Three-dimensional face recognition method for underground coal mine

Abstract

The invention relates to a three-dimensional face recognition method for an underground coal mine, and belongs to the technical field of multi-dimensional face recognition. According to the method, the two-dimensional face image is converted into the three-dimensional points under the camera coordinate system, the rotation matrix is combined for rotation operation, the side face data of the rotating face at multiple angles are generated, and a rich data basis is provided for three-dimensional reconstruction. PointNet + + and an expansion structure thereof are combined with an auto-encoder frame to establish a three-dimensional reconstruction model, compact feature representation can be extracted from a three-dimensional point cloud, the features are utilized to reconstruct a three-dimensional face image with details and geometric structures, the three-dimensional face image is divided into a plurality of recognition areas, and feature recognition matching is performed respectively; during recognition, a face recognition result can be directly given only by screening clear areas for image reconstruction and feature matching. According to the invention, the accuracy and fidelity of reconstruction are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of multi-dimensional face recognition, and relates to a three-dimensional face recognition method used in coal mines. Background Art

[0002] In the field of coal mine safety monitoring, the integration and application of information technology has become an important means to improve the level of mine safety management and prevent accidents. With the increasing complexity and variability of the mine operating environment, traditional safety monitoring methods can no longer meet the needs of modern coal mines for efficient and accurate safety management. In recent years, biometric recognition technology, especially face recognition technology, has gradually emerged in coal mine safety monitoring due to its unique non-contact, high convenience and user-friendliness. However, the application of face recognition technology to the special environment of coal mines still faces a series of technical challenges and difficulties.

[0003] Existing face recognition technologies are mainly based on two-dimensional image processing and basic feature extraction methods, such as edge detection and feature point positioning. These technologies can achieve good recognition results under ideal conditions, such as sufficient lighting, correct angles, and natural expressions. However, the coal mine working environment is complex and changeable, with poor lighting conditions, high dust concentrations, and workers' rich and changeable facial expressions. These factors greatly affect the accuracy and robustness of face recognition.

[0004] First, in terms of lighting changes, insufficient or excessive lighting in coal mines will cause the quality of facial images to deteriorate, thus affecting recognition accuracy. Existing two-dimensional image processing technologies are often sensitive to lighting changes and it is difficult to maintain stable recognition performance under complex lighting conditions.

[0005] Secondly, existing face recognition technology is also unable to cope with different angles and facial expressions. When coal miners are working, their head posture and facial expressions will constantly change, which requires the recognition system to have a high degree of adaptability to angles and expressions. However, the current technology often has a significant drop in recognition accuracy when dealing with large angle deflections or complex expressions.

[0006] In addition, the occlusion problem is also a challenge that cannot be ignored in coal mine face recognition. Due to the special working environment, coal miners may wear protective equipment such as helmets and masks. These occlusions will seriously affect the integrity of the face image and the accuracy of feature extraction. When dealing with occlusion problems, existing technologies often cannot accurately identify the facial features of the occluded part, resulting in recognition failure.

[0007] With the continuous increase of coal mine safety monitoring data, existing face recognition algorithms are also facing severe challenges in terms of processing speed and real-time performance. In the context of big data, how to quickly and accurately process and analyze massive amounts of face image data has become an urgent problem to be solved in the field of coal mine safety monitoring. Summary of the invention

[0008] In view of this, the purpose of the present invention is to provide a three-dimensional face recognition method for use in coal mines to solve the following problems: (1) Existing two-dimensional image processing technology is often sensitive to changes in lighting and it is difficult to maintain stable recognition performance under complex lighting conditions; (2) When dealing with large-angle deflection or complex expressions, the recognition accuracy of current technology often drops significantly. (3) When dealing with occlusion problems, existing technologies are often unable to accurately identify the facial features of the occluded part, resulting in recognition failure. The method of the present invention can improve the accuracy and consistency of face recognition.

