CN118644819B - Video monitoring management method and system for face care - Google Patents

Abstract

The present application relates to the field of face matching, and specifically to a video monitoring management method and system for face care, the method comprising: obtaining a patient face image and each template image; calculating the deviation factor between the patient face image and each template image, and screening each candidate template image; obtaining each layer of pyramid images of the patient face image and the candidate template image; obtaining each pyramid rotation template image through the rotation angle range and adaptive rotation step of the pyramid template image; obtaining the matching image of the pyramid search image according to the correlation between the texture feature matrix of the corresponding matching block of the pyramid search image and each pyramid rotation template image of the same layer, calculating the matching degree between the pyramid search image and the candidate image, and obtaining the final matching image of the patient face image through the matching degree. Thus, accurate matching and recognition of the patient's face information is achieved, and the accuracy of the patient's condition warning is improved.

Description

A video monitoring management method and system for face care

Technical Field

The present application relates to the field of face matching, and in particular to a video surveillance management method and system for face care.

Background Art

At present, intelligent medical care has entered people's daily lives, and human-computer interaction and intelligent robots have become hot technologies for medical assistance. With the intensification of the aging trend of the population, the medical care industry has ushered in rapid development. On the road of intelligent development of the medical care industry, it is increasingly important for hospitals to provide patients with a safe, caring, thoughtful and humane ward care environment while ensuring high standards of clinical medical care.

In the wards for critically ill patients, most of them need medical staff to watch over them all the time to prevent abnormal reactions or serious complications in the patients, and there is no one to watch over them in the ward. However, the number of medical staff is limited and relying on medical staff to watch over them in real time may lead to negligence, waste of manpower and material resources, and in serious cases, the patients may lose the best time for treatment. At present, some technologies detect dangerous movements of the patient's limbs to judge the patient's condition. The judgment basis of this process is not accurate enough, and it is not suitable for patients with severe illness or limbs that cannot move, and the scope of application is small.

In summary, the present application proposes a video surveillance management method and system for facial care, which captures facial videos of patients through video cameras, obtains multiple frames of continuous facial images, further matches the patient's face, and provides corresponding early warning prompts to medical staff. It has a wide range of applications so that medical staff can take targeted care and treatment measures.

Summary of the invention

In order to solve the above technical problems, in a first aspect, the present application provides a video monitoring management method for face care, the method comprising:

Acquire the patient's face image and each template image in the face care database, and obtain the deviation factor according to the difference between the centroid coordinates of the patient's face image and the centroid coordinates of the template image; use the template images with deviation factors lower than a preset threshold as candidate template images; and respectively obtain the pyramid images of each layer of the patient's face image and the candidate template image;

Each layer of the pyramid image of the patient's face image is recorded as a pyramid search image in turn, and each layer of the pyramid image of the candidate template image is recorded as a pyramid template image in turn; the rotation angle range of the pyramid template image is obtained according to the difference in face directions between the pyramid search image and the pyramid template image;

A multi-scale Gabor filter is used to filter the patient's face image to obtain multiple patient face feature maps; the complexity of the patient's face image is obtained according to the grayscale distribution of each patient's face feature map, and the adaptive rotation step length of the pyramid template image is calculated;

According to the rotation angle range and the adaptive rotation step size combined with the image rotation transformation algorithm, each pyramid rotation template image of the pyramid template image is obtained;

Each matching block in the pyramid search image is sequentially recorded as a pyramid search matching block, and each matching block in the pyramid rotation template image is sequentially recorded as a pyramid rotation template matching block;

The texture feature matrix of the pyramid search matching block is obtained according to the eigenvalues and eigenvectors of the Hessian matrix of all pixels in the pyramid search matching block, and accordingly, the texture feature matrix of the pyramid rotation template matching block is obtained; the matching image of the pyramid search image is obtained according to the correlation between the texture feature matrices of the corresponding matching blocks in the pyramid rotation template images of the same layer;

The mean of the matching similarity between all layers of pyramid search images and the corresponding matching images is analyzed to obtain the matching degree between the patient's face image and the candidate template image, and the final matching image of the patient's face image is obtained, and an early warning prompt is given according to the matching result.

