Qualification identification method and system for high-altitude operation personnel

Abstract

The embodiment of the invention discloses a qualification identification method and system for high-altitude operation personnel. The method comprises the following steps: in response to a job ticket downloading request, determining a target job ticket, the target job ticket comprising a target face image; determining a to-be-detected face image based on the unmanned aerial vehicle, wherein the to-be-detected face image comprises the face of the to-be-detected aerial work personnel; determining a to-be-detected forward face image based on the face alignment model and the to-be-detected face image; wherein the face alignment model is a pre-trained network model used for realizing a face alignment function; and according to the target face image and the to-be-detected forward face image, carrying out qualification identification on the to-be-detected high-altitude operation personnel. According to the scheme, the to-be-detected face image of the high-altitude operation personnel can be determined through the unmanned aerial vehicle, active cooperation of the operation personnel is not needed, face alignment is performed on the to-be-detected face image through the face alignment model, personnel qualification recognition is performed on the basis of the to-be-detected forward face image after face alignment, and the accuracy and efficiency of qualification verification are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to a method and system for identifying the qualifications of personnel working at heights. Background Art

[0002] The construction site of power transmission lines is complex and dangerous. In order to ensure construction safety, strict qualification review of on-site workers is required. The traditional method is to use a surveillance ball to obtain the facial image of the workers and conduct qualification review based on facial recognition technology. This method often requires the workers to actively cooperate, such as keeping the camera facing the surveillance ball still. However, at the construction site of power transmission lines, workers are often too busy to actively cooperate with facial recognition, which is not conducive to the convenience and practicality of qualification review. In addition, in the application scenario where the surveillance ball recognizes from bottom to top, facial recognition may fail or be misrecognized due to the skewed angle of the face, which greatly reduces the accuracy and efficiency of the qualification review. Summary of the invention

[0003] The present invention provides a qualification identification method and system for high-altitude workers, which can automatically capture the face image to be tested of the high-altitude workers through an unmanned aerial vehicle, without the need for the workers to actively participate and cooperate, and use a face correction model to correct the face image to be tested, so as to perform personnel qualification identification based on the positive face image to be tested after the face correction, thereby greatly improving the accuracy and efficiency of qualification review, and enhancing the convenience and practicality of qualification review.

[0004] According to one aspect of the present invention, a method for identifying the qualifications of a person working at height is provided, the method comprising:

[0005] In response to a job ticket download request, determining a target job ticket, wherein the target job ticket includes a target face image;

[0006] Determine a face image to be tested based on the drone, wherein the face image to be tested includes the face of a high-altitude worker to be tested;

[0007] Determining a positive face image to be tested based on a face rotation model and the face image to be tested; wherein the face rotation model refers to a pre-trained network model for realizing a face rotation function;

[0008] According to the target facial image in the target operation ticket and the positive facial image to be tested, the qualification of the high-altitude operation personnel to be tested is identified.

[0009] According to another aspect of the present invention, a qualification identification system for high-altitude workers is provided, the system comprising a drone, a cloud server and an edge server, wherein:

[0010] The drone is used to determine a face image to be tested, and send the face image to be tested to the edge server, wherein the face image to be tested includes the face of a high-altitude worker to be tested;

[0011] The cloud server is used to respond to the job ticket download request of the edge server, determine the target job ticket, and send the target job ticket to the edge server, wherein the target job ticket includes a target face image;

[0012] The edge server is used to send the work ticket download request to the cloud server, receive the target work ticket sent by the cloud server, and determine the frontal face image to be tested based on the face rotation model and the frontal face image to be tested, and identify the qualifications of the high-altitude worker to be tested according to the target face image and the frontal face image to be tested; wherein the face rotation model refers to a pre-trained network model for realizing the face rotation function.

[0013] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0014] at least one processor; and

[0015] a memory communicatively connected to the at least one processor; wherein,

[0016] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the method for identifying the qualifications of aerial work personnel described in any embodiment of the present invention.

[0017] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the method for identifying the qualifications of aerial workers described in any embodiment of the present invention when executed.

[0018] The technical solution of the embodiment of the present invention, in response to a work ticket download request, first determines the target work ticket, the target work ticket includes a target facial image, then determines the face image to be tested based on the drone, the face image to be tested includes the face of the high-altitude worker to be tested; then determines the positive face image to be tested based on the face correction model and the face image to be tested; wherein the face correction model refers to a pre-trained network model for realizing the face correction function; and then performs qualification identification on the high-altitude worker to be tested based on the target facial image in the target work ticket and the positive face image to be tested. This technical solution can automatically capture the facial images of high-altitude workers through drones, without the need for active participation and cooperation of high-altitude workers, thus improving the efficiency of qualification review. The facial images to be tested are corrected through the face correction model, so that personnel qualifications can be identified based on the positive face images to be tested after the face correction. This can effectively avoid facial recognition failures or misidentification of faces caused by facial angle deviation, greatly improving the accuracy of personnel qualification review, while enhancing the convenience and practicality of qualification review, helping to improve the safety of transmission line construction sites, and reducing safety risks caused by personnel qualification issues, thereby providing strong guarantees for the sustainable development of the power industry. In addition, this technical solution can also be applied to other scenarios that require non-cooperative qualification identification, such as security monitoring of public places such as large-scale events and transportation hubs.