[0009] In order to achieve the above object, the present invention provides the following technical solutions:

[0010] A three-dimensional face recognition method for underground coal mines, specifically comprising the following steps:

[0011] S1: Use a high-resolution camera to collect diverse and wide-coverage original face images under different lighting and angles, and preprocess them, including filtering and denoising, grayscale normalization, etc., to output pure data images; then use the ICP registration method to perform feature point registration and fusion on the same face image collected under different conditions to obtain a second face image, thereby improving the accuracy of subsequent processing;

[0012] S2: Divide the second face image into a training set, a validation set and a test set in proportion, which are used for model learning, tuning and performance evaluation respectively; at the same time, generate feature point cloud data for establishing a three-dimensional facial image from the second face image, including rotated face profile data at multiple angles;

[0013] S3: Build a 3D reconstruction model, using PointNet++ and its extended structure combined with an autoencoder framework to extract features from 3D point clouds and reconstruct 3D facial images. Through training, the model learns to extract useful information from point clouds and generate realistic 3D facial images.

[0014] S4: The target faces of underground coal mine workers are identified through the trained 3D reconstruction model, and the 3D facial image is divided into several recognition areas, and feature recognition and matching are performed separately; in the information database, the facial information of each person is split and stored according to the recognition area, with identity information; during recognition, only clear areas need to be selected for image reconstruction and feature matching, and the face recognition result can be directly given.

[0015] Furthermore, in step S1, the original face image is preprocessed, specifically including: filtering and denoising, grayscale normalization, edge detection or size normalization.

[0016] Furthermore, in step S1, the ICP registration method is to perform geometric transformation on the same face images obtained at different times, different image blocks or different shooting angles so that the image points with the same name completely overlap in position and orientation; the specific calculation steps are:

[0017] 1) Initialization: Select a reference image I from N facial images of the same person taken at different times or angles. 0 As a benchmark, a feature point set P is extracted from the image. 0 ; For the other N-1 images I 1 ,I 2 ,…,I N-1 Extract the corresponding feature point set P respectively 1 ,P 2 ,…,P N-1 ;

[0018] 2) Registration process: For each pair of feature point sets P 0 and P i , perform the following steps:

[0019] 2.1) Find the nearest neighbor point: For P 0 Every point p in 0 , find P i The nearest point p i , can be measured using the Euclidean distance:

[0020] p i =arg min||p 0 -q||,q∈P i

[0021] 2.2) Calculate the transformation matrix: According to the nearest neighbor point pair found (p 0 ,p i ) Calculate the optimal rigid body transformation X i , rigid body transformation X i Usually contains the rotation matrix R i and the translation vector t i ; This can be solved using the least squares method:

[0022]

[0023] Where t is the translation vector; R is the rotation matrix. The calculation process of the rotation matrix is ​​to first center each point set, then calculate its covariance matrix, perform singular value SVD decomposition on the covariance matrix, and obtain the rotation matrix from the SVD decomposition;

[0024] 2.3) Apply the transformation: Apply the calculated transformation R i and t i Applied to feature point set Pi , and get a new set of points:

[0025] P i ′=R i ·P i +t i

[0026] 3) Iterative update: Repeat step 2) until the feature point set P i With P 0 The alignment error is less than a certain threshold or the maximum number of iterations is reached.

[0027] Through the above steps, the feature points in multiple images of the same person collected at different times and angles can be registered and fused, so that key parts such as eyes, nose, mouth, etc. are aligned in different images to obtain a second face image. This can improve the accuracy and consistency of subsequent processing steps.

[0028] Further, in step S2, the second face image is used to generate feature point cloud data for establishing a three-dimensional face image, specifically including: knowing that the pixel value of each pixel in the second face image is I(x, y), the initial depth value of each pixel is z(x, y), and a virtual camera is set based on the second face image, whose internal parameter matrix is ​​K and whose focal length relative to the second face image is f x ,f y , the center point of the image under the camera is c x ,c y , then each pixel (x, y) in the second face image is converted into a three-dimensional point (X c ,Y c ,Z c ):

[0029]

[0030] Z c =z(x,y)

[0031] Then, use the rotation matrix R to rotate each point; assume that the rotation matrix R contains three rotation angles θ x ,θ y ,θ z , then the rotation matrix can be expressed as:

[0032] R=R z (θ z )·R y (θ y )·R x (θ x )

[0033] Among them, R x ,Ry ,R z They are the rotation matrices around the x, y, and z axes, respectively, and can be expressed as:

[0034]

[0035] Through the rotation matrix R, each point in the camera coordinate system can be rotated:

[0036]

[0037] Among them, (X w ,Y w ,Z w ) is a point in the world coordinate system.