In some embodiments, the deviation factor corresponds to the calculation formula: ; In the formula, is the deviation factor between the patient's face image and the template image k, is an exponential function with the natural constant e as base, To take the absolute value sign, are the x-coordinate difference and y-coordinate difference of the centroid coordinates of the patient's face image and the template image k, respectively.

In some embodiments, the process of obtaining the rotation angle range of the pyramid template image is as follows:

Obtain the angle between the line connecting the center points of the two eye matching blocks in the pyramid search image and the x-axis of the image coordinate system, which is recorded as the face direction of the pyramid search image. Accordingly, obtain the face direction of the pyramid template image.

The rotation angle range of the pyramid template image is calculated by the difference between the face directions of the pyramid search image and the pyramid template image, wherein the rotation angle range of the pyramid template image is positively correlated with the difference between the face directions of the pyramid search image and the pyramid template image.

In some embodiments, the process of obtaining the complexity of the patient's face image is:

Obtain the gray-level co-occurrence matrix corresponding to each patient's facial feature map, and calculate the complexity of the patient's facial feature map according to the feature quantity of the gray-level co-occurrence matrix, wherein the complexity of the patient's facial feature map is positively correlated with the entropy of the corresponding gray-level co-occurrence matrix, and is inversely correlated with the energy and inverse variance of the corresponding gray-level co-occurrence matrix;

The complexity mean of all patients' facial feature maps is taken as the complexity of the patient's facial image.

In some embodiments, the adaptive rotation step size of the pyramid template image corresponds to the calculation formula: ; In the formula, is the adaptive rotation step size of the pyramid template image, is the initial rotation step size, To round down, is the rotation step length control factor, is the complexity of the patient's face image.

In some embodiments, the step of obtaining the texture feature matrix of the pyramid search matching block according to the eigenvalues and eigenvectors of the Hessian matrix of all pixels in the pyramid search matching block comprises the following specific steps:

Calculate the Hessian matrix of each pixel in the pyramid search matching block to obtain the eigenvalues and eigenvectors of the Hessian matrix of each pixel, use the eigenvalue with a large absolute value as the texture change index corresponding to each pixel, and obtain the direction of the eigenvector corresponding to the eigenvalue with a large absolute value as the texture direction index of each pixel, and combine the texture change index and the texture direction index to form a texture feature binary of each pixel;

A texture feature matrix of the pyramid search matching block is constructed according to texture features of all pixels in the pyramid search matching block, wherein the texture feature binary of each pixel in the pyramid search matching block is used as each row of the texture feature matrix.

In some embodiments, the process of obtaining the matching image of the pyramid search image is:

Each pyramid rotation template image in the same layer as the pyramid search image is recorded as the same layer pyramid rotation template image in turn; the correlation coefficient between the texture feature matrix of each matching block in the pyramid search image and the texture feature matrix of each corresponding matching block in the same layer pyramid rotation template image is calculated, and the sum of all correlation coefficients is taken as the matching similarity between the pyramid search image and the same layer pyramid rotation template image;

A matching image of the pyramid search image is determined based on the matching similarity.

In some embodiments, the matching image of the pyramid search image is a pyramid rotated template image of the same level with the highest matching similarity.

In some embodiments, the process of obtaining the final matching image of the patient's facial image is: taking the candidate template image with the highest matching degree as the final matching image of the patient's facial image.

In the second aspect, an embodiment of the present application also provides a video surveillance management system for facial care, the system comprising a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the steps of any one of the above methods when executing the computer program.

This application has at least the following beneficial effects:

This application can monitor the patient's condition by matching the patient's facial information with the facial information in the facial care database. It has a wide range of applications and is particularly applicable to patients with severe illnesses and those who cannot move their limbs, reducing the time and energy consumption of medical staff.

In order to reduce the system computing cost and improve the matching speed, the present application extracts each matching block of the face, solves the problem of large computational complexity and slow matching speed when the entire face area is involved in the matching process, and the matching process is easily affected by irrelevant areas, resulting in low matching accuracy; at the same time, a pyramid image is constructed, combined with the matching of multi-resolution images, to improve the matching accuracy of the patient's face image; considering that the patient's face angle in the collected patient's face image is arbitrary, in order to avoid directly matching and analyzing the patient's face information through each matching block in the collected patient's face image, resulting in low matching accuracy and increased matching difficulty, the present application will rotate the candidate template image, and adaptively set the rotation angle range and rotation step of the candidate template image according to the complexity of the patient's face image, to improve matching accuracy and reduce the system's computing complexity.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

FIG1 is a flow chart of a video monitoring management method for face care provided by the present application;

FIG2 is a schematic diagram of the specific process of video surveillance management for face care.

