A method, apparatus, system and medium for human helmet detection

Abstract

Embodiments of the present application provide a kind of helmet detection method, device, storage medium and system, its method includes: obtaining the sample image set of target object;Sample image set is carried out attribute identification, to obtain first label information and second label information;Based on first label, first label confidence, second label and second label confidence, initial neural network model is carried out model training operation, to obtain target neural network model;Through the target neural network model, label confidence analysis is carried out to target image information, to obtain the label confidence information of target image;In the case where the first confidence is greater than the first threshold value, the first label is contained in the label confidence information, and posture adjustment instruction is sent to target object;In the case where the second confidence is greater than the first threshold value, the second label is contained in the label confidence information, and alarm operation is executed.The present application solves the problem of low helmet recognition accuracy, thereby improving the accuracy of helmet recognition.

Description

Technical Field

[0001] This invention relates to the field of image detection technology, and in particular to a method, apparatus, system and medium for detecting human helmets. Background Technology

[0002] To regulate the riding of shared electric bikes, a special campaign has been launched to crack down on traffic violations involving shared electric bikes, targeting riders and passengers not wearing helmets in key urban areas. Therefore, wearing a helmet correctly is now a required step before unlocking a shared electric bike.

[0003] In existing technologies, the methods for detecting whether a helmet is worn correctly are relatively crude. Generally, as long as a helmet is present within the camera's detection range, it can be determined that the user has worn the helmet correctly. This detection method is difficult to avoid situations such as the helmet not being worn properly or not being worn on the user's head. In actual vehicle use, these irregular situations are often misdetected as correct wearing, leading to improper helmet use.

[0004] Therefore, it is necessary to propose a helmet detection method, device, system, and medium to solve or at least alleviate the above-mentioned defects. Summary of the Invention

[0005] The main purpose of this application is to provide a helmet detection method, device, system, and medium to solve the problem that existing technologies cannot prevent helmets from being used improperly, such as when they are not properly worn or are not worn on the user's head, from passing the helmet detection process, thus resulting in non-standard helmet use in actual use.

[0006] According to an embodiment of the present invention, a method, apparatus, system, and medium for detecting helmets are provided, applicable to shared equipment including helmet devices, comprising:

[0007] Obtain a sample image set of the target object;

[0008] The sample image set is subjected to attribute recognition to obtain first label information and second label information. The first label information includes a first label that the target object contains helmet equipment information and a first confidence level of the first label. The second label information includes a second label that the target object does not contain helmet equipment information and a second confidence level of the second label.

[0009] The first label, the confidence score of the first label, the second label, and the confidence score of the second label are input into the initial neural network model to perform model training operations on the initial neural network model and obtain the target neural network model.

[0010] The target image information is analyzed by the target neural network model to obtain the label confidence information of the target image, wherein the label confidence information includes at least one of the first label, the first label confidence, the second label, and the second label confidence.

[0011] If the tag confidence information contains the first tag and the first confidence is greater than the first threshold, the helmet of the target object is confirmed to be in the first state, and a posture adjustment command is sent to the target object.

[0012] If the tag confidence information includes the second tag and the second confidence is greater than the first threshold, the helmet of the target object is confirmed to be in the second state, and an alarm operation is performed.

[0013] In an exemplary embodiment, the step of performing attribute recognition on the sample image set to obtain first label information and second label information includes:

[0014] The sample image set is segmented using a preset segmentation model to obtain initial helmet information;

[0015] The initial helmet information is labeled with attributes by using a preset initial neural network model to obtain the first label information and the second label information.

[0016] In an exemplary embodiment, when the label confidence information includes the first label and the first confidence level is greater than a first threshold, the method further includes:

[0017] The system acquires the initial position information and target position information of the helmet device, wherein the initial position information includes the initial coordinate information and initial height information of the helmet device within a target time period, and the target position information includes the final coordinate information and final height information of the helmet device within the target time period; based on the initial coordinate information and final coordinate information, the system determines the coordinate trajectory information of the helmet device; based on the initial height information and final height information, the system determines the height trajectory information of the helmet device.

[0018] If the coordinate trajectory information satisfies the second condition and / or the altitude trajectory information satisfies the third condition, the device start-up information is fed back to the external control center.

[0019] In an exemplary embodiment, determining the height trajectory information of the helmet device based on the initial height information and the final height information includes:

[0020] The detection reference plane is determined based on the initial height information;

[0021] Based on the initial height information and the final height information, the trajectory plane of the helmet device is determined;

[0022] Based on the trajectory plane and the detection reference plane, the included angle of the motion trajectory is determined, and the included angle of the motion trajectory is used as the height trajectory information.