[0038] Specifically, step S2 generates rotated face profile data including 4 angles, from which face data with 30 different postures at 10°, 30° and 45° are selected as a test set, and the comparison error between the face data and the reference posture face data is calculated.

[0039] Furthermore, in step S3, the established three-dimensional reconstruction model specifically includes:

[0040] The encoder part is mainly composed of PointNet++, which is responsible for extracting features from the input 3D point cloud. This feature extraction process is divided into multiple levels, and each level captures the geometric structure information in the point cloud by aggregating local features. Specifically, each point interacts with its neighboring points to calculate the local feature representation of the point. These local features are aggregated at multiple levels to finally obtain the global feature representation of the entire point cloud.

[0041] The decoder is responsible for converting the feature representation extracted by the encoder into a 3D mesh model or voxel model; this conversion process is usually achieved through a series of deconvolution operations or upsampling operations, which gradually restore the feature representation to points or voxels in 3D space; during the decoding process, the model uses the features extracted by the encoder to generate a 3D facial image with details and geometric structure;

[0042] Combining the encoder and decoder results in an autoencoder framework, in which the model first extracts features from the input 3D point cloud through the encoder, and then restores these features to a 3D mesh model or voxel model through the decoder.

[0043] Furthermore, in step S3, training the three-dimensional reconstruction model specifically includes: by training the model, one can learn how to effectively extract useful information from the point cloud and use the information to generate realistic three-dimensional facial images.

[0044] The training process involves a large amount of 3D point cloud data and the corresponding 3D mesh model or voxel model. These data are usually obtained through 3D scanners or other acquisition devices. During the training process, the model will try to minimize the reconstruction error, that is, to make the generated 3D model as close to the real 3D model as possible; through continuous iteration and optimization, the model can gradually learn how to accurately reconstruct 3D facial images from point clouds.

[0045] Further, in step S4, the recognition area at least includes an eye recognition area, a nose recognition area, a mouth recognition area, an eye and nose recognition area, a mouth and nose recognition area, an ear and nose recognition area, and a face recognition area; the eye recognition area includes a left eye recognition sub-area and a right eye recognition sub-area, which respectively recognize and match the left and right eye features; the nose recognition area recognizes and matches the nose features; the mouth recognition area recognizes and matches the mouth features; the eye and nose recognition area refers to the facial area between the eyes and the nose, which are the left eye nose recognition area and the right eye nose recognition area, respectively, which mainly judge the obvious features of the facial area between the eyes and the nose and the nose The mouth and nose recognition area refers to the facial area between the mouth and the nose, which mainly judges the obvious features of the facial area between the nose and the mouth, the distance features between the nose and the mouth, and the lateral length features of the mouth relative to the nose; the ear and nose recognition area refers to the facial area between the ears and the nose, which includes the left ear and nose recognition sub-area and the right ear and nose recognition sub-area, and mainly judges the facial area between the ears and the nose and the distance features between the ears and the nose; the facial recognition area mainly recognizes facial contour features, including the left facial contour recognition sub-area, the right facial contour recognition sub-area, the upper facial contour recognition sub-area and the lower facial contour recognition sub-area.

[0046] The beneficial effects of the present invention are:

[0047] (1) The present invention improves the accuracy and consistency of face recognition: by using a high-resolution camera to collect face images under different lighting and angles, combined with pre-processing methods such as filtering and denoising, grayscale normalization, the diversity and high quality of image data are ensured, providing a reliable basis for subsequent processing. The ICP registration method is used to register and fuse the feature points in multiple images of the same person collected at different times and angles, so that the key parts are aligned in different images, significantly improving the accuracy and consistency of subsequent processing steps.

