Personnel facial form recognition method, electronic equipment and storage medium

Abstract

The invention discloses a person facial form recognition method, electronic equipment and a storage medium. The method belongs to the field of intelligent identification, and comprises the following steps: obtaining environment information of an environment in which a target person is located in a historical time period, the environment information at least comprising occurrence times of different environments; based on the environment information, determining a facial form detection environment of the target person, the occurrence frequency of the facial form detection environment in the environment information being greater than or equal to a preset frequency; in response to the fact that the environment type of the face shape detection environment is a preset type, a face image, collected by an image collection device, of the target person is obtained, the image collection device is deployed in the face shape detection environment, and the frequency of using the image collection device by the target person is larger than or equal to the preset frequency; and recognizing the face image based on the face recognition model to obtain a face recognition result of the target person. According to the invention, the technical problem of low accuracy of human face recognition in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent recognition, and in particular to a method for recognizing human face shapes, an electronic device and a storage medium. Background Art

[0002] In the mining industry, the safety and health of operators are always the primary focus. The mine environment is harsh, and pollutants such as dust and harmful gases suspended in the air pose a serious threat to the respiratory system of operators. In order to ensure the life safety and physical health of operators, it is necessary to use scientific methods to determine the facial features of operators in order to select appropriate self-priming filter respiratory protective equipment to improve the protection effect.

[0003] At present, the face shape of mine workers is mainly determined by manual measurement and classification against standards. This process is not only time-consuming, but also has large errors, especially in the mine environment with insufficient light and narrow space, the measurement accuracy is difficult to guarantee, thus affecting the selection and use of respiratory protection equipment.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present invention provide a method for recognizing a person's face shape, an electronic device, and a storage medium, so as to at least solve the technical problem of low accuracy in recognizing a person's face shape in the related art.

[0006] According to one aspect of an embodiment of the present invention, a method for face recognition of a person is provided, comprising: obtaining environmental information of an environment in which a target person is located during a historical time period, wherein the environmental information includes at least the number of times different environments appear; determining a face detection environment of the target person based on the environmental information, wherein the number of times the face detection environment appears in the environmental information is greater than or equal to a preset number; in response to an environmental type of the face detection environment being a preset type, obtaining a face image of the target person acquired by an image acquisition device, wherein the image acquisition device is deployed in the face detection environment, and a frequency of the target person using the image acquisition device is greater than or equal to a preset frequency; and recognizing the face image based on a face recognition model to obtain a face recognition result of the target person.

[0007] Furthermore, the method also includes: detecting the influencing parameters included in the face shape detection environment to obtain parameter influence indicators, wherein the influencing parameters are used to characterize the parameters that will affect the quality of the collected image when the image is collected in the face shape detection environment, and the parameter influence indicators are used to characterize the degree of influence of the influencing parameters on the collected image; in response to the parameter influence indicator being greater than or equal to a preset indicator, determining that the environment type of the face shape detection environment is a preset type.

[0008] Furthermore, the influencing parameters include at least: light intensity and dust concentration of the face detection environment. The influencing parameters contained in the face detection environment are detected to obtain parameter influence indicators, including: light intensity detection of the face detection environment to obtain a light intensity index; dust concentration detection of the face detection environment to obtain a dust concentration index; weighted fusion of the light intensity index and the dust concentration index to obtain a parameter influence index.

[0009] Furthermore, the face image is recognized based on the face shape recognition model to obtain the face shape recognition result of the target person, including: extracting features from the face image to obtain multiple face image features; constructing an initial face shape based on the multiple face image features; matching the initial face shape with multiple preset face shapes respectively to determine the target face shape of the target person from the multiple preset face shapes; marking the target face shape based on the person identification of the target person to obtain the face shape recognition result.

[0010] Furthermore, the initial face shape is matched with multiple preset face shapes respectively, and the target face shape of the target person is determined from the multiple preset face shapes, including: obtaining face shape matching values ​​corresponding to different preset face shapes, wherein the face shape matching values ​​are used to characterize the degree of matching between the initial face shape and the different preset face shapes; based on the face shape matching values, a first face shape is screened out from the multiple preset face shapes, wherein the face shape matching value corresponding to the first face shape is greater than or equal to the preset value; outputting the first face shape, and receiving a face shape selection result, wherein the face shape selection result includes a second face shape selected by the target person from the first face shape; and determining the second face shape as the target face shape.

