Method, device and equipment for detecting wearing condition of worker goggles and medium
By combining real scene monitoring pictures and face detection data sets, the face detection and classification models are trained, the problem of sample imbalance in complex environments is solved, the accuracy and generalization ability of goggles wear detection are improved, and false alarms are reduced.
Patent Information
- Application Number
- CN202411919253.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-24
- Publication Date
- 2025-03-18
AI Technical Summary
The existing goggle detection method based on deep learning cannot effectively solve the problem of sample imbalance in complex and changeable actual production scenarios, resulting in low detection accuracy and frequent false alarms.
By combining the real scene pictures of staff who are not wearing goggles in real scene monitoring of industrial production with pictures of staff who are not wearing goggles, and introducing a face detection data set, generating a target data set, training the preset face detection model and face classification model, the trained model is obtained to realize the detection of the goggles worn by staff who are being inspected and monitored.
It improves the accuracy of goggles wearing detection in complex environments, reduces false alarms, enhances the model's face detection and classification capabilities, and is suitable for complex and changeable actual production scenarios.
Smart Images
Figure CN119649304A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of deep learning technology, and in particular to a method, device, equipment and medium for detecting the wearing condition of goggles of staff. Background Art
[0002] At present, the use of deep learning-based detection technology to determine whether workers are wearing goggles can save manual work and achieve efficient detection of goggles wearing. However, the current deep learning-based goggles detection methods are limited to some simple and ideal small-scale scenes and cannot handle complex and changeable actual production scenes, such as special operation platforms and offshore operation platforms. These scenes are very complex and are affected by a variety of factors. Ordinary detection design schemes cannot achieve the same results in small scenes. In addition, the actual production scenes have complex and changeable scenes, difficult sample collection, and serious imbalance of positive and negative samples. As a result, the accuracy of detecting whether workers are wearing goggles is low and false alarms are prone to occur.
[0003] It can be seen from the above that how to solve the problem of sample imbalance in complex environments, improve the accuracy of detection of goggles worn by staff under detection and monitoring, and reduce false alarms are problems to be solved in this field. Summary of the invention
[0004] In view of this, the purpose of the present invention is to provide a method, device, equipment and medium for detecting the wearing condition of goggles of staff, which can solve the problem of sample imbalance in complex environments, improve the accuracy of detecting the wearing condition of goggles of staff to be monitored, and reduce false alarms. The specific scheme is as follows:
[0005] In a first aspect, the present application discloses a method for detecting the wearing condition of goggles of a worker, comprising:
[0006] Combine real scene pictures of workers not wearing goggles in real scene monitoring of industrial production with pictures of people not wearing goggles, and introduce a face detection data set to generate a target data set, and use the target data set to train a preset face detection model and a face classification model to obtain the trained face detection model and the face classification model;
[0007] Based on the trained face detection model, face detection is performed on the staff to be monitored to obtain a face image; the face detection includes image conversion, feature extraction, and feature fusion;
[0008] The face images are classified using the trained face classification model to obtain face images of people who are not wearing goggles in the monitoring to be detected, so as to detect the goggles wearing status of the staff in the monitoring to be detected.
[0009] Optionally, the real scene pictures of workers not wearing goggles in real scene monitoring of industrial production are combined with pictures of people not wearing goggles, and a face detection data set is introduced to generate a target data set, including:
[0010] Segment real scene images of workers not wearing goggles from real scene monitoring of industrial production;
[0011] Cut out images of staff members not wearing goggles and images of network personnel not wearing goggles from the real scene monitoring and network images, and generate images of personnel not wearing goggles based on the images of staff members and network personnel;
[0012] Paste the picture of the person without goggles into the real scene picture of the staff without goggles, and introduce the face detection dataset to generate the target dataset; the face detection dataset is the WiderFace dataset.
[0013] Optionally, the using the target data set to train a preset face detection model and a face classification model includes:
[0014] The target data set is used to train and test the preset face detection model and face classification model; the face detection model is a YOLOv8 model; the face classification model is a MobileNetV3 model.