DETAILED DESCRIPTION

In order to further explain the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the following, in combination with the accompanying drawings and preferred embodiments, describes in detail a video surveillance management method and system for face care proposed in the present application, its specific implementation method, structure, features and effects. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

The following is a detailed description of a specific solution of a video surveillance management method and system for face care provided by the present application in conjunction with the accompanying drawings.

An embodiment of the present application provides a video surveillance management method and system for face care.

Specifically, the video surveillance management method and system for face care of this embodiment provide the following video surveillance management method for face care, please refer to FIG1, the method comprises the following steps:

Step S001, obtain a patient face video, extract a patient face image from the patient face video, and obtain a face care database.

First, in this embodiment, the patient's face video is collected by a video monitoring device to obtain the patient's face image so as to monitor the patient's face information in real time. In this embodiment, the video monitoring device is a monitoring camera, which is used to collect the patient's face in real time. The monitoring camera model, shooting range and viewing angle of the video monitoring device can be set by the implementer. In order to realize the analysis of the patient's face information, this embodiment will automatically practice the curvature changes of facial features such as cheeks, eyebrows, mouth corners, forehead, etc., and then use OpenGL rendering to obtain a basic face image rich in various basic information. The face information includes but is not limited to frowning, painful closing of eyes, loose and drooping cheek muscles, frowning and tilted eyebrows, etc. All basic face images are stored in the face care database as the benchmark data for patient face analysis, and the face images in the face care library are all used as template images. The number of template images contained in the face care database is set by the implementer. It should be noted that the establishment of a specific face care database can be realized by existing technology, and the specific process is not described in detail one by one in this embodiment.

At this point, the patient's face video can be obtained, the patient's face image can be obtained as the patient's face image for real-time monitoring of the patient's face information, and the face care database can be obtained for analysis and matching of the patient's face information.

Step S002, establish a patient face image matching model, obtain a final matching image that matches the patient's face information, and implement analysis of the patient's face information.

For the patient's face image, this embodiment will process and analyze the patient's face image, screen the patient's face image matching blocks, and in order to improve the matching accuracy, during the matching process, rotate each template image so that the face in the template image is consistent with the face in the patient's face image. The matching speed of the matching block is improved while ensuring the matching accuracy, and the face image processing matching model is established:

In order to reduce the amount of calculation in the system matching process, this embodiment will perform a preliminary screening of the template images in the face care database, adaptively obtain candidate template images for matching the corresponding patient facial information, reduce the computational cost in the patient facial information matching and identification process, and improve the matching speed.

First, the centroid coordinates of the collected patient face image are calculated, and the centroid coordinates of each template image are obtained. The deviation factor between the patient face image and each template image is calculated by the difference between the centroid coordinates of the patient face image and the centroid coordinates of each template image. The expression is: ; In the formula, is the deviation factor between the patient's face image and the template image k, is an exponential function with the natural constant e as base, To take the absolute value sign, are the x-coordinate difference and y-coordinate difference of the centroid coordinates of the patient's face image and the template image k, respectively.

It can be understood that the larger the deviation factor between the patient's facial image and the template image, the greater the difference between the facial information in the patient's facial image and the facial information in the template image.

Furthermore, a deviation factor threshold is set to screen out template images whose deviation factors with the patient's facial image are higher than the threshold, and retain template images whose deviation factors with the patient's facial image are lower than the threshold as candidate template images for matching the patient's facial image. It should be noted that the deviation factor threshold can be set by the implementer, and in this embodiment, it is set to 0.5.