[0048] (2) Training and evaluation of the optimization model of the present invention:

[0049] Generate high-quality 3D facial images: By converting 2D facial images into 3D points in the camera coordinate system and performing rotation operations in combination with rotation matrices, rotated face profile data including multiple angles is generated, providing a rich data foundation for 3D reconstruction. Using PointNet++ and its extended structure combined with the autoencoder framework to establish a 3D reconstruction model, it is possible to extract compact feature representations from 3D point clouds and use these features to reconstruct 3D facial images with details and geometric structures, improving the accuracy and realism of reconstruction.

[0050] Adapt to the complex environment of coal mines: In view of the situation that coal mine workers may wear masks, helmets and other equipment or their faces may be blocked by stains, the three-dimensional facial image is divided into several recognition areas, and feature recognition and matching are performed separately, which effectively improves the recognition accuracy and robustness in complex environments. The corresponding identity information can be known by identifying the area separately, without splicing all area images, which simplifies the recognition process and improves the recognition efficiency. This solution can accurately and quickly identify coal mine workers, which helps to improve the safety management level and overall work efficiency of coal mining enterprises.

[0051] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0053] Figure 1 A flow chart of a three-dimensional face recognition method for use in coal mines provided by the present invention. DETAILED DESCRIPTION

[0054] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0055] See also Figure 1The present invention provides a three-dimensional face recognition method for coal mines, which specifically includes the following steps:

[0056] S1: Use a high-resolution camera to collect the first face image under different lighting and angles to ensure the diversity and coverage of the image; the image preprocessing method is filtering denoising, grayscale normalization, edge detection or size normalization; through image processing, a relatively pure data image is output to eliminate a large number of irrelevant areas, wherein the face matching model method is the ICP registration method, which is an operation of geometrically transforming the same face images obtained at different times, different image blocks, and different shooting angles so that the same-name image points are completely overlapped in position and orientation.

[0057] Specific calculation steps:

[0058] 1) Initialization: Select a reference image I from N facial images of the same person taken at different times and angles. 0 As a benchmark, a feature point set P is extracted from the image. 0 ; For the other N-1 images I 1 ,I 2 ,…,I N-1 Extract the corresponding feature point set P respectively 1 ,P 2 ,…,P N-1 .

[0059] 2) Registration process: For each pair of feature point sets P 0 and P i , perform the following steps:

[0060] 2.1) Find the nearest neighbor point: For P 0 Every point p in 0 , find P i The nearest point p i , can be measured using the Euclidean distance:

[0061] p i =arg min||p 0 -q||,q∈P i

[0062] 2.2) Calculate the transformation matrix: According to the nearest neighbor point pair found (p 0 ,p i ) Calculate the optimal rigid body transformation X i , rigid body transformation X i Usually contains the rotation matrix R i and the translation vector t i . This can be solved using the least squares method:

[0063]

[0064] Among them, R is the rotation matrix, t is the translation vector, and the calculation process of the rotation matrix is ​​to first center each point set, then calculate its covariance matrix, perform singular value SVD decomposition on the covariance matrix, and obtain the rotation matrix from the SVD decomposition.

[0065] 2.3) Apply the transformation: Apply the calculated transformation R i and t i Applied to feature point set P i , and get a new set of points:

[0066] P i ′=R i ·P i +t i

[0067] 3) Iterative update: Repeat step 2) until the feature point set P i With P 0 The alignment error is less than a certain threshold or the maximum number of iterations is reached.

[0068] Through the above steps, the feature points in multiple images of the same person collected at different times and angles can be registered and fused, so that key parts such as eyes, nose, mouth, etc. are aligned in different images to obtain a second face image. This can improve the accuracy and consistency of subsequent processing steps.