[0015] Optionally, performing face detection on a staff member to be monitored based on the trained face detection model to obtain a face image includes:
[0016] Obtain the image of the staff member to be monitored;
[0017] The image is input into the trained face detection model to perform image conversion, feature extraction, and feature fusion on the image to output a face image.
[0018] Optionally, the step of inputting the image into the trained face detection model to perform image conversion, feature extraction, and feature fusion on the image to output a face image includes:
[0019] Using the trained face detection model to perform image conversion on the image, and using the backbone network in the trained face detection model to perform feature extraction on the converted image to obtain image feature information of different scales;
[0020] Using the trained Neck module in the face detection model to perform feature fusion on image feature information of different scales to obtain fused features;
[0021] The fused features are processed using the detection head module in the trained face detection model to output a face image.
[0022] Optionally, the method of using the trained face classification model to classify the face image to obtain the face image of the person who is not wearing goggles in the monitoring to be detected includes:
[0023] The face image is subjected to image dimensionality increase, convolution processing, feature enhancement, feature dimensionality reduction, residual splicing, and face part classification using multiple basic modules in the trained face classification model to obtain a face image of a person who is not wearing goggles in the monitoring to be detected.
[0024] Optionally, the method of using the trained face classification model to classify the face image to obtain the face image of the person who is not wearing goggles in the monitoring to be detected includes:
[0025] The trained face classification model is used to classify face images to obtain face images without goggles, face images with goggles, blurred face images, and back face images in the monitoring to be detected.
[0026] In a second aspect, the present application discloses a device for detecting the wearing condition of goggles for staff, comprising:
[0027] A data set generation module, for combining real scene pictures of workers not wearing goggles in real scene monitoring of industrial production with pictures of people not wearing goggles, and introducing a face detection data set to generate a target data set, and using the target data set to train a preset face detection model and a face classification model to obtain the trained face detection model and the face classification model;
[0028] A face detection module is used to perform face detection on the staff to be monitored based on the trained face detection model to obtain a face image; the face detection includes image conversion, feature extraction, and feature fusion;
[0029] The face image classification module is used to classify the face images using the trained face classification model to obtain face images of people who are not wearing goggles in the monitoring to be detected, so as to detect the goggles wearing status of the staff in the monitoring to be detected.
[0030] In a third aspect, the present application discloses an electronic device, including:
[0031] Memory, used to store computer programs;
[0032] A processor is used to execute the computer program to implement the aforementioned method for detecting the wearing condition of goggles by staff.
[0033] In a fourth aspect, the present application discloses a computer storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the steps of the aforementioned method for detecting the wearing condition of goggles of staff disclosed above are implemented.
[0034] It can be seen that the present application provides a method for detecting the wearing status of goggles of staff, including combining real scene pictures of staff not wearing goggles in real scene monitoring of industrial production with pictures of staff not wearing goggles, and introducing a face detection data set to generate a target data set, and using the target data set to train a preset face detection model and a face classification model to obtain the trained face detection model and the face classification model; based on the trained face detection model, face detection is performed on the staff in the monitoring to be detected to obtain a face picture; the face detection includes picture conversion, feature extraction, and feature fusion; the face picture is classified using the trained face classification model to obtain a face picture of the staff not wearing goggles in the monitoring to be detected, so as to realize the detection of the wearing status of goggles of the staff in the monitoring to be detected. In order to solve the problem of serious sample imbalance caused by difficulty in collecting samples in special and complex scenes, the present application combines real scene pictures of staff not wearing goggles collected through real scene monitoring with pictures of staff not wearing goggles. On this basis, in order to avoid the overfitting problem, a face detection dataset is introduced to generate a target dataset. The target dataset is used to train a face detection model and a face classification model to enhance the model's face detection and classification capabilities. The trained face detection model is used to perform face detection on the staff in the monitoring to be detected to obtain face pictures. The trained face classification model is used to classify the face pictures to obtain face pictures of the staff not wearing goggles in the monitoring to be detected, thereby detecting the goggles wearing condition of the staff in the monitoring to be detected. The face position can be detected and separated, and then the face parts can be classified. This design can obtain better detection effects and generalization capabilities in complex and changeable environments, improve the accuracy of detection of the goggles wearing condition of the staff in the monitoring to be detected, and reduce false alarms. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0036] Figure 1 A flow chart of a method for detecting the wearing condition of goggles for staff disclosed in this application;