Then, this embodiment will match the patient's face image with each candidate template image to obtain the final matching image corresponding to the patient's face image. Considering that it is more in line with the visual characteristics of the human eye observing an object, first roughly obtaining the outline of the object and then looking at the details of the object, and considering that matching and analyzing the patient's face information through multi-scale patient face images can improve the accuracy of patient face matching, this embodiment will obtain each layer of the patient's face image and each layer of the candidate template image through the image pyramid algorithm, and perform multi-scale downsampling processing on the image to obtain patient face pyramid images of different resolutions and candidate template images of different resolutions. The implementer can set the number of pyramid layers by himself. In this embodiment, the number of pyramid layers is set to 3. There are many methods for downsampling when constructing a pyramid image, such as Gaussian sampling method, nearest neighbor method and mean sampling method. In this embodiment, the mean sampling method is used to construct the pyramid image. The image pyramid algorithm and the specific pyramid image construction process are existing known technologies.

Each layer of the pyramid image of the patient's face image is recorded as a pyramid search image, and each layer of the pyramid image of the candidate template image is recorded as a pyramid template image. The pyramid search image and the pyramid template image are matched with facial information. If all areas in the image are involved in the matching analysis of the patient's facial information, the system will have a large amount of calculation and a long matching time. At the same time, the matching process is easily affected by irrelevant areas. Therefore, this embodiment sets key facial areas. The key facial areas in this embodiment specifically include: forehead, eyebrows, eyes, nose, mouth and both cheeks. The 9 key facial areas are used as matching blocks to analyze the patient's facial expressions. For example, this embodiment will obtain a pyramid search image through the target detection network. The bounding boxes of each matching block in the target detection network, the training process of the target detection network, and obtaining the pyramid search image The process of generating the bounding boxes of each matching block in the target detection network is a well-known technique and will not be described in detail in this embodiment. , the output is the pyramid search image The bounding box information of each matching block, the mouth bounding box information is recorded as , The coordinate information of the center point of the mouth bounding box. are the width and height of the mouth bounding box respectively.

By repeating the above method, the bounding box information of each matching block in each pyramid search image and each pyramid template image can be obtained through the object detection network.

So far, each matching block in each pyramid search image and each pyramid template image can be obtained according to the above method.

Further, considering that when collecting patient face images, the angle of the patient face in the collected patient face images is arbitrary, if the patient face information is directly matched and analyzed through each matching block in each pyramid template image, the matching accuracy will be low and the matching difficulty will be increased. Therefore, when matching the matching blocks of the patient face, the traditional matching process mostly rotates the template image by setting a fixed rotation step to obtain rotated template images of different angles. However, the selection of the rotation step is mostly too random, which will affect the matching accuracy. At the same time, the rotation angle range is mostly [0°, 360°], and the rotation angle range is large, resulting in too many rotated template images, a large amount of calculation in the matching process, and reduced matching efficiency. Therefore, in this embodiment, the pyramid template image is adaptively rotated to obtain pyramid rotated template images of different angles to ensure that it is consistent with the patient face angle as much as possible when matching the patient face information, thereby improving the matching speed and matching accuracy. In this embodiment, the rotation angle range and rotation step of the pyramid template image are adaptively set, and the specific process is as follows:

The rotation angle range of the pyramid template image is obtained according to the difference in face directions between the pyramid search image and the pyramid template image. First, the angle between the line connecting the center points of two eye matching blocks in the pyramid search image and the x-axis of the image coordinate system is obtained, which is recorded as the face direction of the pyramid search image; the angle between the line connecting the center points of two eye matching blocks in the pyramid template image and the x-axis of the image coordinate system is obtained, which is recorded as the face direction of the pyramid template image; the rotation angle range of the pyramid template image is calculated by the difference between the face direction of the pyramid search image and the face direction of the pyramid template image, wherein the rotation angle range of the pyramid template image is positively correlated with the difference between the face directions of the pyramid search image and the pyramid template image.

Preferably, as an embodiment of the present application, the expression of the rotation angle range can be: ; In the formula, is the rotation angle range of the pyramid template image, Search the face orientation of the pyramid image S, is the face orientation of the pyramid template image T, is an angle bias term greater than zero, which is used to appropriately expand the rotation angle to prevent the small rotation angle range from causing the inability to accurately obtain a template image that is consistent with the facial angle in the patient's facial image. The value is set by the implementer. In the embodiment of this application, it is set to .

It should be noted that the rotation angle ranges of the pyramid template images at each layer of the same candidate template image are the same.