[0019] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 is a flow chart of a method for identifying qualifications of personnel working at heights provided in accordance with the first embodiment of the present invention;

[0022] Figure 2 is a flow chart of a method for identifying qualifications of personnel working at heights provided in accordance with a second embodiment of the present invention;

[0023] Figure 3 This is a workflow diagram of a method for identifying qualifications of personnel working at heights provided in accordance with Embodiment 2 of the present invention;

[0024] Figure 4This is a workflow diagram of another method for identifying qualifications of personnel working at heights provided in accordance with the second embodiment of the present invention;

[0025] Figure 5 This is a schematic diagram of the structure of a qualification identification system for high-altitude workers provided according to Embodiment 3 of the present invention;

[0026] Figure 6 The present invention is a schematic diagram of the structure of an electronic device for implementing a method for identifying qualifications of personnel working at heights according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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 should fall within the scope of protection of the present invention.

[0028] It should be noted that the terms "first", "second", "target", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0029] Embodiment 1

[0030] Figure 1 This is a flowchart of a method for identifying qualifications of high-altitude workers provided in the first embodiment of the present invention. This embodiment is applicable to the case of non-cooperative qualification identification of high-altitude workers. The method can be executed by a qualification identification system for high-altitude workers. The qualification identification system for high-altitude workers can be implemented in the form of hardware and / or software. The qualification identification system for high-altitude workers can be configured in an electronic device with data processing capabilities. Figure 1 As shown, the method includes:

[0031] S110, in response to the job ticket download request, determining a target job ticket, wherein the target job ticket includes a target face image.

[0032] The technical solution provided by the embodiment of the present invention is implemented based on the data processing of the drone, edge server and cloud server and the information transmission between the three. The cloud server stores the personal information of the power construction personnel, which includes the qualification information of the construction personnel and the standard face image. The edge server deploys a face correction model, which can be used to provide computing power support. The drone is used as a shooting tool, and is equipped with a video monitoring device (such as a high-definition camera) and an image processing device.

[0033] The work ticket download request may refer to an instruction for requesting to download a construction work ticket. The target work ticket may refer to a work ticket used to guide this construction, specifically including the personal information of the high-altitude workers who need to be on-site for construction. The target face image refers to a standard face image of the high-altitude workers who need to be on-site for construction.

[0034] Specifically, the edge server sends a job ticket download request to the cloud server. After receiving the job ticket download request, the cloud server responds, determines the target job ticket required for this construction, and sends the target job ticket to the edge server.

[0035] S120, determining a face image to be tested based on the drone, where the face image to be tested includes a face of a high-altitude worker to be tested.

[0036] The high-altitude workers to be tested may refer to high-altitude workers who are waiting for qualification identification. In this embodiment, after determining the target work ticket, a face image to be tested may be determined based on a drone, and the face image to be tested includes the face of the high-altitude workers to be tested. Optionally, determining the face image to be tested based on a drone includes the following steps A1-A3:

[0037] A1. Collect video stream data of high-altitude workers to be tested through drones.

[0038] When the UAV flies over the transmission line operation surface, the construction workers on the operation surface (i.e., the high-altitude workers to be measured) are taken as the collection objects, and the video monitoring device of the UAV is used to collect video stream data.

[0039] A2. Perform frame processing on the video stream data to obtain the frame images to be tested.

[0040] Among them, the framed images to be tested may refer to multiple frames of images obtained after the video stream data of the high-altitude workers to be tested are framed. After obtaining the video stream data, the acquired video stream data may be framed using the image processing device of the drone to obtain the framed images to be tested. Furthermore, in order to improve the efficiency of personnel qualification recognition, after the video stream data is framed, part of the images may be extracted from the framed images according to a preset sampling ratio as the framed images to be tested.

[0041] A3. Perform face detection on the framed images to be tested, and obtain the face images to be tested corresponding to the high-altitude workers to be tested.

[0042] Among them, a face detection model is pre-deployed in the image processing device of the drone. After obtaining the frame images to be tested, the face detection model can be used to extract the corresponding face images from each frame image to be tested as the face images to be tested. It should be noted that this embodiment does not specifically limit the face detection model, as long as the network model that can realize the face detection function in the image is available.