[0069] S2: Divide the preprocessed second face image into a training set, a validation set, and a test set according to a certain ratio. The training set is used for model learning, the validation set is used for model tuning, and the test set is used to evaluate model performance. In addition, feature point cloud data for establishing a three-dimensional facial image is generated based on the processed second face image, wherein the specific method is:

[0070] It is known that the pixel value of each pixel in the second face image is I(x, y), the initial depth value of each pixel is z(x, y), and a virtual camera is set based on the second face image, whose internal parameter matrix is ​​K and whose focal length relative to the second face image is f x ,f y , the center point of the image under the camera is c x ,c y , then each pixel (x, y) in the second face image is converted into a three-dimensional point (X c ,Y c ,Z c ):

[0071]

[0072] Z c=z(x,y)

[0073] Then, each point is rotated using the rotation matrix R. Assume that the rotation matrix R contains three rotation angles θ x ,θ y ,θ z , then the rotation matrix can be expressed as:

[0074] R=R z (θ z )·R y (θ y )·R x (θ x )

[0075] Among them, R x ,R y ,R z They are the rotation matrices around the x, y, and z axes, respectively, and can be expressed as:

[0076]

[0077] Through the rotation matrix R, each point in the camera coordinate system can be rotated:

[0078]

[0079] Among them, (X w ,Y w ,Z w ) is a point in the world coordinate system.

[0080] Specifically, step S2 generates rotated face profile data including 4 angles, from which face data with 30 different postures at 10°, 30° and 45° are selected as a test set, and the comparison error between the face data and the reference posture face data is calculated.

[0081] S3: Establish a 3D reconstruction model from the 3D point cloud (X w ,Y w ,Z w ) to extract features and build a 3D facial image. The 3D reconstruction model uses PointNet++ and its extended structure, combined with the Autoencoder framework. PointNet++ is a deep learning model for processing point cloud data. It extends the original PointNet model and enhances the ability to understand the local structure of the point cloud by introducing local feature aggregation. When combined with the Autoencoder, this model can extract compact feature representations from 3D point clouds and use these features to reconstruct a 3D mesh model or voxel model.

[0082] The encoder part is mainly composed of PointNet++, which is responsible for extracting features from the input 3D point cloud. This process is divided into multiple levels, and each level captures the geometric structure information in the point cloud by aggregating local features. Specifically, each point interacts with its neighboring points to calculate the local feature representation of the point. These local features are aggregated at multiple levels to finally obtain the global feature representation of the entire point cloud.

[0083] The decoder is responsible for converting the feature representation extracted by the encoder into a 3D mesh model or voxel model. This process is usually achieved through a series of deconvolution operations or upsampling operations, which gradually restore the feature representation to points or voxels in 3D space. During the decoding process, the model uses the features extracted by the encoder to generate a 3D facial image with details and geometric structure.

[0084] Combining the encoder and decoder together forms an autoencoder framework. In this framework, the model first extracts features from the input 3D point cloud through the encoder, and then restores these features to a 3D mesh model or voxel model through the decoder. Through training, the model can learn how to effectively extract useful information from the point cloud and use this information to generate realistic 3D facial images.

[0085] The training process involves a large amount of 3D point cloud data and the corresponding 3D mesh model or voxel model. These data are usually obtained through 3D scanners or other acquisition devices. During the training process, the model will try to minimize the reconstruction error, that is, to make the generated 3D model as close to the real 3D model as possible. Through continuous iteration and optimization, the model can gradually learn how to accurately reconstruct 3D facial images from point clouds.

[0086] S4: Use the trained 3D reconstruction model to identify the target face of the coal mine workers. Since coal mine workers usually wear masks, helmets, soundproof headphones and other equipment, or their faces are covered by stains due to environmental reasons, such as coal ash, etc., it may lead to unrecognition, resulting in misjudgment and affecting overall work efficiency.