Repeat the above method to obtain the rotation angle range of each candidate template image, and thus obtain the rotation angle range of each pyramid template image corresponding to each candidate template image.

Further, for the rotation step of the pyramid template image, if the rotation step is too small, more rotated template images will be generated, which will reduce the overall matching speed; when the rotation step is too large, it may be impossible to obtain a template image that is consistent with the face angle in the patient's face pyramid image, resulting in reduced matching accuracy and mismatching. Therefore, this embodiment analyzes the complexity of the patient's face image. The simpler the patient's face image, the simpler the matching process, the higher the matching accuracy, and the rotation step of the pyramid template image can be appropriately increased. Conversely, in order to ensure matching accuracy, the rotation step should be reduced. The adaptive rotation step calculation process of the pyramid template image is specifically as follows:

For the patient's face image, this embodiment will analyze its complexity. First, this embodiment will filter the patient's face image through a multi-scale Gabor filter. In this embodiment, a multi-scale Gabor filter with 6 scales in 4 directions is used to obtain 24 Gabor filters. Each Gabor filter is convolved with the patient's face image to obtain the patient's face feature map after convolution. In this embodiment, 24 patient's face feature maps can be obtained. The patient's face feature map can characterize the features such as the texture details of the patient's face in the patient's face image. The scale and direction of the multi-scale Gabor filter can be set by the implementer. The specific filtering process of the multi-scale Gabor filter is an existing well-known technology and is not within the protection scope of this embodiment and will not be elaborated in detail.

The complexity of each patient's facial feature map is calculated. Considering that when the patient's facial information in the patient's facial feature map is more complex, the more uneven the texture in the corresponding patient's facial feature map is, the greater the complexity of the grayscale distribution is. Therefore, the grayscale co-occurrence matrix corresponding to each patient's facial feature map is obtained, and the feature value of the grayscale co-occurrence matrix is calculated. The complexity of the patient's facial feature map is calculated based on the feature value of the grayscale co-occurrence matrix, where the complexity of the patient's facial feature map is positively correlated with the entropy of the corresponding grayscale co-occurrence matrix, and inversely correlated with the energy and inverse variance of the corresponding grayscale co-occurrence matrix.

Preferably, as an embodiment of the present application, the complexity expression of the patient's facial feature map i can be: ; In the formula, is the complexity of the patient’s facial feature map i, is the entropy, energy, and inverse variance of the gray-level co-occurrence matrix corresponding to the patient's facial feature map i, To avoid parameters with a denominator of zero, the implementer can set it by himself. In this embodiment, it is set to The complexity mean of all patients’ facial feature maps is taken as the complexity of the patient’s facial image. ,The higher the complexity of the patient’s face image, the more complex the patient’s face information contained in the image;

According to the calculated complexity of the patient's face image, the rotation step of the pyramid template image is adaptively set. The adaptive rotation step expression of the pyramid template image is specifically: ; In the formula, is the adaptive rotation step size of the pyramid template image, To round down, is the initial rotation step length, which can be set by the implementer. In this embodiment, it is set to , The rotation step length control factor is used to control the rotation step length range. The implementer can set it by himself. This application sets it to .

In order to avoid the problem that the adaptive rotation step of the pyramid template image is too large, which leads to exceeding the rotation angle range and failing to accurately obtain the pyramid template image after rotation, this embodiment sets the maximum rotation step: ,in, is the maximum rotation step length, and p is a limiting coefficient, which is used to limit the maximum rotation step length. The implementer can set it by himself. In this embodiment, p is set to 3;

It should be noted that the adaptive rotation step sizes of the pyramid template images at each level of the same candidate template image are the same;

Repeat the above method to obtain the adaptive rotation step length of each candidate template image, and thus obtain the adaptive rotation step length of each pyramid template image corresponding to each candidate template image.

At this point, the rotation angle range and adaptive rotation step size of each pyramid template image can be obtained according to the above method.

According to the rotation angle range and the adaptive rotation step length combined with the image rotation transformation algorithm, each pyramid rotation template image of the pyramid template image is obtained. The pyramid template image is rotated and transformed according to the rotation angle range and the adaptive rotation step length by the image rotation transformation algorithm, and several pyramid rotation template images corresponding to the pyramid template image can be obtained. The several pyramid rotation template images of the pyramid template image are recorded as a pyramid rotation template image set of the pyramid template image. It should be noted that according to the rotation angle range and the adaptive rotation step length of the pyramid template image, through the image rotation transformation algorithm, each pyramid template image can obtain multiple pyramid rotation template images.