[0043] Exemplarily, the face detection model can adopt the LFFD model. Among them, the LFFD model is a face detection model based on deep learning. The goal of this model is to reduce the size of the model as much as possible and improve the model detection speed while maintaining high accuracy. Specifically, the network of the LFFD model consists of 25 convolutional layers, multiple residual modules and ReLU activation functions, and the convolution kernel size of the convolutional layer is 3×3 or 1×1. The network of the LEED model has 8 branches, each of which is used to detect faces of different sizes; each branch is divided into two sub-branches, which are used to perform face binary classification tasks and prediction box regression tasks respectively. For the face binary classification task, the softmax function with cross loss is used as the loss function; for the prediction box regression task, the L2 function is used as the loss function. The input of the face detection model is the frame image to be tested, and the output is the face prediction box contained in the frame image to be tested. For the face prediction box output by the face detection model, the opencv function is used to intercept it, and the face image to be tested corresponding to the high-altitude worker to be tested can be obtained.

[0044] S130, determining a frontal face image to be tested based on the face normalization model and the face image to be tested.

[0045] Among them, the face rotation model refers to a pre-trained network model for realizing the face rotation function, which can automatically adjust the face angle in the image so that it faces the video surveillance device, thereby eliminating the difficulty of face recognition caused by the deviation of the face angle. Exemplarily, the face rotation model can be obtained by training DR-GAN (Distribution Regularization for Text-to-Image Generation). Among them, DR-GAN is composed of a generator and a discriminator. The generator adopts an encoder-decoder structure, and the discriminator is a multi-task CNN (Convolutional Neural Network). Furthermore, the encoder is symmetrical with the encoder, both have 14 layers of convolution, and the convolution kernel size is 3×3, with a step size of 1 or 2.

[0046] In this embodiment, after determining the face image to be tested, the face image to be tested is sent to the edge server by the drone. The edge server inputs the received face image to be tested into the face normalization model, performs a face normalization operation on the face image to be tested through the face normalization model, and outputs the positive face image to be tested corresponding to the face image to be tested.

[0047] In this embodiment, optionally, the training process of the face rotation model includes the following steps B1-B4:

[0048] B1. Obtain a face dataset and add label information to each reference face image in the face dataset.

[0049] The label information includes identity information and posture information. The face dataset includes multiple reference face images, which are real face images. Specifically, before starting the formal training of the face normalization model, it is necessary to first obtain a face dataset including multiple reference face images, and manually add label information (including identity information and posture information) to each reference face image.

[0050] B2. Input each reference face image into the generator of the distribution regularized generative adversarial network to obtain multiple generated face images.

[0051] The generated face image may refer to a simulated image generated by a generator of a distribution regularized generative adversarial network, and the generated face image has a positive facial posture. Specifically, after adding label information to each reference face image in the face data set, each reference face image carrying the label information may be sequentially input into the generator of the distribution regularized generative adversarial network, and the generator outputs a plurality of corresponding generated face images. Optionally, step B2 specifically includes the following steps B21-B22:

[0052] B21. Input each reference face image into the encoder of the generator to obtain the reference face features corresponding to each reference face image.

[0053] The reference facial features may refer to facial features extracted from the reference facial images. Specifically, each reference facial image is input into the distribution regularized generative adversarial network, and the encoder processes each reference facial image to generate corresponding facial features as the reference facial features.

[0054] B22. The reference facial features are integrated with the preset noise information and the preset facial posture information through the decoder of the generator, and the generated facial image is obtained after deconvolution processing.

[0055] The preset noise information may refer to a preset noise signal. The preset facial posture information may refer to a preset facial posture signal. Specifically, after the encoder outputs the reference facial features, the reference facial features are input into the decoder, and the decoder connects and integrates the reference facial features with the preset noise information and the preset facial posture information, and then outputs a generated facial image with a positive facial posture after deconvolution processing.

[0056] B3. Input the generated face image and its corresponding reference face image into the discriminator of the distribution regularized generative adversarial network for image discrimination, and determine the generator loss value and the discriminator loss value in the discrimination process.

[0057] Specifically, after obtaining the generated face image, the generated face image and its corresponding reference face image are input into the discriminator of the distribution regularized generative adversarial network, and the discriminator is used to determine the authenticity of the image, and the face image judged as true is classified according to identity and posture. In the image discrimination process, it is necessary to calculate the loss value of the generator loss function (i.e., the generator loss value) and the loss value of the discriminator loss function (i.e., the discriminator loss value). Among them, the optimization goal of the generator loss function is to maximize the probability that the discriminator will judge the generated face image as true, and to maximize the probability that the generated face image is classified into the correct identity category and posture category; the optimization goal of the discriminator loss function is to maximize the probability that the reference face image is judged as true, and to maximize the probability that the generated face image is judged as false.

[0058] For example, the generator loss function can be expressed as follows:

[0059]

[0060] Where G represents the generator; D represents the discriminator; E represents the expected value of the distribution function; x represents the real sample (i.e., the reference face image); y represents the real sample label (i.e., the label information of the reference face image); p d represents the distribution of real samples and real sample labels; p z represents the distribution of noise; p c represents the distribution of facial pose encoding; y d is the identity label; yt is the specified posture label; z is the preset noise information, and c is the preset facial posture information; Indicates the identity category of the identity tag yd. Indicates the posture category of the specified posture label yt.