[0037] Figure 2 This is an example diagram of a problem existing in a complex scenario disclosed in this application;
[0038] Figure 3 This is an overall framework diagram of a worker goggles wearing condition detection method disclosed in the present application;
[0039] Figure 4 This is a schematic diagram of the structure of a device for detecting the wearing condition of goggles for workers disclosed in this application;
[0040] Figure 5 A structural diagram of an electronic device provided for this application. DETAILED DESCRIPTION
[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0042] At present, the use of deep learning-based detection technology to determine whether the operator is wearing goggles can save manpower and achieve efficient detection of goggles. However, the current goggles detection methods based on deep learning are limited to some simple and ideal small-scale scenes, and cannot handle complex and changeable actual production scenes, such as special operation platforms, offshore operation platforms, etc. These scenes are very complex and are affected by a variety of factors. Ordinary detection design schemes cannot achieve the same results in small scenes. In addition, the actual production scenes have complex and changeable scenes, difficult sample collection, and serious imbalance of positive and negative samples. As a result, the accuracy of detecting whether the operator is wearing goggles is low, and false alarms are prone to occur. As can be seen from the above, how to solve the problem of sample imbalance in complex environments, improve the accuracy of detecting the goggles of staff in detection and monitoring, and reduce false alarms are problems to be solved in this field.
[0043] See also Figure 1 As shown, the embodiment of the present invention discloses a method for detecting the wearing condition of goggles of a worker, which may specifically include:
[0044] Step S11: Combine the real scene pictures of workers not wearing goggles in the real scene monitoring of industrial production with the pictures of people not wearing goggles, and introduce a face detection data set to generate a target data set, and use the target data set to train the preset face detection model and face classification model to obtain the trained face detection model and face classification model.
[0045] In this embodiment, real scene pictures of workers not wearing goggles are segmented from real scene monitoring of industrial production; pictures of workers not wearing goggles and network pictures of people not wearing goggles are cropped from the real scene monitoring and network pictures, and pictures of people not wearing goggles are generated based on the workers pictures and the network pictures; the pictures of people not wearing goggles are pasted into the real scene pictures of workers not wearing goggles, and a face detection data set is introduced to generate a target data set, and a preset face detection model and a face classification model are trained using the target data set to obtain the trained face detection model and the face classification model, and then the preset face detection model and the face classification model are trained and tested using the target data set; the face detection data set is a WiderFace data set; the face detection model is a YOLOv8 model; and the face classification model is a MobileNetV3 model.
[0046] Specifically, key frames are extracted from the video data collected by real scene monitoring to obtain a real scene data set including real scene pictures. Then, in order to address the problem of small number of samples without goggles, artificial methods are used to produce samples. Specifically, pictures of network personnel who are not wearing goggles and close to the scene and pictures of staff who are not wearing goggles in real scene monitoring are found to generate pictures of people without goggles. Then, the pictures of people without goggles are combined with real scene pictures of staff who are not wearing goggles in real scene monitoring to obtain scene pictures of people without goggles. On this basis, in order to avoid overfitting problems, the WiderFace data set is mixed in to enhance the face detection ability of the model, generate a target data set, use the target data set to train a preset face detection model to obtain a trained face detection model, process the target data set into a general classification data set form, and then train a face classification model to obtain a trained face classification model.
[0047] Step S12: performing face detection on the staff to be monitored based on the trained face detection model to obtain a face image; the face detection includes image conversion, feature extraction, and feature fusion.