Repeat the above method to obtain a set of pyramid rotation template images for each pyramid template image.

Finally, according to the pyramid rotation template image set of each pyramid template image, the matching image of the pyramid search image is selected to match and identify the patient's facial information. Each matching block in the pyramid search image is recorded as a pyramid search matching block, and each matching block in the pyramid rotation template image is recorded as a pyramid rotation template matching block. For the pyramid search image, the texture feature matrix of each matching block is extracted in turn. The extraction process of the texture feature matrix is described in detail using the pyramid search image matching block as an example:

The Hessian matrix of each pixel in the pyramid search matching block is calculated to obtain the eigenvalues and eigenvectors of the Hessian matrix of each pixel, and then the eigenvalue with the largest absolute value of the Hessian matrix of each pixel is obtained as the texture change index of each pixel, and the direction of the eigenvector corresponding to the eigenvalue with the largest absolute value is obtained as the texture direction index of each pixel. The texture change index and texture direction index of each pixel in the pyramid search matching block are used as the texture feature binary of each pixel.

Then, the texture feature matrix of the pyramid search matching block is constructed according to the texture features of all pixels in the pyramid search matching block, wherein the texture feature binary of each pixel in the pyramid search matching block is used as each row of the texture feature matrix. The texture feature matrix expression of the pyramid search matching block is: ; In the formula, is the texture feature matrix of the pyramid search matching block j, S represents the pyramid search image, Search for the texture feature binary of the first pixel in the pyramid matching block j, are the texture change index and texture direction index of the first pixel in the pyramid search matching block j, respectively. is the texture feature binary of the nth pixel in the pyramid search matching block j, n is the number of pixels in the pyramid search matching block j, They are the texture change index and texture direction index of the nth pixel in the pyramid search matching block j, respectively.

The texture feature matrix of the matching block is mainly used to characterize the face texture detail information in the matching block. It should be noted that the calculation of the pixel Hessian matrix and the eigenvalue and eigenvector of the Hessian matrix is a known technology.

Repeat the above method to obtain the texture feature matrix of each matching block of the pyramid search image. Similarly, obtain the texture feature matrix of each matching block of each pyramid rotation template image in the pyramid rotation template image set.

In this embodiment, the matching image of the pyramid search image is selected to extract the final matching image of the patient's face image. The matching image of the pyramid search image is obtained according to the Pearson coefficient between the texture feature matrix of the corresponding matching block in each pyramid rotation template image of the same layer, specifically:

Each pyramid rotation template image in the same layer as the pyramid search image is recorded as the same-layer pyramid rotation template image in turn, and the correlation coefficient between the texture feature matrix of each matching block in the pyramid search image and the corresponding texture feature matrix of each matching block in the same-layer pyramid rotation template image is calculated; the sum of all correlation coefficients is taken as the matching similarity between the pyramid search image and the pyramid rotation template image in the same layer.

Preferably, as an embodiment of the present application, the correlation coefficient is analyzed using the Pearson correlation coefficient. It should be noted that the calculation method of the Pearson coefficient is an existing public technology; as other implementation methods, other correlation measurement methods of the existing technology can be used to calculate the correlation coefficient between matrices, and the present application does not impose any special restrictions.

Repeat the above method to calculate the matching similarity between the pyramid search image and each pyramid rotation template image at the same level; and use the pyramid rotation template image at the same level with the highest matching similarity as the matching image of the pyramid search image.

The above method is repeated to obtain matching images of each layer of pyramid search images, that is, to obtain matching similarities between each layer of pyramid search images and corresponding matching images.

The mean of the matching similarities between all layers of pyramid search images and the corresponding matching images is calculated, and the mean is used as the matching degree between the patient's face image and the corresponding candidate template image.

Repeat the above method to obtain the matching degree between the patient's face image and each candidate template image.

The candidate template image with the highest matching degree is used as the final matching image of the patient's face image, thereby realizing the recognition of the patient's face information.

Step S003, based on the matching results, corresponding early warning prompts are given to medical staff so that they can provide targeted nursing treatment to the patients.