[0061] For example, the discriminator loss function can be expressed as follows:

[0062]

[0063] Among them, G represents the generator; D represents the discriminator; E represents the expected value of the distribution function; x represents the real sample; y represents the real sample label; p d represents the distribution of real samples and real sample labels; y d is the identity label; p is the posture label; Represents the identity label y d The identity category; Represents the pose label y p The posture category; z is the preset noise information, c is the preset facial posture information; p z represents the distribution of noise; p c represents the distribution of facial posture; N d is the total number of identity categories of all face images in the face dataset, N d +1 is a category outside of all identity categories in the face dataset, and is used to store face images that are judged as false by the discriminator.

[0064] B4. When the generator loss value and the discriminator loss value reach convergence, the generator is used as the face normalization model.

[0065] Specifically, after determining the generator loss value and the discriminator loss value, it is necessary to determine whether the generator loss value and the discriminator loss value have reached convergence (decreased and stabilized). If both the generator loss value and the discriminator loss value have reached convergence, the corresponding generator at this time will be used as the face normalization model.

[0066] S140, performing qualification identification on the high-altitude worker to be tested based on the target face image and the positive face image to be tested.

[0067] In this embodiment, after determining the frontal face image to be tested, the qualification of the high-altitude worker to be tested can be identified based on the target face image and the frontal face image to be tested. Optionally, the qualification of the high-altitude worker to be tested is identified based on the target face image and the frontal face image to be tested, including the following steps C1-C2:

[0068] C1. Match the facial features of the tested positive face image with the target face image.

[0069] Among them, a face recognition model is also deployed in the edge server, and the face recognition model can be used to perform feature matching on faces in images. Exemplarily, the edge server inputs each positive face image to be tested and the target face image into the face recognition model, and matches the face features of the positive face image to be tested with the target face image through the face recognition model, thereby obtaining a face matching result. Among them, the face matching result is a face matching success or a face matching failure.

[0070] C2. Determine the qualification recognition result of the high-altitude worker to be tested based on the facial feature matching result.

[0071] Specifically, if the face matching result is a successful face match, it can be determined that the qualification recognition result of the high-altitude worker to be tested corresponding to the forward face image to be tested is passed the qualification verification; if the face matching result is a failed face match, it can be determined that the qualification recognition result of the high-altitude worker to be tested corresponding to the forward face image to be tested is failed the qualification verification.

[0072] The technical solution of the embodiment of the present invention, in response to a work ticket download request, first determines the target work ticket, the target work ticket includes a target facial image, then determines the face image to be tested based on the drone, the face image to be tested includes the face of the high-altitude worker to be tested; then determines the positive face image to be tested based on the face correction model and the face image to be tested; wherein the face correction model refers to a pre-trained network model for realizing the face correction function; and then performs qualification identification on the high-altitude worker to be tested based on the target facial image and the positive face image to be tested. This technical solution can automatically capture the facial images of high-altitude workers through drones, without the need for active participation and cooperation of high-altitude workers, thus improving the efficiency of qualification review. The facial images to be tested are corrected through the face correction model, so that personnel qualifications can be identified based on the positive face images to be tested after the face correction. This can effectively avoid facial recognition failures or misidentification of faces caused by facial angle deviation, greatly improving the accuracy of personnel qualification review, while enhancing the convenience and practicality of qualification review, helping to improve the safety of transmission line construction sites, and reducing safety risks caused by personnel qualification issues, thereby providing strong guarantees for the sustainable development of the power industry. In addition, this technical solution can also be applied to other scenarios that require non-cooperative qualification identification, such as security monitoring of public places such as large-scale events and transportation hubs.

[0073] Embodiment 2

[0074] Figure 2 A flowchart of a method for identifying qualifications of personnel working at heights provided in the second embodiment of the present invention, this embodiment is optimized based on the above-mentioned embodiment. The specific optimization is: before determining the positive face image to be tested based on the face rotation model and the face image to be tested, it also includes: determining whether there is an occlusion on the face in the face image to be tested; if so, determining the unobstructed face image to be tested based on the face occlusion elimination model and the face image to be tested; wherein the face occlusion elimination model refers to a pre-trained network model for realizing the face occlusion elimination function; the face image to be tested is updated according to the unobstructed face image to be tested.

[0075] like Figure 2 As shown, the method of this embodiment specifically includes the following steps:

[0076] S210, in response to a job ticket download request, determining a target job ticket, wherein the target job ticket includes a target face image.

[0077] S220, determining a face image to be tested based on the drone, where the face image to be tested includes a face of a high-altitude worker to be tested.

[0078] S230, determining whether there is any obstruction on the face in the face image to be tested.