[0048] In this embodiment, a picture of the staff member to be monitored is obtained; the picture is input into the trained face detection model to perform picture conversion, feature extraction, and feature fusion on the picture to output a face picture.
[0049] Among them, the specific process of image conversion, feature extraction and feature fusion of the image is: using the trained face detection model to convert the image, and using the backbone network in the trained face detection model to extract features from the converted image to obtain image feature information of different scales; using the Neck module in the trained face detection model to fuse the feature information of images of different scales to obtain fused features; using the detection head module in the trained face detection model to process the fused features to output a face image.
[0050] After comprehensive consideration and comparison of various target detection models, in order to meet the requirements of real-time and as little video memory usage as possible, the YOLOv8 model is selected as the face detection model. In order to better adapt to the task requirements, the rotating box target detection solution is selected. The specific process is as follows: first, the input image is converted to 640*640 size, and the label is converted proportionally at the same time, and then input into the backbone network for feature extraction. Here, the C2F module is used as the basic module to reduce the number of parameters and improve the calculation efficiency; then the image feature information of different scales in the backbone network is sent to the Neck module for processing, and the feature information of different scales is fused to improve the model's detection ability for targets of different scales and achieve feature enhancement. Finally, the features are processed by the Head module to obtain the prediction box and classification results.
[0051] Step S13: classify the face images using the trained face classification model to obtain face images of people who are not wearing goggles in the monitoring to be detected, so as to detect the goggles wearing status of the staff in the monitoring to be detected.
[0052] In this embodiment, multiple basic modules in the trained face classification model are used to perform image dimensionality upgrade, convolution processing, feature enhancement, feature dimensionality reduction, residual stitching, and face part classification on the face image to obtain face images without goggles, face images with goggles, blurred face images, and back face images in the monitoring to be detected, so as to detect the goggles wearing status of the staff in the monitoring to be detected.
[0053] In this embodiment, in order to classify the facial parts obtained by face detection and find out the faces without goggles, the binary classification method is abandoned, that is, instead of simply classifying the pictures of faces with goggles and without goggles, the face categories are divided into four categories, namely, with goggles, without goggles, blurred faces and back faces. This design is to meet the needs of complex environments. Due to reasons such as lighting, lens blur, water stains, etc., the pictures obtained are often blurred, and the facial parts obtained are also difficult to distinguish. At the same time, blurred faces are closer to faces that can be judged whether they are wearing goggles. Therefore, these faces that cannot be judged are classified as blurred categories for distinction.
[0054] This application adopts the existing classification network. In order to improve the processing speed and meet the real-time requirements, MobileNetV3 with fewer parameters and calculations is selected as the face classification model. MobileNetV3 adds SE modules and improves the activation function. When processing the image, the image is first adjusted to a suitable size and then input into the MobileNetV3 model for processing. MobileNetV3 is mainly composed of multiple Blocks (basic modules) with similar structures. The operation process of each Block for image features is as follows: for image features with an input size of H×W×C, first use a 1×1 convolution to increase the dimension of the feature to obtain a feature map with a size of H×W×d, then use a depth-separable convolution to learn the feature, and then apply the channel attention mechanism to give weights to each channel to enhance the feature, and finally use a 1×1 convolution to reduce the dimension of the feature to obtain a feature map with a size of H×W×f, and perform residual connection. After multiple Block operations, the final high-dimensional image features are obtained. The final classification result is obtained by processing the high-dimensional image features using global average pooling.
[0055] The current goggles detection methods based on deep learning are limited to some simple and ideal small-scale scenarios and cannot handle complex and changeable actual production scenarios, such as special operation platforms and offshore operation platforms. The problems existing in complex scenarios are as follows: Figure 2 As shown in the figure, these scenes are very complex and are affected by a variety of factors. Ordinary detection design schemes cannot achieve the same results on small scenes. The detection method for staff goggles wearing proposed in this application is mainly divided into three parts: data set preparation, face detection, and face classification. The overall framework is as follows Figure 3As shown: In the first part, we need to address the problem of serious sample imbalance caused by difficulty in collecting samples in special and complex scenes. We use real scene data, map data and WiderFace dataset to create an available dataset that can support training, namely the target dataset, and train the preset face detection model and face classification model. In the second part, we can use the existing target detection network to implement it. We use the trained face detection model to perform face detection on the staff in the monitoring to be detected, and can accurately detect the faces in the pictures. In the third part, we use the trained face classification model to classify the face pictures, and obtain the face pictures of the people who are not wearing goggles in the monitoring to be detected, which can classify the face pictures obtained in the second part.