[0023] In an exemplary embodiment, determining the height trajectory information of the helmet device based on the initial height information and the final height information includes:

[0024] Based on the initial height information and the final height information, the height difference of the helmet device within the target time is determined, and the height difference is used as the height movement trajectory.

[0025] In one exemplary embodiment, before performing the trajectory determination operation when the helmet attribute information satisfies the first condition, the method further includes:

[0026] If the first timing meets the preset period, a tracking command is sent to the helmet device to instruct the helmet device to provide feedback information;

[0027] Upon receiving a response from the helmet device, a communication link is established with the helmet device, wherein the communication link is used to perform the trajectory determination operation.

[0028] According to another embodiment of the present invention, a helmet detection device is provided, applied to a shared device including a helmet device, comprising:

[0029] The initial image acquisition module is used to acquire image information of the target object;

[0030] An attribute recognition module is used to perform attribute recognition on the sample image set to obtain first label information and second label information, wherein the first label information includes a first label that the target object contains helmet device information and a first confidence level of the first label, and the second label information includes a second label that the target object does not contain helmet device information and a second confidence level of the second label;

[0031] The model training module is used to perform model training operations on the initial neural network model based on the first label, the confidence level of the first label, the second label, and the confidence level of the second label, so as to obtain the target neural network model.

[0032] The information analysis module is used to perform label confidence analysis on the target image information through the target neural network model to obtain the label confidence information of the target image;

[0033] The instruction sending module is used to send an attitude adjustment instruction to the target object when the tag confidence information contains the first tag and the first confidence is greater than a first threshold.

[0034] An alarm module is used to perform an alarm operation when the tag confidence information contains the second tag and the second confidence is greater than a first threshold.

[0035] In one exemplary embodiment, the attribute recognition module includes:

[0036] The image segmentation unit is used to segment the sample image set using a preset segmentation model to obtain initial helmet information;

[0037] The attribute recognition unit is used to perform tag attribute recognition on the initial helmet information through a preset initial neural network model to obtain the first tag information and the second tag information.

[0038] The present invention also provides a helmet detection system, the system including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the helmet detection accuracy evaluation method described above.

[0039] The present invention also provides a storage medium storing a computer program, which, when executed by a processor, implements the aforementioned helmet detection method.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] This invention provides a method, device, system, and medium for detecting helmets. It trains an initial neural network model using a sample image set and outputs label information for image judgment using the trained target neural network model, thereby determining whether the target object is wearing a helmet correctly. This solves the problem of inaccurate helmet detection and improves the accuracy of helmet detection. Attached Figure Description

[0042] Figure 1 This is a flowchart of a helmet detection method according to an embodiment of the present invention;

[0043] Figure 2 A structural block diagram of a helmet detection device according to an embodiment of the present invention. Detailed Implementation

[0044] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0046] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0047] Furthermore, the use of terms such as "first" and "second" in this invention is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. Additionally, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. When the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0048] Firstly, please refer to the appendix. Figure 1 The present invention provides a helmet detection method in one embodiment, which is applied to a shared device including a helmet device, wherein the shared device may be (but is not limited to) a shared bicycle, a shared electric bicycle, etc., and the method includes:

[0049] Step S102: Obtain a sample image set of the target object;

[0050] In this embodiment, the acquired sample image set is sample image data used for training the neural network model. It can be (but is not limited to) a random combination of images containing a helmet and a face. For example, the face is a side profile and the helmet is an image occluded by a mask, or the face is a frontal view and the helmet is occluded by a mask, or the face is a side profile and the helmet is an image rotated at a certain angle, etc. The face can be randomly selected as male or female, and the style and color of the helmet can also be diverse, such as tank helmets, pilot helmets, construction helmets, vehicle helmets, etc. Training the model with various types of images is to improve the training accuracy of the model.

[0051] It should be noted that the sample image set can be obtained from the Internet through web crawlers, or it can be obtained by manually splicing images randomly, or by automatically cutting and splicing a certain number of images using an automatic masking model or splicing model. In order to ensure the training effect, the sample image set should contain a massive amount of image data.

[0052] Step S104: Perform attribute recognition on the sample image set to obtain first label information and second label information, wherein the first label information includes a first label that the target object contains helmet device information and a first confidence level of the first label, and the second label information includes a second label that the target object does not contain helmet device information and a second confidence level of the second label;

[0053] In this embodiment, obtaining the first label and the second label, along with their corresponding confidence scores, is to filter the model training results and select pseudo-labels that can play a significant role in the model training process, thereby improving the accuracy of model training.