[0079] It should be noted that due to the complexity of the activities of the operators and the complexity of the construction site environment, the facial images to be tested collected by the drone may not only have the problem of facial angle deviation, but also the problem of partial facial occlusion. These problems will affect the accuracy and efficiency of facial recognition, and thus affect the accuracy and efficiency of personnel qualification recognition. Therefore, in order to further improve the accuracy of personnel qualification recognition, after determining the facial image to be tested based on the drone, it is necessary to determine whether there is an obstruction on the face in the facial image to be tested. For example, the obstruction is a helmet, a mask, etc.

[0080] Among them, a face occlusion detection model is also deployed in the edge server, which can be used to detect whether there is occlusion on the face in the image. Exemplarily, the face occlusion positions can be pre-set to 5 areas, including the left eye, right eye, nose, mouth and chin, and the face occlusion detection model is obtained through model training based on the Tensorflow+Keras generative adversarial network. The edge server inputs the face image to be tested into the face occlusion detection model, and according to the model output, it can be determined whether there is an occlusion on the face in the test face image. Performing face occlusion detection on the face image to be tested helps to build a high-quality face recognition database.

[0081] S240: If so, determine the unobstructed face image to be tested based on the face occluder removal model and the face image to be tested.

[0082] Among them, the face occluder removal model refers to a pre-trained network model for realizing the face occluder removal function, which can automatically identify and remove face occluders in the image to ensure the clarity and recognizability of facial features.

[0083] Exemplarily, the face occlusion removal model can be obtained based on the training of a generative adversarial network. The generative adversarial network is composed of a generative network and an adversarial network. The generative network can realize the function of converting an occluded face image into an unoccluded face image, and there is no need for any pairing relationship between the training image sets used by the generative network and the adversarial network, thereby greatly reducing the difficulty of preparing training data.

[0084] Specifically, the input of the generative network is a real occluded face image, and its training goal is to optimize and adjust the network parameters of the generative network so that the output of the generative network is a simulated unobstructed face image corresponding to the real occluded face image; the input of the adversarial network is a real unobstructed face image and a simulated unobstructed face image output by the generative network, and its training goal is to optimize and adjust the network parameters of the adversarial network so that the adversarial network can distinguish the real unobstructed face image from the simulated unobstructed face image output by the generative network (i.e., perform binary classification). The generative network and the adversarial network are trained simultaneously, influencing and promoting each other.

[0085] In this embodiment, if it is determined that there is an occluder on the face in the face image to be tested, the face image to be tested needs to be input into the face occluder removal model, and the face occluder removal operation is performed on the face image to be tested by the face occluder removal model, thereby obtaining a face image to be tested without occlusion. In addition, if it is determined that there is no occluder on the face in the face image to be tested, step S260 is directly executed.

[0086] In this embodiment, optionally, the training process of the face occlusion removal model includes steps D1-D4:

[0087] D1. Input a reference occluded face image in the first training image set into a preset generation network to generate an unoccluded face image.

[0088] The preset generation network refers to an artificial neural network for converting an occluded face image into an unobstructed face image. The first training image set can be used as a training data set for the preset generation network, including multiple real occluded face images. The reference occluded face image can refer to any real occluded face image in the first training image set. The generated unobstructed face image can refer to a simulated occluded face image corresponding to the reference occluded face image generated by the preset generation network.

[0089] Specifically, firstly, a real occluded face image is randomly selected from the first training image set as a reference occluded face image X, and the reference occluded face image X is input into the preset generation network G1 to obtain a generated unoccluded face image Y1=G1(X).

[0090] D2. Input a reference unobstructed face image and a generated unobstructed face image in the second training image set into a preset adversarial network to obtain a first probability and a second probability.

[0091] The preset adversarial network refers to an artificial neural network used to classify and discriminate the reference unobstructed face image and the generated unobstructed face image, the first probability refers to the probability of discriminating the reference unobstructed face image as the unobstructed face image, and the second probability refers to the probability of discriminating the generated unobstructed face image as the unobstructed face image. The second training image set can be used as a training data set for the preset adversarial network, including multiple real unobstructed face images. The reference unobstructed face image can refer to any real unobstructed face image in the second training image set.

[0092] Specifically, a real unobstructed face image is randomly selected from the second training image set as the reference unobstructed face image Y2, and the reference unobstructed face image Y2 and the generated unobstructed face image Y1 are respectively input into the preset adversarial network D1 to obtain the probability of judging that the two face images are unobstructed face images: the first probability P1=D1(Y1)=D1(G1(X)) and the second probability P2=D2(Y2). Furthermore, the reference unobstructed face image Y2 and the generated unobstructed face image Y1 are simultaneously input into the preset adversarial network D1 to judge whether D1 can distinguish and classify Y1 and Y2. If the distinction cannot be made, stop training D1 and use D1 as a face occlusion elimination model; if the distinction can be made, continue to train G1 and D1 using the training image set until D1 cannot distinguish and classify Y1 and Y2.