According to the final matching image corresponding to the patient's facial image, the matching recognition of the patient's face is realized, and an early warning prompt is issued according to the result of the patient's facial matching, so that medical staff can accurately understand the patient's condition, take targeted nursing measures, and provide corresponding nursing treatment to the patient. It should be noted that the implementer can set the specific early warning prompt content corresponding to the patient's facial information. For example, when the facial expression corresponding to the final matching image of the patient's facial image is frowning or opening the mouth, the system will issue an early warning prompt of the patient's discomfort, so that medical staff can accurately understand the patient's condition.

Preferably, as an embodiment of the present application, a schematic diagram of the specific process of video surveillance management for face care is shown in FIG2 .

It is worth noting that the positive correlation described in this application is used to characterize that the changing trends between variables are the same, that is, the dependent variable increases (decreases) as the independent variable increases (decreases). The specific calculation relationship is determined by the implementer based on the actual application scenario, and can be an additive relationship, multiplicative relationship or other positive correlation relationship. This application does not impose any special restrictions.

Based on the same inventive concept as the above method, an embodiment of the present application also provides a video surveillance management system for facial care, including a memory, a processor, and a computer program stored in the memory and running on the processor, and when the processor executes the computer program, it implements the steps of any one of the above-mentioned video surveillance management methods for facial care.

To summarize, the embodiment of the present application can monitor the patient's condition by matching the patient's facial information with the facial information in the facial care database. It has a wide range of applications and is particularly suitable for patients with severe illnesses and those who cannot move their limbs, thereby reducing the consumption of time and energy of medical staff.

In order to reduce the system computing cost and improve the matching speed, the present application extracts each matching block of the face, solves the problem of large computational complexity and slow matching speed when the entire face area is involved in the matching process, and the matching process is easily affected by irrelevant areas, resulting in low matching accuracy; at the same time, a pyramid image is constructed, combined with the matching of multi-resolution images, to improve the matching accuracy of the patient's face image; considering that the patient's face angle in the collected patient's face image is arbitrary, in order to avoid directly matching and analyzing the patient's face information through each matching block in the collected patient's face image, resulting in low matching accuracy and increased matching difficulty, the present application will rotate the candidate template image, and adaptively set the rotation angle range and rotation step of the candidate template image according to the complexity of the patient's face image, to improve matching accuracy and reduce the system's computing complexity.

It should be noted that the above sequence of the embodiments of the present application is for description only and does not represent the advantages and disadvantages of the embodiments. The above is a description of a specific embodiment of this specification. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present application should be included in the protection scope of the present application.

Claims (10)

1. The video monitoring management method for face care is characterized by comprising the following steps:

Obtaining a face image of a patient and each template image in a face nursing database, and obtaining a deviation factor according to the difference between the centroid coordinates of the face image of the patient and the centroid coordinates of the template image; taking template images with deviation factors lower than a preset threshold value as candidate template images; respectively acquiring pyramid images of each layer of face images of a patient and candidate template images;

Each layer of pyramid images of the face image of the patient are sequentially marked as pyramid search images, and each layer of pyramid images of the candidate template images are sequentially marked as pyramid template images; obtaining a rotation angle range of the pyramid template image according to the face direction difference of the pyramid search image and the pyramid template image;

Filtering the face images of the patients by adopting a multi-scale Gabor filter to obtain a plurality of face feature images of the patients; obtaining the complexity of the face image of each patient according to the gray distribution condition of the face feature image of each patient, and calculating the self-adaptive rotation step length of the pyramid template image;

obtaining pyramid rotation template images of the pyramid template images according to the rotation angle range and the self-adaptive rotation step length in combination with an image rotation transformation algorithm;

each matching block in the pyramid search image is sequentially marked as a pyramid search matching block, and each matching block in the pyramid rotation template image is sequentially marked as a pyramid rotation template matching block;

obtaining a texture feature matrix of the pyramid search matching block according to the feature values and feature vectors of all pixel point hessian matrixes in the pyramid search matching block, and correspondingly obtaining the texture feature matrix of the pyramid rotation template matching block; obtaining a matched image of the pyramid search image according to the correlation between the pyramid search image and the texture feature matrix of the corresponding matched block in each pyramid rotation template image of the same layer;

And analyzing the average value of the matching similarity between the pyramid search images of all layers and the corresponding matching images to obtain the matching degree of the face image of the patient and the candidate template image, obtaining the final matching image of the face image of the patient, and carrying out early warning prompt according to the matching result.