[0054] The first label indicates that a helmet is present in an initial image, and the second label indicates that a helmet is not present in the initial image. Correspondingly, the first confidence level indicates the sample probability of the first label (e.g., 0.6), and the second confidence level indicates the sample probability of the second label (e.g., 0.3). The first and second labels can be manually labeled or automatically labeled using software. The first and second labels can be (but are not limited to) bounding boxes, 3D boxes, or labeled lines and dots. Regardless of whether they are bounding boxes, lines, or dots, they can move with the target object or be labeled after the target object appears. The target object includes (but is not limited to) users of the shared device, specifically including (but not limited to) men and women.

[0055] It should be noted that the model can be trained using either supervised or unsupervised methods, and the specific method can be selected based on the usage environment.

[0056] Step S106: Input the first label, the confidence level of the first label, the second label, and the confidence level of the second label into the initial neural network model to perform model training operation on the initial neural network model and obtain the target neural network model;

[0057] In this embodiment, once the target neural network model is obtained, it can be used to predict and identify real-world scenarios without further training, which greatly reduces the workload of manual identification and significantly improves equipment management efficiency.

[0058] It should be noted that both the initial neural network model and the target neural network model after training can be models such as YOLO3 and YOLO5, and those skilled in the art can choose according to actual needs.

[0059] Step S108: Using the target neural network model, perform label confidence analysis on the target image information to obtain label confidence information of the target image, wherein the label confidence information includes at least one of the first label, the first label confidence, the second label, and the second label confidence.

[0060] Step S1010: If the tag confidence information contains the first tag and the first confidence is greater than the first threshold, confirm that the helmet of the target object is in the first state and send a posture adjustment command to the target object.

[0061] In this embodiment, when a first label is included and the first confidence level is greater than the first threshold, it indicates that the target object in the target image is wearing a helmet, and the confidence level of the recognition result meets the threshold condition. However, the wearing posture may be inaccurate, so it is necessary to remind the user to adjust the wearing posture. The posture adjustment instruction can be displayed in the form of graphics and text, or in the form of voice (such as continuous voice broadcast), or in other ways.

[0062] It should be noted that after obtaining the target neural network model, performing helmet and face recognition on the target image and labeling it accordingly enables the control module to respond quickly, reducing the computing power required for the calculation process.

[0063] Step S1012: If the tag confidence information includes the second tag and the second confidence is greater than the first threshold, confirm that the helmet of the target object is in the second state and perform an alarm operation.

[0064] In this embodiment, if the second tag is included and the confidence level is greater than the first threshold, it indicates that the target object is not wearing a helmet. At this time, an alarm needs to be triggered. The alarm can be triggered by means of a buzzer alarm, or by sending a command to the control module to make the shared device unusable, or by continuously flashing lights to provide an alarm prompt.

[0065] By labeling and classifying images, and then using this information to train a model, and finally using the trained target neural network model for image recognition, errors in human recognition can be reduced. At the same time, it can accurately identify whether users are wearing helmets or whether their wearing posture meets the requirements, thereby improving the efficiency and accuracy of shared equipment management and reducing the cost of shared equipment management.

[0066] The first state includes (but is not limited to) the target being wearing a helmet, but in an incorrect posture; the second state includes (but is not limited to) the target being not wearing a helmet.

[0067] In an exemplary embodiment, the step of performing attribute recognition on the sample image set to obtain first label information and second label information includes:

[0068] Step S1042: The sample image set is segmented using a preset segmentation model to obtain initial helmet information;

[0069] Step S1044: The initial helmet information is subjected to tag attribute recognition through a preset initial neural network model to obtain the first tag information and the second tag information.

[0070] In this embodiment, segmenting the image and then recognizing each segment is to reduce interference when recognizing a specific target, thereby improving the recognition accuracy of the specific target.

[0071] In this process, both the segmentation model and the label attribute recognition can be achieved by combining models such as YOLO3 and YOLO5 with other models; the label attribute can be (but is not limited to) the code or identifier of the first label or the second label.