[0093] D3. Update network parameters of the preset generative network and the preset adversarial network according to the first probability and the second probability.

[0094] After obtaining the first probability and the second probability, the network parameters of the preset generation network and the preset adversarial network can be updated according to the first probability and the second probability. Exemplarily, the network parameters W1 of the preset generation network and the network parameters W2 of the preset adversarial network can be updated using the following formula:

[0095] minw1V(W1, W2)=-log(D1(G1(X)));

[0096] maxw2V(W1, W2)=logD2(Y2)+log(1-D1(G1(X))).

[0097] D4. Repeat the above network parameter updating process. When the network parameters reach convergence, use the preset generated network as the face occlusion removal model.

[0098] Repeat the above D1-D3 process until the values ​​of network parameters W1 and W2 no longer change significantly, it is determined that the network parameters have reached convergence, and the preset generation network at this time is determined as the face occlusion removal model.

[0099] S250: updating the face image to be tested according to the unobstructed face image to be tested.

[0100] After determining the unobstructed face image to be tested, the unobstructed face image to be tested is used as the updated face image to be tested. At this time, there is no face occlusion in the face image to be tested.

[0101] S260, determining a frontal face image to be tested based on the face normalization model and the updated face image to be tested.

[0102] The face normalization model refers to a pre-trained network model for realizing the face normalization function. It should be noted that the positive face image to be tested at this time is an unobstructed positive face image.

[0103] S270, performing qualification identification on the high-altitude worker to be tested based on the target face image and the positive face image to be tested.

[0104] Figure 3 The following is a workflow diagram of a method for identifying the qualifications of high-altitude workers provided in the second embodiment of the present invention. Figure 3 As shown, first, the target work ticket is downloaded from the cloud server through the edge server, and then the face image of the high-altitude worker to be tested is obtained through the drone, and then the face image to be tested is input into the face occlusion detection model through the edge server to determine whether there is face occlusion in the face image to be tested. If there is occlusion, the face image to be tested is input into the face occlusion elimination model to obtain the unobstructed face image to be tested, and the unobstructed face image to be tested is input into the face correction model to obtain the positive face image to be tested; if there is no occlusion, the face image to be tested is directly input into the face correction model to obtain the positive face image to be tested. Then, the positive face image to be tested and the target face image are input into the face recognition model together to obtain the face matching result, and finally the qualification recognition result of the high-altitude worker to be tested is determined according to the face matching result.

[0105] Figure 4 The following is a flowchart of another method for identifying the qualifications of high-altitude workers provided in the second embodiment of the present invention. Figure 4As shown, when the drone arrives above the transmission line, it shoots a video of the high-altitude workers to be tested, and performs face detection on the captured video. If a face is detected, face extraction is performed; if no face is detected, video shooting and face extraction need to be performed again. It is determined whether the extracted face image to be tested is blocked. If there is an occlusion, the occlusion is removed, and the face image to be tested is corrected after the removal; if there is no occlusion, the face image to be tested is directly corrected. Then, facial feature matching is performed based on the corrected face image to be tested and the target face image. If the match is successful, it is determined that the qualification verification of the high-altitude workers to be tested has passed; if the match fails (i.e., it does not match), it is determined that the qualification verification of the high-altitude workers to be tested has not passed.

[0106] The technical solution of the embodiment of the present invention, in response to a work ticket download request, first determines the target work ticket, the target work ticket includes the target face image; then determines the face image to be tested based on the drone, the face image to be tested includes the face of the high-altitude worker to be tested; then determines whether there is an occlusion on the face in the face image to be tested; if so, determines the unobstructed face image to be tested based on the face occlusion elimination model and the face image to be tested, and updates the face image to be tested according to the unobstructed face image to be tested; wherein the face occlusion elimination model refers to a pre-trained network model for realizing the face occlusion elimination function; then determines the frontal face image to be tested based on the face rotation model and the updated face image to be tested; wherein the face rotation model refers to a pre-trained network model for realizing the face rotation function; finally, performs qualification identification on the high-altitude worker to be tested according to the target face image and the frontal face image to be tested. This technical solution can automatically capture the facial images of high-altitude workers through drones, without the need for active participation and cooperation of high-altitude workers, thus improving the efficiency of qualification review. The facial occlusion removal model and the facial rotation model are used to remove occlusions and rotate the faces of the people to be tested, respectively, so as to identify the qualifications of the people based on the unobstructed forward facial images. This can effectively avoid facial recognition failures or misidentifications caused by facial angle deviation and facial occlusion, further improving the accuracy of personnel qualification review, while enhancing the convenience and practicality of qualification review, helping to improve the safety of transmission line construction sites, and reducing safety risks caused by personnel qualification issues, thereby providing strong guarantees for the sustainable development of the power industry. In addition, this technical solution can also be applied to other scenarios that require non-cooperative qualification recognition, such as security monitoring of public places such as large-scale events and transportation hubs.