[0087] Based on this, after obtaining the reconstructed three-dimensional facial image according to the above steps, the three-dimensional facial image is divided into several recognition areas, and the recognition areas at least include eye recognition area, nose recognition area, mouth recognition area, eye and nose recognition area, mouth and nose recognition area, ear and nose recognition area and face recognition area. The eye recognition area includes a left eye recognition sub-area and a right eye recognition sub-area, which respectively recognize and match the features of the left and right eyes; the nose recognition area recognizes and matches the features of the nose; the mouth recognition area recognizes and matches the features of the mouth; the eye and nose recognition area refers to the facial area between the eyes and the nose, which are the left eye nose recognition area and the right eye nose recognition area respectively, which mainly judge the obvious features of the facial area between the eyes and the nose and the distance features between the nose and the eyes; the mouth and nose recognition area refers to the facial area between the mouth and the nose, which mainly judges the obvious features of the facial area between the nose and the mouth, the distance features between the nose and the mouth, and the lateral length features of the mouth relative to the nose; the ear and nose recognition area refers to the facial area between the ears and the nose, which includes the left ear and nose recognition sub-area and the right ear and nose recognition sub-area, which mainly judges the facial area between the ears and the nose and the distance features between the ears and the nose; the face recognition area mainly recognizes facial contour features, including the left facial contour recognition sub-area, the right facial contour recognition sub-area, the upper facial contour recognition sub-area, and the lower facial contour recognition sub-area.

[0088] In the established information database, the pre-stored facial information of each person is split as above, and each part after the split carries corresponding identity information, that is, there is no need to splice the images of all areas, and the identity information corresponding to the area can be known only through a separate identification area.

[0089] Therefore, when performing facial recognition on staff members in mines, it is only necessary to screen out the clear area features of the person's face for image reconstruction, and then perform feature matching on the partial facial image of the reconstructed clear area with the corresponding recognition area. After a successful match, the facial recognition result is directly given based on the identity information in the corresponding recognition area.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A three-dimensional face recognition method for underground coal mines, characterized in that: The method specifically comprises the following steps: S1: Use a high-resolution camera to collect diverse and wide-coverage original face images under different lighting and angles, and preprocess them to output pure data images; then use the ICP registration method to perform feature point registration and fusion on the same face images collected under different conditions to obtain a second face image; S2: Divide the second face image into a training set, a validation set and a test set in proportion, which are used for model learning, tuning and performance evaluation respectively; at the same time, generate feature point cloud data for establishing a three-dimensional facial image from the second face image, including rotated face profile data at multiple angles; S3: Build a 3D reconstruction model, using PointNet++ and its extended structure combined with an autoencoder framework to extract features from 3D point clouds and reconstruct 3D facial images. Through training, the model learns to extract useful information from point clouds and generate realistic 3D facial images. S4: The target faces of underground coal mine workers are identified through the trained 3D reconstruction model, and the 3D facial image is divided into several recognition areas, and feature recognition and matching are performed separately; in the information database, the facial information of each person is split and stored according to the recognition area, with identity information; during recognition, only clear areas need to be selected for image reconstruction and feature matching, and the face recognition result can be directly given.

2. The three-dimensional face recognition method for underground coal mines according to claim 1, characterized in that: In step S1, the original face image is preprocessed, specifically including: filtering and denoising, grayscale normalization, edge detection or size normalization.

3. The three-dimensional face recognition method for underground coal mines according to claim 1, characterized in that: In step S1, the ICP registration method is to perform geometric transformation on the same face images obtained at different times, different image blocks or different shooting angles so that the image points with the same name completely overlap in position and orientation; the specific calculation steps are: 1) Initialization: Among N facial images of the same face taken at different times or angles, select a reference image I0 as the benchmark and extract the feature point set P0 from the image; for the other N-1 images I1, I2, …, I N-1 Extract the corresponding feature point sets P1, P2, ..., P respectively N-1 ; 2) Registration process: For each pair of feature point sets P0 and P i , perform the following steps: 2.1) Find the nearest neighbor: For each point p0 in P0, find P i The nearest point p i Use Euclidean distance to measure: p i =argmin||p0-q||,q∈P i 2.2) Calculate the transformation matrix: According to the nearest neighbor point pair found (p0, p i ) Calculate the optimal rigid body transformation X i , rigid body transformation X i Contains the rotation matrix R i and the translation vector t i ; Solved by the least squares method: Where t is the translation vector; R is the rotation matrix. The calculation process of the rotation matrix is ​​to first center each point set, then calculate its covariance matrix, perform singular value SVD decomposition on the covariance matrix, and obtain the rotation matrix from the SVD decomposition; 2.3) Apply the transformation: Apply the calculated transformation R i and t i Applied to feature point set P i , and get a new set of points: P i ′=R i ·P i +t i 3) Iterative update: Repeat step 2) until the feature point set P i The alignment error with P0 is less than a certain threshold or the maximum number of iterations is reached.

