] In this embodiment, the convolutional neural network (CNN) architecture uses VGGNet, ResNet, etc. as backbone network, which contains multiple convolutional layers and pooling layers for automatically extracting deep features of images. The convolutional layer slides on the image through the convolution kernel for convolution operation to extract local features of the image, such as, where x i is the input feature, k i is the convolution kernel, b is the bias, and f is the activation function. The pooling layer then down-samples the output of the convolution layer to reduce the data dimension, such as max pooling or average pooling.
[0064] On the basis of the backbone network, a detection head is added for detection and positioning of the helmet. As in Faster R-CNN, the region proposal network (RPN) generates candidate regions that may contain the helmet, and then the classifier and regressor classify and regress the boundary box of the candidate region to determine the class and position of the helmet.
[0065] In this embodiment, the method further includes training the convolutional neural network model, wherein in order to enhance the diversity of data and prevent model overfitting, data enhancement operations are performed on the training data, including random cropping, flipping, rotation, scaling, color jittering, etc. During the training of the convolutional neural network model, classification loss function and regression loss function are used.
[0066] Commonly used loss functions include classification loss and regression loss. For classification loss, such as cross-entropy loss function where yi is a real label, is a predicted probability. For regression loss, such as a smooth L1 loss function, is used to calculate the error between the predicted bounding box and the real bounding box. By back propagation algorithm, the parameters of the model are updated to minimize the loss function.
[0067] The helmet wearing detection method can be used for electric vehicle helmet AI monitoring and linkage barrier management. The smart helmet recognition AI camera, tablet helmet recognition terminal and helmet recognition barrier can monitor and analyze the electric vehicle riders in real time, and immediately give voice warning and barrier closing when the helmet is not worn.
[0068] In addition, the helmet wearing detection method can monitor the helmet wearing situation of electric vehicle personnel in the area in real time, facilitate event statistics and business management, provide comprehensive illegal behavior statistics analysis and objective and quantitative data business evaluation support, and rank and statistically report the helmetless rate of each department and equipment.
[0069] In summary, the helmet wearing detection method adopts the above technical solutions, obtains target image information of the target area detected by the smart helmet recognition AI camera, the target image information includes target electric vehicle information and image information of the electric vehicle riding user in the target area, and the target electric vehicle information is electric vehicles with and without sunshades; if the target electric vehicle is an electric vehicle without a sunshade, the image information of the electric vehicle riding user is analyzed to obtain helmet wearing information of the electric vehicle riding user, the helmet wearing information includes wearing a helmet or not wearing a helmet; if the analysis result is that the electric vehicle riding user does not wear a helmet, the barrier is closed, and voice warning and safety helmet wearing voice propaganda are triggered, which improves the helmet wearing detection efficiency and reduces the management cost.
[0070] To achieve the above purpose, the present application also provides a helmet wearing detection system, which comprises a memory, a processor and a helmet wearing detection program stored in the processor. The helmet wearing detection program is run by the processor to execute the steps of the method as described above. Here, the steps are described.
[0071] To achieve the above purpose, the present application also provides a computer readable storage medium, which stores a helmet wearing detection program. The helmet wearing detection program is run by the processor to execute the steps of the method as described above. Here, the steps are not described.
[0072] The above merely describes the preferred embodiments of the present application, and is not intended to limit the patent scope of the present application. Any equivalent structural changes made under the concept of the present application, or direct / indirect application in other related technical fields, using the content of the present application specification and drawings, are included in the patent protection scope of the present application.
Claims
1. A helmet wearing detection method, characterized in that: The method is applied to a helmet wearing detection system, which includes an intelligent helmet recognition AI camera set in an electric vehicle riding area. The method includes the following steps: Step S10, obtaining target image information of a target area detected in real time by the smart helmet recognition AI camera, wherein the target image information includes target electric vehicle information in the target area and image information of an electric vehicle rider, wherein the target electric vehicle information includes electric vehicles with awnings and electric vehicles without awnings; If the target electric vehicle is an electric vehicle without a sunshade, directly executing step S20, parsing the image information of the electric vehicle rider to obtain helmet wearing information of the electric vehicle rider, the helmet wearing information including whether the rider is wearing a helmet or not wearing a helmet; Step S30: If the analysis result shows that the electric vehicle rider is not wearing a helmet, the gate is closed, and a voice alarm prompt and a voice promotion for wearing a helmet safely are triggered.