[0056] The innovation of this application lies in: proposing a method for detecting the wearing of goggles by workers in complex and changeable real production scenes, which can solve the problems of complex and diverse scenes, difficult sample collection, and serious imbalance of positive and negative samples, including a data set preparation method for sample imbalance and a two-stage method for detecting the wearing of goggles by workers; the data set preparation method takes real scene monitoring frame extraction pictures as the main body, and at the same time, the pictures of people not wearing goggles in real scenes and network people not wearing goggles in network pictures are segmented and pasted into the real scene pictures of workers not wearing goggles to increase the number of samples not wearing goggles, and finally, in order to enhance network performance and prevent overfitting, the WiderFace data set is mixed in to reduce the overfitting problem and solve the problems of difficult sample collection and sample imbalance; the one-stage goggles detection suitable for simple scenes in the traditional scheme is transformed into a two-stage goggles detection method, the first stage focuses on detecting the position of the face and separating it, and the second stage classifies the face parts. This design can obtain better detection effect and generalization ability in complex and changeable environments, enhance the accuracy of goggles detection, and reduce false alarms.
[0057] In this embodiment, real scene pictures of workers not wearing goggles in real scene monitoring of industrial production are combined with pictures of workers not wearing goggles, and a face detection data set is introduced to generate a target data set. The preset face detection model and face classification model are trained using the target data set to obtain the trained face detection model and face classification model; face detection is performed on the workers in the monitoring to be detected based on the trained face detection model to obtain face pictures; the face detection includes picture conversion, feature extraction, and feature fusion; the face pictures are classified using the trained face classification model to obtain face pictures of people not wearing goggles in the monitoring to be detected, so as to realize the detection of goggles wearing status of the workers in the monitoring to be detected. In order to solve the problem of serious sample imbalance caused by difficulty in collecting samples in special and complex scenes, the present application combines real scene pictures of staff not wearing goggles collected through real scene monitoring with pictures of staff not wearing goggles. On this basis, in order to avoid the overfitting problem, a face detection dataset is introduced to generate a target dataset. The target dataset is used to train a face detection model and a face classification model to enhance the model's face detection and classification capabilities. The trained face detection model is used to perform face detection on the staff in the monitoring to be detected to obtain face pictures. The trained face classification model is used to classify the face pictures to obtain face pictures of the staff not wearing goggles in the monitoring to be detected, thereby detecting the goggles wearing condition of the staff in the monitoring to be detected. The face position can be detected and separated, and then the face parts can be classified. This design can obtain better detection effects and generalization capabilities in complex and changeable environments, improve the accuracy of detection of the goggles wearing condition of the staff in the monitoring to be detected, and reduce false alarms.
[0058] See also Figure 4 As shown, the embodiment of the present invention discloses a device for detecting the wearing condition of goggles of staff, which may specifically include:
[0059] The data set generation module 11 is used to combine the real scene pictures of workers not wearing goggles in the real scene monitoring of industrial production with the pictures of people not wearing goggles, and introduce the face detection data set to generate a target data set, and use the target data set to train the preset face detection model and face classification model to obtain the trained face detection model and face classification model;
[0060] A face detection module 12 is used to perform face detection on the staff to be monitored based on the trained face detection model to obtain a face image; the face detection includes image conversion, feature extraction, and feature fusion;
[0061] The face image classification module 13 is used to classify the face images using the trained face classification model to obtain face images of people who are not wearing goggles in the monitoring to be detected, so as to detect the goggles wearing status of the staff in the monitoring to be detected.