[0072] In an exemplary embodiment, when the label confidence information includes the first label and the first confidence level is greater than a first threshold, the method further includes:

[0073] Step S10102: Obtain the initial position information and target position information of the helmet device, wherein the initial position information includes the initial coordinate information and initial height information of the helmet device within a target time period, and the target position information includes the final coordinate information and final height information of the helmet device within the target time period; determine the coordinate trajectory information of the helmet device based on the initial coordinate information and the final coordinate information; determine the height trajectory information of the helmet device based on the initial height information and the final height information.

[0074] Step S10104: If the coordinate trajectory information satisfies the second condition and / or the height trajectory information satisfies the third condition, the device start-up information is fed back to the external control center.

[0075] In this embodiment, in order to further detect the image recognition results and ensure the accuracy of the recognition results, the process of the helmet moving from the shared device itself to the user is tracked to determine whether the user is indeed wearing the helmet. The tracking results are then fused with the recognition results to achieve high-precision recognition of the helmet.

[0076] The movement tracking of the helmet can be achieved by tracking changes in its position and height. For example, in the shared electric vehicle sector, helmets are typically fixed to the vehicle body. When a user puts the helmet on their head, its height changes by about 30-50 centimeters, and its horizontal position changes by about 30 centimeters. Therefore, by monitoring changes in position coordinates and height using motion monitoring modules such as Bluetooth, the movement of the helmet can be tracked. Initial position information and initial height information include (but are not limited to) GPS coordinate information and horizontal height information when the helmet is not in use. Correspondingly, final position information and final height information include (but are not limited to) GPS coordinate information and horizontal height information after the helmet is in use and before the shared device is activated. Initial position information, final position information, and final height information can all be collected by using a Bluetooth module built into the helmet in conjunction with a GPS / BeiDou positioning module, a height sensor, or a speed sensor. Correspondingly, the shared device can be equipped with a communication module that works with the Bluetooth module to receive position and height information.

[0077] The second condition may (but is not limited to) be that the spatial range of the starting and ending points of the helmet's movement is less than or equal to a preset range. The third condition may (but is not limited to) be that the height of the helmet's movement is less than or equal to a preset value, or that the angle between the line connecting the helmet's ending and starting points and the plane where the starting point is located is less than a preset range, or other conditions.

[0078] An external control center can be (but is not limited to) an external control system or a central control center that communicates with the control terminal or central processing module of the shared equipment. This method enables staff to remotely monitor the shared equipment, thereby allowing the shared equipment to be effectively monitored when it is not in use, thus preventing the theft of the shared equipment.

[0079] It is conceivable that the fusion of recognition results and tracking results includes (but is not limited to) retrieving tracking results when the recognition confidence information contains the first label or the second label and the corresponding confidence is greater than the first threshold, and matching the label confidence information with the tracking results or directly judging the tracking results, thereby achieving the fusion of recognition results and tracking results.

[0080] In an exemplary embodiment, determining the height trajectory information of the helmet device based on the initial height information and the final height information includes:

[0081] Step S101022: Determine the detection reference plane based on the initial height information;

[0082] Step S101024: Based on the initial height information and the final height information, determine the trajectory plane of the helmet device;

[0083] Step S101026: Determine the motion trajectory angle based on the trajectory plane and the detection reference plane, and use the motion trajectory angle as the height trajectory information.

[0084] In this embodiment, the detection reference plane can be (but is not limited to) a horizontal or vertical plane at the initial height of the helmet, so that the influence of slope and other factors on the height of the shared device can be ignored; the trajectory plane can be the plane where the straight line between the initial height position and the final height position is located, and the spatial angle between the two planes is used as the motion trajectory angle.

[0085] It is conceivable that if the angle is too large or too small, it means that the helmet is not worn in the correct position, and then you can take actions such as calling the alarm.

[0086] It should be noted that selecting only the straight line between the initial and final height positions to determine the trajectory plane is to avoid complex motion trajectory calculations during helmet retrieval, thereby saving the computing power required for tracking.

[0087] In an exemplary embodiment, determining the height trajectory information of the helmet device based on the initial height information and the final height information includes:

[0088] Step S101028: Based on the initial height information and the final height information, determine the height difference of the helmet device within the target time, and use the height difference as the height movement trajectory.

[0089] In this embodiment, the height trajectory information can be determined simply by calculating the height difference before and after the helmet change. This method requires very little computing power and can greatly save on configuration costs.

[0090] In one exemplary embodiment, before performing the trajectory determination operation when the helmet attribute information satisfies the first condition, the method further includes:

[0091] Step S101002: When the first timing meets the preset period, a tracking command is sent to the helmet device to instruct the helmet device to provide feedback information;

[0092] Step S101004: Upon receiving the response information from the helmet device, establish a communication link with the helmet device, wherein the communication link is used to perform the trajectory determination operation.