[0011] Furthermore, the method also includes: obtaining multiple training images matching different preset face shapes, wherein the number of multiple training images corresponding to different preset face shapes is the same, and the number of multiple training images is greater than or equal to a preset number; training the initial recognition model based on multiple training images according to a preset number of training times, a preset learning rate and a preset number of samples to obtain a face shape recognition model, wherein the preset number of training times is greater than or equal to a first value, the preset learning rate is a second value, the preset learning rate decreases as the number of training times increases according to a preset value, the preset number of samples is used to characterize the number of samples for updating the model weights of the initial recognition model, and the preset number of samples is greater than or equal to a third value.

[0012] Furthermore, the face shape recognition model is a deep neural network model, the size of the convolution kernel included in the face shape recognition model is less than or equal to the fourth value, and the number of network structure layers of the face shape recognition model is greater than or equal to the fifth value.

[0013] According to another aspect of an embodiment of the present invention, a device for recognizing a person's face shape is also provided, comprising: a first acquisition module, used to acquire environmental information of the environment in which a target person is located during a historical time period, wherein the environmental information at least includes the number of times different environments appear; an environment determination module, used to determine a face detection environment of the target person based on the environmental information, wherein the number of times the face detection environment appears in the environmental information is greater than or equal to a preset number; a second acquisition module, used to acquire a face image of the target person acquired by an image acquisition device in response to the environment type of the face detection environment being a preset type, wherein the image acquisition device is deployed in the face detection environment, and the frequency with which the target person uses the image acquisition device is greater than or equal to a preset frequency; and a face recognition module, used to recognize the face image based on a face recognition model to obtain a face recognition result of the target person.

[0014] According to another aspect of an embodiment of the present invention, there is further provided an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein the method in each embodiment of the present invention is executed when the program is running.

[0015] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in various embodiments of the present invention.

[0016] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0017] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0018] According to another aspect of the embodiments of the present invention, a computer program is further provided. When the computer program is executed by a processor, the methods in the embodiments of the present invention are implemented.

[0019] In an embodiment of the present invention, the environmental information of the target person's environment in a historical time period is obtained; based on the environmental information, the face detection environment of the target person is determined; in response to the environmental type of the face detection environment being a preset type; based on the face recognition model, the face image is recognized to obtain the face recognition result of the target person, and the face detection environment is screened and matched with the specific environmental type of the above face detection environment to reduce the negative impact of complex environmental factors on the face detection of the person. Further, based on the face recognition model, the face images under the above different environmental types are intelligently determined, and the specific face categories under different environmental types can be determined more accurately, achieving the purpose of reducing the error of face determination in the related technology, thereby achieving the technical effect of improving the accuracy of face determination, and then solving the technical problem of low accuracy of face recognition of the related technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0021] Figure 1 is a flow chart of a method for face recognition according to an embodiment of the present invention;

[0022] Figure 2 is a detailed flow chart of a method for face recognition according to an embodiment of the present invention;

[0023] Figure 3 is a schematic diagram of a person face recognition device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] 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.

[0025] It should be noted that the terms "first", "second", 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 that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] According to an embodiment of the present invention, an embodiment of a method for facial recognition of a person is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0027] Figure 1 is a flow chart of a method for face recognition according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0028] Step S102, obtaining environmental information of the target person's environment in a historical time period, wherein the environmental information at least includes the number of times different environments appear.

[0029] The target person may refer to a person who needs to perform face recognition. The historical time period may refer to a time period before the current moment. The environmental information may refer to information that affects the face recognition of a person in different environments, and the environmental information may include at least the number of times different environments appear, but is not limited thereto.

[0030] In an optional embodiment, considering that the target person may appear in multiple different environments, if the face recognition of the target person is only performed in a specific environment, it may be difficult to apply the face recognition method to different environments, resulting in a decrease in the recognition accuracy of the face recognition method when the specific environment changes. However, if all possible scenes of the target person are considered, the face recognition efficiency may be reduced. Therefore, by collecting environmental information of the target person's environment in a historical time period, it is helpful to determine an environment with a high demand for face recognition, thereby improving the efficiency and accuracy of subsequent face recognition of personnel.

[0031] For example, in order to accurately obtain the above environmental information, the staff can deploy multiple environmental sensors in different environments in the mine in advance. The above environmental sensors can be light sensors, temperature and humidity sensors, and air quality sensors, but are not limited to these. The face detection system (hereinafter referred to as the detection system) can determine the environmental parameter information of the environment in which the above target person is located during the historical time period based on the above environmental sensors. In addition, the staff can also deploy multiple high-definition cameras in the above different environments in advance. The detection system can count the number of times the above target person appears in different environments based on the image data captured by the above multiple high-definition cameras.