[0062] In some specific embodiments, the data set generating module 11 may specifically include:
[0063] A segmentation module is used to segment the real scene images of workers who are not wearing goggles from the real scene monitoring of industrial production;
[0064] A personnel picture generation module, used to cut out pictures of staff members not wearing goggles and pictures of network personnel not wearing goggles from the real scene monitoring and network pictures, and generate pictures of personnel not wearing goggles based on the pictures of staff members and the pictures of network personnel;
[0065] The pasting module is used to paste the picture of the person not wearing goggles into the real scene picture of the staff not wearing goggles, and introduce the face detection dataset to generate the target dataset; the face detection dataset is the WiderFace dataset.
[0066] In some specific embodiments, the data set generating module 11 may specifically include:
[0067] The training and testing module is used to train and test the preset face detection model and face classification model using the target data set; the face detection model is the YOLOv8 model; the face classification model is the MobileNetV3 model.
[0068] In some specific embodiments, the face detection module 12 may specifically include:
[0069] An image acquisition module is used to acquire images of the staff to be monitored;
[0070] The image input module is used to input the image into the trained face detection model to perform image conversion, feature extraction, and feature fusion on the image to output a face image.
[0071] In some specific embodiments, the face detection module 12 may specifically include:
[0072] An image conversion and feature extraction module, used to perform image conversion on the image using the trained face detection model, and to perform feature extraction on the converted image using the backbone network in the trained face detection model, so as to obtain image feature information of different scales;
[0073] A feature fusion module, used to perform feature fusion on image feature information of different scales using the trained Neck module in the face detection model to obtain fused features;
[0074] The feature processing module is used to process the fused features using the detection head module in the trained face detection model to output a face image.
[0075] In some specific embodiments, the face image classification module 13 may specifically include:
[0076] The face image processing module is used to use multiple basic modules in the trained face classification model to perform image dimensionality upgrade, convolution processing, feature enhancement, feature dimensionality reduction, residual splicing, and face part classification on the face image to obtain a face image of the person not wearing goggles in the monitoring to be detected.
[0077] In some specific embodiments, the face image classification module 13 may specifically include:
[0078] The classification module is used to classify face images using the trained face classification model to obtain face images without goggles, face images with goggles, blurred face images and back face images in the monitoring to be detected.
[0079] Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the method for detecting the wearing condition of goggles of a worker performed by the electronic device disclosed in any of the aforementioned embodiments.
[0080] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.
[0081] In addition, the memory 22, as a carrier for storing resources, can be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon include an operating system 221, a computer program 222 and data 223, etc. The storage method can be temporary storage or permanent storage.
[0082] Among them, the operating system 221 is used to manage and control the hardware devices and computer programs 222 on the electronic device 20, so as to realize the operation and processing of the data 223 in the memory 22 by the processor 21, which can be Windows, Unix, Linux, etc. In addition to including a computer program that can be used to complete the staff goggles wearing condition detection method performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks. In addition to including data transmitted from an external device received by the staff goggles wearing condition detection device, the data 223 can also include data collected by its own input and output interface 25, etc.
[0083] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
[0084] Furthermore, an embodiment of the present application also discloses a computer-readable storage medium, in which a computer program is stored. When the computer program is loaded and executed by a processor, the steps of the method for detecting the wearing condition of goggles of staff disclosed in any of the aforementioned embodiments are implemented.
[0085] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used 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 presence of other identical elements in the process, method, article or device including the elements.
[0086] The above is a detailed introduction to a method, device, equipment and storage medium for detecting the wearing condition of goggles for staff provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, according to the idea of the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.