[0107] Embodiment 3

[0108] Figure 5This is a structural diagram of a system for identifying qualifications of high-altitude workers provided in Embodiment 3 of the present invention. The system can execute the method for identifying qualifications of high-altitude workers provided in any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

[0109] like Figure 5 As shown, the system includes a drone 310, a cloud server 320 and an edge server 330, wherein:

[0110] The drone 310 is used to determine a face image to be tested and send the face image to be tested to the edge server 330, wherein the face image to be tested includes the face of the high-altitude worker to be tested;

[0111] The cloud server 320 is used to respond to the job ticket download request of the edge server 330, determine the target job ticket, and send the target job ticket to the edge server 330, wherein the target job ticket includes a target face image;

[0112] The edge server 330 is used to send the work ticket download request to the cloud server 320, and receive the target work ticket sent by the cloud server 320, and determine the frontal face image to be tested based on the face rotation model and the face image to be tested, and perform qualification identification on the high-altitude worker to be tested according to the target face image and the frontal face image to be tested; wherein the face rotation model refers to a pre-trained network model for realizing the face rotation function.

[0113] Optionally, the drone 310 is also used for:

[0114] Collect video stream data of the high-altitude workers to be tested;

[0115] Performing frame processing on the video stream data to obtain a frame image to be tested;

[0116] Perform face detection on the framed image to be tested to obtain a face image to be tested corresponding to the high-altitude worker to be tested.

[0117] Optionally, the training process of the face rotation model includes:

[0118] Acquire a face data set, and add label information to each reference face image in the face data set; wherein the label information includes identity information and posture information;

[0119] Inputting each of the reference face images into a generator of a distribution regularized generative adversarial network to obtain a plurality of generated face images; wherein the generated face images have a positive facial posture;

[0120] Inputting the generated face image and the corresponding reference face image into the discriminator of the distribution regularized generative adversarial network for image discrimination, and determining the generator loss value and the discriminator loss value during the discrimination process;

[0121] The generator when the generator loss value and the discriminator loss value reach convergence is used as the face normalization model.

[0122] Optionally, each of the reference face images is input into a generator of a distribution regularized generative adversarial network to obtain a plurality of generated face images, including:

[0123] Input each of the reference facial images into the encoder of the generator to obtain reference facial features corresponding to each reference facial image;

[0124] The reference facial features are integrated with preset noise information and preset facial posture information through the decoder of the generator, and a generated facial image is obtained after deconvolution processing.

[0125] Optionally, the edge server 330 is further configured to:

[0126] Performing facial feature matching between the forward face image to be tested and the target face image;

[0127] The qualification recognition result of the high-altitude worker to be tested is determined based on the facial feature matching result.

[0128] Optionally, the edge server 330 is further configured to:

[0129] Before determining the frontal face image to be tested based on the face rotation model and the face image to be tested, determining whether there is an obstruction on the face in the face image to be tested;

[0130] If so, determining the unobstructed face image to be tested based on the face occluder removal model and the face image to be tested; wherein the face occluder removal model refers to a pre-trained network model for realizing the face occluder removal function;

[0131] The face image to be tested is updated according to the unobstructed face image to be tested.

[0132] Optionally, the training process of the face occlusion removal model includes:

[0133] Inputting a reference occluded face image in the first training image set into a preset generation network to generate an unoccluded face image; wherein the preset generation network refers to an artificial neural network for converting the occluded face image into the unoccluded face image;

[0134] Inputting a reference unobstructed face image in the second training image set and the generated unobstructed face image into a preset adversarial network to obtain a first probability and a second probability; wherein the preset adversarial network refers to an artificial neural network for classifying and distinguishing the reference unobstructed face image and the generated unobstructed face image, the first probability refers to the probability of distinguishing the reference unobstructed face image as an unobstructed face image, and the second probability refers to the probability of distinguishing the generated unobstructed face image as an unobstructed face image;

[0135] Updating network parameters of the preset generation network and the preset adversarial network according to the first probability and the second probability;

[0136] The above network parameter updating process is repeated, and when the network parameters reach convergence, the preset generation network is used as the face occlusion removal model.

[0137] A system for identifying qualifications of high-altitude workers provided in an embodiment of the present invention can execute a method for identifying qualifications of high-altitude workers provided in any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

[0138] Embodiment 4

[0139] Figure 6 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0140] like Figure 6As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0141] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0142] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as the qualification identification method for aerial workers.

[0143] In some embodiments, the qualification identification method for high-altitude workers can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the qualification identification method for high-altitude workers described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to execute the qualification identification method for high-altitude workers in any other appropriate manner (for example, by means of firmware).