[0032] Step S104, determining a face shape detection environment of the target person based on the environmental information, wherein the number of times the face shape detection environment appears in the environmental information is greater than or equal to a preset number of times.

[0033] The face shape detection environment may be an environment where the target person often appears, such as a work environment or a living environment, but is not limited thereto. The preset number of times may be pre-set and used to determine the value of the face shape detection environment.

[0034] In an optional embodiment, considering that when performing face shape detection on a target person, in order to ensure the accuracy of the detection result, the target person is usually required to go to a designated detection environment and use a high-precision detection device to perform face shape detection, but these designated detection environments may not be the target person's daily work, life, and other environments. When there are many people in the designated detection environment who need to perform face shape detection, a long waiting time may be generated, resulting in the target person needing to spend more time cost when performing face shape detection, affecting the target person's face shape detection efficiency. Therefore, in order to improve the efficiency of face shape detection on the target person, the environment where the target person often appears, that is, the above-mentioned face shape detection environment, can be used instead of the designated detection environment, so that the target person can conveniently perform face shape detection in environments such as daily work and life, without worrying about consuming a lot of time when performing face shape detection. Based on this, in order to accurately determine the detection environment that can conveniently perform face detection, after obtaining the environmental information of the target person's environment in the historical time period, the detection system can use the number of times different environments appear in the environmental information as a judgment standard, and compare the number of times different environments appear with the preset number of times on the basis of a preset judgment index, such as a preset number of times, so as to screen out the target environment where the target person often appears from the environment where the target person appears, as the face detection environment for face detection of the target person, so that the target person does not need to spend more time to perform face detection in the specified detection environment, thereby reducing the time cost of face detection and improving the efficiency of face detection. Among them, when screening the face detection environment from the environment where the user appears, the number of times different environments appear can be compared with the preset number of times. If the number of times the environment appears is greater than or equal to the preset number of times, it can be considered that the target person will often appear in the environment, and at this time, the environment can be determined as the above-mentioned face detection environment; if the number of times the environment appears is less than the preset number of times, it can be considered that the target person will occasionally appear in the environment, and at this time, the environment can not be determined as the above-mentioned face detection environment.

[0035] For example, the detection system has determined the number of times the above-mentioned different environments appear. At this time, in order to reduce the cost of face shape detection and improve the efficiency of face shape detection, the detection system can determine the environment whose number of appearances in the above-mentioned environmental information is greater than or equal to the above-mentioned preset number as the face shape detection environment of the above-mentioned target person, so that the target person can perform face shape detection in the face shape detection environment, and no longer needs to enter the designated area to detect the target person's face shape through the designated detection machine.

[0036] Step S106, in response to the environment type of the face shape detection environment being a preset type, obtaining a face image of the target person captured by the image acquisition device, wherein the image acquisition device is deployed in the face shape detection environment, and the target person uses the image acquisition device at a frequency greater than or equal to a preset frequency.

[0037] The above-mentioned preset type may be pre-set to indicate the type of environment where the face of a person cannot be recognized by conventional means, for example, it may be an environment with weak lighting or an environment with high dust concentration, but is not limited thereto. The above-mentioned image acquisition device may be a device set for a preset type of face detection environment, and is used to obtain clear facial images in an environment where the face of a person cannot be recognized by conventional means, for example, it may be a device such as a face clock-in system, but is not limited thereto. The above-mentioned preset frequency may be the minimum frequency set in advance for use or triggering of the above-mentioned image acquisition device in order to achieve the expected data collection goal or system performance indicator.

[0038] In an optional embodiment, considering the above-mentioned face detection environment, that is, the environment where the target person appears more frequently, it may be a specific environment where the face of the person cannot be identified by conventional means, for example, it may be an environment with weak light or high dust concentration. At this time, if the face of the above-mentioned target person is to be recognized, the detection system needs to obtain a clear face image of the target person with the help of the above-mentioned image acquisition device. Therefore, in order to accurately identify the face of the target person in the above-mentioned specific environment, the staff can pre-select the face detection device deployed in the above-mentioned specific environment and the target person's use frequency greater than or equal to the above-mentioned preset frequency as the image acquisition device. For example, the face punching system that the target person needs to use frequently can be selected as the above-mentioned image acquisition device. At the same time, the detection system can pre-classify the types of the above-mentioned specific environments into preset types, and then the detection system can determine the environmental type of the current face detection environment. If the detection system determines that the environmental type of the current face detection environment belongs to one of the above-mentioned preset types, the detection system can obtain a clear face image of the target person through the selected image acquisition device, thereby improving the accuracy of face recognition of the target person. If the detection system determines that the environmental type of the current face detection environment does not belong to any of the preset types, the detection system can directly obtain the face image of the target person through conventional face detection equipment, such as cameras, and then complete the face recognition of the target person.