4. The three-dimensional face recognition method for underground coal mines according to claim 3, characterized in that: In step S2, the second face image is used to generate feature point cloud data for establishing a three-dimensional face image, specifically including: given that the pixel value of each pixel in the second face image is I(x, y), the initial depth value of each pixel is z(x, y), a virtual camera is set based on the second face image, its internal parameter matrix is ​​K, and its focal length relative to the second face image is f x ,f y , the center point of the image under the camera is c x ,c y , then each pixel (x, y) in the second face image is converted into a three-dimensional point (X c ,Y c ,Z c ): Z c =z(x,y) Then, use the rotation matrix R to rotate each point; assume that the rotation matrix R contains three rotation angles θ x ,θ y ,θ z , then the rotation matrix is ​​expressed as: R=R z (i z )·R y (i y )·R x (i x ) Among them, R x ,R y ,R z They are the rotation matrices around the x, y, and z axes, respectively, expressed as: By rotating the rotation matrix R, each point in the camera coordinate system is rotated: Among them, (X w ,Y w ,Z w ) is a point in the world coordinate system.

5. The three-dimensional face recognition method for underground coal mines according to claim 1, characterized in that: In step S3, the established three-dimensional reconstruction model specifically includes: The encoder part is composed of PointNet++, which is responsible for extracting features from the input 3D point cloud. This feature extraction process is divided into multiple levels, and each level captures the geometric structure information in the point cloud by aggregating local features. Specifically, each point interacts with its neighboring points to calculate the local feature representation of the point. These local features are aggregated at multiple levels to finally obtain the global feature representation of the entire point cloud. The decoder part is responsible for converting the feature representation extracted by the encoder into a 3D mesh model or voxel model; this conversion process is achieved through a series of deconvolution operations or upsampling operations, which gradually restore the feature representation to points or voxels in 3D space; during the decoding process, the model uses the features extracted by the encoder to generate a 3D facial image with details and geometric structure; Combining the encoder and decoder results in an autoencoder framework, in which the model first extracts features from the input 3D point cloud through the encoder, and then restores these features to a 3D mesh model or voxel model through the decoder.

6. The three-dimensional face recognition method for underground coal mines according to claim 1, characterized in that: In step S3, training the three-dimensional reconstruction model specifically includes: during the training process, the model will try to minimize the reconstruction error, that is, to make the generated three-dimensional model close to the real three-dimensional model; through continuous iteration and optimization, the model gradually learns to accurately reconstruct the three-dimensional facial image from the point cloud.

7. The three-dimensional face recognition method for underground coal mines according to claim 1, characterized in that: In step S4, the recognition area includes an eye recognition area, a nose recognition area, a mouth recognition area, an eye and nose recognition area, a mouth and nose recognition area, an ear and nose recognition area, and a face recognition area; the eye recognition area includes a left eye recognition sub-area and a right eye recognition sub-area, which respectively recognize and match the left and right eye features; the nose recognition area recognizes and matches the nose features; the mouth recognition area recognizes and matches the mouth features; the eye and nose recognition area refers to the facial area between the eyes and the nose, which are the left eye and nose recognition area and the right eye and nose recognition area, respectively, to judge the obvious features of the facial area between the eyes and the nose and the nose The mouth and nose recognition area refers to the facial area between the mouth and the nose, and determines the obvious features of the facial area between the nose and the mouth, as well as the distance features between the nose and the mouth, and the lateral length features of the mouth relative to the nose; the ear and nose recognition area refers to the facial area between the ears and the nose, which includes the left ear and nose recognition sub-area and the right ear and nose recognition sub-area, and determines the facial area between the ears and the nose and the distance features between the ears and the nose; the facial recognition area recognizes facial contour features, including the left facial contour recognition sub-area, the right facial contour recognition sub-area, the upper facial contour recognition sub-area, and the lower facial contour recognition sub-area.

Images