[0093] In this embodiment, a communication link is established before data transmission to ensure the stability and accuracy of subsequent data transmission.

[0094] The first timing can be (but is not limited to) implemented through a periodic timing module within the shared device. The preset period can be a breathing timing period, such as 1 second or 10 minutes.

[0095] According to another embodiment of the invention, such as Figure 2 As shown, a helmet detection device is provided, applied to a shared device that includes a helmet, comprising:

[0096] The initial image acquisition module 22 is used to acquire a sample image set of the target object.

[0097] The attribute recognition module 24 is used to perform attribute recognition on the sample image set to obtain first label information and second label information, wherein the first label information includes a first label that the target object contains helmet equipment information and a first confidence level of the first label, and the second label information includes a second label that the target object does not contain helmet equipment information and a second confidence level of the second label;

[0098] The model training module 26 is used to input the first label, the confidence level of the first label, the second label, and the confidence level of the second label into the initial neural network model, so as to perform model training operation on the initial neural network model and obtain the target neural network model;

[0099] The information analysis module 28 is used to perform label confidence analysis on the target image information through the target neural network model to obtain the label confidence information of the target image, wherein the label confidence information includes at least one of the first label, the first label confidence, the second label, and the second label confidence;

[0100] The instruction sending module 210 is used to confirm that the helmet of the target object is in a first state when the tag confidence information contains the first tag and the first confidence is greater than a first threshold, and to send a posture adjustment instruction to the target object.

[0101] The alarm module 212 is used to confirm that the helmet of the target object is in a second state and to perform an alarm operation when the tag confidence information contains the second tag and the second confidence is greater than a first threshold.

[0102] In one exemplary embodiment, the attribute recognition module includes:

[0103] Image segmentation unit 242 is used to segment the sample image set using a preset segmentation model to obtain initial helmet information;

[0104] The attribute recognition unit 244 is used to perform tag attribute recognition on the initial helmet information through a preset initial neural network model to obtain the first tag information and the second tag information.

[0105] In an optional embodiment, the device further includes:

[0106] The information acquisition module 216 is configured to, when the tag confidence information includes the first tag and the first confidence is greater than a first threshold, acquire the initial position information and target position information of the helmet device, wherein the initial position information includes the initial coordinate information and initial height information of the helmet device within a target time, and the target position information includes the final coordinate information and final height information of the helmet device within the target time; determine the coordinate trajectory information of the helmet device based on the initial coordinate information and the final coordinate information; and determine the height trajectory information of the helmet device based on the initial height information and the final height information.

[0107] The information feedback module 218 is used to feed back device start information to the external control center when it is determined that the coordinate trajectory information meets the second condition and / or the height trajectory information meets the third condition.

[0108] In an optional embodiment, the information acquisition module 216 includes:

[0109] Reference plane determination unit 2162 is used to determine a detection reference plane based on the initial height information;

[0110] The trajectory plane determination unit 2164 is used to determine the trajectory plane of the helmet device based on the initial height information and the final height information;

[0111] The height trajectory determination unit 2166 is used to determine the angle of the motion trajectory based on the trajectory plane and the detection reference plane, and use the angle of the motion trajectory as the height trajectory information.

[0112] In an optional embodiment, the information acquisition module 216 further includes:

[0113] The height difference determination unit 2168 is used to determine the height difference of the helmet device within a target time based on the initial height information and the final height information, and to use the height difference as the height movement trajectory.

[0114] In one exemplary embodiment, the apparatus further includes:

[0115] The tracking unit 2102 is used to send a tracking instruction to the helmet device before performing the trajectory determination operation when the helmet attribute information meets the first condition, so as to instruct the helmet device to provide feedback response information.

[0116] The communication link unit 2104 is used to establish a communication link with the helmet device upon receiving response information from the helmet device, wherein the communication link is used to perform the trajectory determination operation.

[0117] Embodiments of the present invention also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to perform the steps in any of the above method embodiments when executed.

[0118] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0119] Embodiments of the present invention also provide a helmet detection system, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0120] In one exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0121] Specific examples in this embodiment can be found in the examples described in the above embodiments and exemplary implementations, and will not be repeated here.