Claims
1. A method for detecting the wearing condition of goggles by staff, characterized in that: include: Combine real scene pictures of workers not wearing goggles in real scene monitoring of industrial production with pictures of people not wearing goggles, and introduce a face detection data set to generate a target data set, and use the target data set to train a preset face detection model and a face classification model to obtain the trained face detection model and the face classification model; Based on the trained face detection model, face detection is performed on the staff to be monitored to obtain a face image; the face detection includes image conversion, feature extraction, and feature fusion; The face images are classified using the trained face classification model to obtain face images of people who are not wearing goggles in the monitoring to be detected, so as to detect the goggles wearing status of the staff in the monitoring to be detected.
2. The method for detecting the wearing condition of goggles of staff according to claim 1, characterized in that: The real scene pictures of workers not wearing goggles in real scene monitoring of industrial production are combined with pictures of people not wearing goggles, and a face detection data set is introduced to generate a target data set, including: Segment real scene images of workers not wearing goggles from real scene monitoring of industrial production; Cut out images of staff members not wearing goggles and images of network personnel not wearing goggles from the real scene monitoring and network images, and generate images of personnel not wearing goggles based on the images of staff members and network personnel; Paste the picture of the person without goggles into the real scene picture of the staff without goggles, and introduce the face detection dataset to generate the target dataset; the face detection dataset is the WiderFace dataset.
3. The method for detecting the wearing condition of goggles of staff according to claim 1, characterized in that: The using the target data set to train a preset face detection model and a face classification model includes: The target data set is used to train and test the preset face detection model and face classification model; the face detection model is a YOLOv8 model; the face classification model is a MobileNetV3 model.
4. The method for detecting the wearing condition of goggles of workers according to claim 1, characterized in that: The face detection of the staff to be detected and monitored is performed based on the trained face detection model to obtain a face image, including: Obtain the image of the staff member to be monitored; The image is input into the trained face detection model to perform image conversion, feature extraction, and feature fusion on the image to output a face image.
5. The method for detecting the wearing condition of goggles of workers according to claim 4, characterized in that: The step of inputting the image into the trained face detection model to perform image conversion, feature extraction, and feature fusion on the image to output a face image includes: Using the trained face detection model to perform image conversion on the image, and using the backbone network in the trained face detection model to perform feature extraction on the converted image to obtain image feature information of different scales; Using the trained Neck module in the face detection model to perform feature fusion on image feature information of different scales to obtain fused features; The fused features are processed using the detection head module in the trained face detection model to output a face image.
6. The method for detecting the wearing condition of goggles of workers according to claim 1, characterized in that: The method of using the trained face classification model to classify the face image to obtain the face image of the person who is not wearing goggles in the monitoring to be detected includes: The face image is subjected to image dimensionality increase, convolution processing, feature enhancement, feature dimensionality reduction, residual splicing, and face part classification using multiple basic modules in the trained face classification model to obtain a face image of a person who is not wearing goggles in the monitoring to be detected.
7. The method for detecting the wearing condition of goggles of workers according to any one of claims 1 to 6, characterized in that: The method of using the trained face classification model to classify the face image to obtain the face image of the person who is not wearing goggles in the monitoring to be detected includes: The trained face classification model is used to classify face images to obtain face images without goggles, face images with goggles, blurred face images, and back face images in the monitoring to be detected.
8. A device for detecting the wearing condition of goggles for workers, characterized in that: include: A data set generation module, for combining real scene pictures of workers not wearing goggles in real scene monitoring of industrial production with pictures of people not wearing goggles, and introducing a face detection data set to generate a target data set, and using the target data set to train a preset face detection model and a face classification model to obtain the trained face detection model and the face classification model; A face detection module is used to perform face detection on the staff to be monitored based on the trained face detection model to obtain a face image; the face detection includes image conversion, feature extraction, and feature fusion; The face image classification module is used to classify the face images using the trained face classification model to obtain face images of people who are not wearing goggles in the monitoring to be detected, so as to detect the goggles wearing status of the staff in the monitoring to be detected.
9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor is used to execute the computer program to implement the method for detecting the wearing condition of goggles by staff as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: Used to store computer programs; wherein, when the computer program is executed by a processor, it implements the method for detecting the wearing condition of goggles for staff as described in any one of claims 1 to 7.