[0144] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0145] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0146] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0147] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0148] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0149] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0150] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0151] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for identifying the qualifications of personnel working at heights, characterized in that: The method comprises: In response to a job ticket download request, determining a target job ticket, wherein the target job ticket includes a target face image; Determine a face image to be tested based on the drone, wherein the face image to be tested includes the face of a high-altitude worker to be tested; Determining a positive face image to be tested based on a face rotation model and the face image to be tested; wherein the face rotation model refers to a pre-trained network model for realizing a face rotation function; The qualification of the high-altitude worker to be tested is identified based on the target facial image and the positive facial image to be tested.

2. The method according to claim 1, characterized in that: Determine the face image to be tested based on the drone, including: Collect video stream data of high-altitude workers to be tested through drones; Performing frame processing on the video stream data to obtain a frame image to be tested; Perform face detection on the framed image to be tested to obtain a face image to be tested corresponding to the high-altitude worker to be tested.

3. The method according to claim 1, characterized in that The training process of the face rotation model includes: Acquire a face data set, and add label information to each reference face image in the face data set; wherein the label information includes identity information and posture information; Inputting each of the reference face images into a generator of a distribution regularized generative adversarial network to obtain a plurality of generated face images; wherein the generated face images have a positive facial posture; Inputting the generated face image and the corresponding reference face image into the discriminator of the distribution regularized generative adversarial network for image discrimination, and determining the generator loss value and the discriminator loss value during the discrimination process; The generator when the generator loss value and the discriminator loss value reach convergence is used as the face normalization model.

4. The method according to claim 3, characterized in that Input each of the reference face images into the generator of the distribution regularized generative adversarial network to obtain multiple generated face images, including: Input each of the reference facial images into the encoder of the generator to obtain reference facial features corresponding to each reference facial image; The reference facial features are integrated with preset noise information and preset facial posture information through the decoder of the generator, and a generated facial image is obtained after deconvolution processing.

5. The method according to claim 1, characterized in that According to the target face image and the positive face image to be tested, the qualification of the high-altitude worker to be tested is identified, including: Performing facial feature matching between the forward face image to be tested and the target face image; The qualification recognition result of the high-altitude worker to be tested is determined based on the facial feature matching result.

6. The method according to any one of claims 1 to 5, characterized in that: Before determining the frontal face image to be tested based on the face normalization model and the face image to be tested, the method further includes: Determine whether there is an obstruction on the face of the person in the face image to be tested; If so, determining the unobstructed face image to be tested based on the face occluder removal model and the face image to be tested; wherein the face occluder removal model refers to a pre-trained network model for realizing the face occluder removal function; The face image to be tested is updated according to the unobstructed face image to be tested.

7. The method according to claim 6, characterized in that The training process of the face occlusion removal model includes: Inputting a reference occluded face image in the first training image set into a preset generation network to generate an unoccluded face image; wherein the preset generation network refers to an artificial neural network for converting the occluded face image into the unoccluded face image; Inputting a reference unobstructed face image in the second training image set and the generated unobstructed face image into a preset adversarial network to obtain a first probability and a second probability; wherein the preset adversarial network refers to an artificial neural network for classifying and distinguishing the reference unobstructed face image and the generated unobstructed face image, the first probability refers to the probability of distinguishing the reference unobstructed face image as an unobstructed face image, and the second probability refers to the probability of distinguishing the generated unobstructed face image as an unobstructed face image; Updating network parameters of the preset generative network and the preset adversarial network according to the first probability and the second probability; The above network parameter updating process is repeated, and when the network parameters reach convergence, the preset generation network is used as the face occlusion removal model.

8. A qualification identification system for high-altitude workers, characterized in that: The system includes a drone, a cloud server and an edge server, wherein: The drone is used to determine a face image to be tested, and send the face image to be tested to the edge server, wherein the face image to be tested includes the face of a high-altitude worker to be tested; The cloud server is used to respond to the job ticket download request of the edge server, determine the target job ticket, and send the target job ticket to the edge server, wherein the target job ticket includes a target face image; The edge server is used to send the work ticket download request to the cloud server, receive the target work ticket sent by the cloud server, and determine the frontal face image to be tested based on the face rotation model and the frontal face image to be tested, and identify the qualifications of the high-altitude worker to be tested according to the target face image and the frontal face image to be tested; wherein the face rotation model refers to a pre-trained network model for realizing the face rotation function.

9. The system according to claim 8, characterized in that The drone is also used to: Collect video stream data of the high-altitude workers to be tested; Performing frame processing on the video stream data to obtain a frame image to be tested; Perform face detection on the framed image to be tested to obtain a face image to be tested corresponding to the high-altitude worker to be tested.

10. The system according to any one of claims 8 or 9, characterized in that: The edge server is also used for: Before determining the frontal face image to be tested based on the face normalization model and the face image to be tested, determining whether there is an obstruction on the face in the face image to be tested; If so, determining the unobstructed face image to be tested based on the face occluder removal model and the face image to be tested; wherein the face occluder removal model refers to a pre-trained network model for realizing the face occluder removal function; The face image to be tested is updated according to the unobstructed face image to be tested.

Images