[0122] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. They can be implemented using computer-executable program code, and thus can be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those described herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

Claims

1. A method for detecting a helmet, characterized in that, Applicable to shared equipment that includes helmet devices, including: Obtain a sample image set of the target object; The sample image set is subjected to attribute recognition to obtain first label information and second label information. The first label information includes a first label that the target object contains helmet equipment information and a first confidence level of the first label. The second label information includes a second label that the target object does not contain helmet equipment information and a second confidence level of the second label. The first label, the confidence score of the first label, the second label, and the confidence score of the second label are input into the initial neural network model to perform model training operations on the initial neural network model and obtain the target neural network model. The target image information is analyzed by the target neural network model to obtain the label confidence information of the target image, wherein the label confidence information includes at least one of the first label, the first label confidence, the second label, and the second label confidence. If the tag confidence information contains the first tag and the first confidence is greater than the first threshold, the helmet of the target object is confirmed to be in the first state, and a posture adjustment command is sent to the target object. If the tag confidence information includes the second tag and the second confidence is greater than the first threshold, the helmet of the target object is confirmed to be in the second state, and an alarm operation is performed.

2. The method according to claim 1, characterized in that, The step of performing attribute recognition on the sample image set to obtain first label information and second label information includes: The sample image set is segmented using a preset segmentation model to obtain initial helmet information; The initial helmet information is labeled with attributes by using a preset initial neural network model to obtain the first label information and the second label information.

3. The method according to claim 1, characterized in that, If the label confidence information includes the first label and the first confidence level is greater than a first threshold, the method further includes: The system acquires the initial position information and target position information of the helmet device, wherein the initial position information includes the initial coordinate information and initial height information of the helmet device within a target time period, and the target position information includes the final coordinate information and final height information of the helmet device within the target time period; based on the initial coordinate information and final coordinate information, the system determines the coordinate trajectory information of the helmet device; based on the initial height information and final height information, the system determines the height trajectory information of the helmet device. If the coordinate trajectory information satisfies the second condition and / or the altitude trajectory information satisfies the third condition, the device start-up information is fed back to the external control center.

4. The method according to claim 3, characterized in that, The process of determining the height trajectory information of the helmet device based on the initial height information and the final height information includes: The detection reference plane is determined based on the initial height information; Based on the initial height information and the final height information, the trajectory plane of the helmet device is determined; Based on the trajectory plane and the detection reference plane, the included angle of the motion trajectory is determined, and the included angle of the motion trajectory is used as the height trajectory information.

5. The method according to claim 3, characterized in that, The process of determining the height trajectory information of the helmet device based on the initial height information and the final height information includes: Based on the initial height information and the final height information, the height difference of the helmet device within the target time is determined, and the height difference is used as the height trajectory information.

6. The method according to claim 1, characterized in that, If the attribute recognition result satisfies the first condition, the method further includes: If the first timing meets the preset period, a tracking command is sent to the helmet device to instruct the helmet device to provide feedback information; Upon receiving a response from the helmet device, a communication link is established with the helmet device.

7. A helmet detection device, characterized in that, Applicable to shared equipment that includes helmet devices, including: The initial image acquisition module is used to acquire a sample image set of the target object; An attribute recognition module is used to perform attribute recognition on the sample image set to obtain first label information and second label information, wherein the first label information includes a first label that the target object contains helmet device information and a first confidence level of the first label, and the second label information includes a second label that the target object does not contain helmet device information and a second confidence level of the second label; The model training module is used to input the first label, the confidence score of the first label, the second label, and the confidence score of the second label into the initial neural network model, so as to perform model training operation on the initial neural network model and obtain the target neural network model; The information analysis module is used to perform label confidence analysis on the target image information through the target neural network model to obtain the label confidence information of the target image, wherein the label confidence information includes at least one of the first label, the first label confidence, the second label, and the second label confidence; The instruction sending module is used to confirm that the helmet of the target object is in a first state when the tag confidence information contains the first tag and the first confidence is greater than a first threshold, and to send a posture adjustment instruction to the target object. An alarm module is used to confirm that the helmet of the target object is in a second state and to perform an alarm operation when the tag confidence information contains the second tag and the second confidence is greater than a first threshold.

8. The apparatus according to claim 7, characterized in that, The attribute recognition module includes: The image segmentation unit is used to segment the sample image set using a preset segmentation model to obtain initial helmet information; The attribute recognition unit is used to perform tag attribute recognition on the initial helmet information through a preset initial neural network model to obtain the first tag information and the second tag information.

9. A helmet detection system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the helmet detection method as claimed in any one of claims 1 to 6.

10. A storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the helmet detection method as claimed in any one of claims 1 to 6.

Images