2. The helmet wearing detection method according to claim 1, characterized in that: The system further includes a flat-panel helmet recognition terminal provided at the gate, and after the step of obtaining target image information of the target area detected in real time by the smart helmet recognition AI camera, the system further includes: If the target electric vehicle is an electric vehicle with a sunshade, the following steps are performed before step S20: Step S101: close the barrier and re-collect image information of the electric vehicle rider through the flat-panel helmet recognition terminal.
3. The helmet wearing detection method according to claim 1, characterized in that: The step S20 of parsing the image information of the electric vehicle rider to obtain the helmet wearing information of the electric vehicle rider includes: Step S201, analyzing the color distribution in the image information of the electric vehicle rider, and extracting the pixel area corresponding to the helmet color; Step S202, extracting edge information of the pixel area corresponding to the helmet color using an edge detection algorithm, and extracting corresponding shape features using a Hough transform method; Step S203, matching the color distribution in the image information of the electric bike rider, the pixel area corresponding to the helmet color, and the corresponding shape features with a predefined helmet template or feature library, and calculating the similarity between the features of the object to be identified and the template features; Step S204: If the similarity exceeds a certain threshold, it is considered that a helmet is detected.
4. The helmet wearing detection method according to claim 1, characterized in that: The step S20 of parsing the image information of the electric vehicle rider to obtain the helmet wearing information of the electric vehicle rider includes: Step S205: extracting texture features from the image information of the electric vehicle riding user, inputting the extracted texture features into a machine learning model, and determining whether helmet information exists in the image information of the electric vehicle riding user.
5. The helmet wearing detection method according to claim 4, characterized in that: The machine learning model uses a support vector machine, a decision tree, or a random forest machine learning algorithm for classification, takes the extracted features as input, and classifies helmet samples and non-helmet samples by finding an optimal hyperplane. During the training process, the parameters of the hyperplane are determined by minimizing the structural risk.
6. The helmet wearing detection method according to claim 4 or 5, characterized in that: The method further includes model training and evaluation: dividing the labeled helmet image and non-helmet image dataset into a training set, a validation set, and a test set; Use the training set to train the model and adjust the model parameters so that the model can accurately classify helmets and non-helmets on the training set; Use the validation set to select the optimal model parameters and adjust the model structure to prevent overfitting; Finally, the performance of the model is evaluated using the test set, and evaluation indicators are calculated, including precision and recall.
7. The helmet wearing detection method according to claim 1, characterized in that: The step S20 of parsing the image information of the electric vehicle rider to obtain the helmet wearing information of the electric vehicle rider includes: Step S206: extracting deep features from the image information of the electric vehicle rider using a convolutional neural network architecture based on a deep learning algorithm, wherein the convolutional neural network architecture uses VGGNet and ResNet as backbone networks; Step S207: Based on the deep-level features, a detection head is added to the backbone network for helmet detection and positioning.
8. The helmet wearing detection method according to claim 7, characterized in that: The method also includes: training the convolutional neural network model, wherein, in order to enhance data diversity and prevent model overfitting, data enhancement operations are performed on the training data, and the data enhancement operations include random cropping, flipping, rotation, scaling, and color jittering. During the training of the convolutional neural network model, a classification loss function and a regression loss function are used.
9. A helmet wearing detection system, characterized in that: The helmet detection system includes a memory, a processor, and a helmet wearing detection program stored in the processor. When the helmet wearing detection program is executed by the processor, the steps of the method according to any one of claims 1 to 8 are executed.
10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a helmet wearing detection program, and when the helmet wearing detection program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are executed.