2. The video monitoring management method for face care according to claim 1, wherein the deviation factor corresponds to a calculation formula: ; in the method, in the process of the invention, As a deviation factor between the patient face image and the template image k,As an exponential function based on a natural constant e,In order to take the sign of the absolute value,The difference value of the x coordinate and the difference value of the y coordinate of the centroid coordinates of the face image of the patient and the template image k are respectively.

3. The video monitoring management method for face care according to claim 1, wherein the acquiring process of the rotation angle range of the pyramid template image is as follows:

Acquiring the angle between the connecting line of the center points of the two eye matching blocks in the pyramid search image and the x axis of the image coordinate system, marking the angle as the face direction of the pyramid search image, and correspondingly, acquiring the face direction of the pyramid template image;

and calculating a rotation angle range of the pyramid template image through the difference between the face direction of the pyramid search image and the face direction of the pyramid template image, wherein the rotation angle range of the pyramid template image is positively associated with the difference between the face directions of the pyramid search image and the pyramid template image.

4. The video monitoring management method for face care according to claim 1, wherein the obtaining process of the complexity of the face image of the patient is:

Acquiring a gray level co-occurrence matrix corresponding to each patient face feature map, and calculating the complexity of the patient face feature map according to the feature amount of the gray level co-occurrence matrix, wherein the complexity of the patient face feature map is in forward association with the entropy of the corresponding gray level co-occurrence matrix, and in reverse association with the energy and inverse variance of the corresponding gray level co-occurrence matrix;

and taking the complexity average value of all the face feature images of the patient as the complexity of the face images of the patient.

5. The video monitoring management method for face care according to claim 1, wherein the adaptive rotation step length of the pyramid template image corresponds to a calculation formula: ; in the method, in the process of the invention, For the adaptive rotation step of the pyramid template image,For the initial rotation step size,In order to perform the rounding-down operation,In order to rotate the step size regulatory factor,Complexity of face images for patients.

6. The video monitoring management method for face care according to claim 1, wherein the obtaining the texture feature matrix of the pyramid search matching block according to the feature values and feature vectors of all pixel point hessian matrices in the pyramid search matching block comprises the following specific steps:

Calculating a hessian matrix of each pixel point in a pyramid search matching block to obtain characteristic values and characteristic vectors of the hessian matrix of each pixel point, taking the characteristic value with a large absolute value as a texture change index corresponding to each pixel point, obtaining the direction of the characteristic vector corresponding to the characteristic value with the large absolute value as a texture trend index of each pixel point, and forming texture feature binary groups of each pixel point by the texture change index and the texture trend index;

And constructing a texture feature matrix of the pyramid search matching block according to the texture features of all the pixel points in the pyramid search matching block, wherein texture feature doublets of all the pixel points in the pyramid search matching block are used as all rows of the texture feature matrix.

7. The video monitoring management method for face care according to claim 1, wherein the acquiring process of the matching image of the pyramid search image is as follows:

Sequentially marking each pyramid rotation template image on the same layer as the pyramid search image as a pyramid rotation template image on the same layer; calculating correlation coefficients between texture feature matrixes of all matching blocks in the pyramid search image and texture feature matrixes of corresponding matching blocks in the same-layer pyramid rotation template image, and taking the sum of all the correlation coefficients as matching similarity between the pyramid search image and the same-layer pyramid rotation template image;

and determining a matching image of the pyramid search image based on the matching similarity.

8. The video monitoring management method for face care according to claim 1, wherein the matching image of the pyramid search image is a same-layer pyramid rotation template image with highest matching similarity.

9. The video monitoring management method for face care according to claim 1, wherein the final matching image acquisition process of the face image of the patient is: and taking the candidate template image with the highest matching degree as a final matching image of the face image of the patient.

10. A video surveillance management system for face care comprising a memory, a processor and a computer program stored in the memory and running on the processor, characterized in that the processor implements the steps of the method according to any of claims 1-9 when executing the computer program.