[0039] For example, in order to accurately capture the facial image of the target person, the image acquisition device may be a camera, which may be pre-deployed in the face detection environment, and the frequency of the target person using the camera is greater than or equal to the preset frequency. Considering that the lighting in the mine may be relatively limited, the preset type may be an environment with poor auxiliary lighting. If the environment type of the face detection environment is the preset type, the detection system may obtain the facial image of the target person captured by the camera.

[0040] For example, the above-mentioned image acquisition device may be a face punching system. The above-mentioned face punching system may be pre-deployed in the above-mentioned face detection environment, and the frequency of the target person using the face punching system is greater than or equal to the above-mentioned preset frequency. Taking into account the particularity of the mine environment, the preset type may also need to consider the impact of the dust concentration in the air on face detection. Therefore, the above-mentioned preset type may be an environment with a high dust concentration in the air. In this environment, it is difficult for the detection system to directly obtain a high-definition face image of the target person through conventional face detection equipment such as a camera. If the environment type of the above-mentioned face detection environment is the above-mentioned preset type, the detection system can obtain the face image of the target person collected by the face punching system for face recognition of the target person, thereby improving the accuracy of face recognition.

[0041] For another example, if the current face shape detection environment does not belong to the above-mentioned preset face shape, the light in the environment is sufficient and the visibility is high, and the camera deployed in the environment can clearly capture the facial image of the target person. At this time, the detection system can directly obtain the facial image of the target person through the above-mentioned camera, thereby completing face shape recognition.

[0042] Step S108, recognizing the face image based on the face recognition model to obtain the face recognition result of the target person.

[0043] The above-mentioned face shape recognition model can be an algorithm or machine learning model for analyzing and classifying face shapes, for example, a deep neural network model or a convolutional neural network model, but is not limited thereto. The above-mentioned face shape recognition result can be a shape category or feature description of the analyzed face obtained after analyzing the face image through the above-mentioned face shape recognition model.

[0044] In an optional embodiment, considering that facial image recognition through manual recognition may be affected by fatigue, emotions or other human factors, it is difficult to maintain a high degree of consistency and reliability, resulting in a decrease in the accuracy of face recognition. Compared with manual recognition, facial images are recognized through a pre-trained face recognition model, which can process a large amount of facial image data in near real time. Especially in large-scale facial image data scenarios, the recognition speed of the face recognition model far exceeds that of manual recognition, and at the same time, it can ensure that the face recognition results have a high degree of accuracy. Therefore, the detection system can recognize facial images based on a pre-trained face recognition model to improve recognition efficiency and accuracy.

[0045] For example, in order to improve the accuracy, efficiency and user experience of face recognition, the face recognition model can be a pre-trained deep neural network model. The detection system can input the face image into the deep neural network model, and the deep neural network model performs pre-processing such as adjusting the image size, graying, normalizing and removing noise on the face image, and then the deep neural network model can output the face category to which the face image belongs.

[0046] For another example, in order to improve recognition speed and accuracy, the face shape recognition model can be a pre-trained convolutional neural network model. The detection system can input the above-mentioned face image into the above-mentioned convolutional neural network model, and the deep neural network model performs pre-processing such as resizing, grayscale or color standardization, correcting facial posture, cropping and aligning the face on the face image, and then the convolutional neural network model can output the face shape category to which the face image belongs.

[0047] In an embodiment of the present invention, the environmental information of the target person's environment in a historical time period is obtained; based on the environmental information, the face detection environment of the target person is determined; in response to the environmental type of the face detection environment being a preset type; based on the face recognition model, the face image is recognized to obtain the face recognition result of the target person, and the face detection environment is screened and matched with the specific environmental type of the above face detection environment to reduce the negative impact of complex environmental factors on the face detection of the person. Further, based on the face recognition model, the face images under the above different environmental types are intelligently determined, and the specific face categories under different environmental types can be determined more accurately, achieving the purpose of reducing the error of face determination in the related technology, thereby achieving the technical effect of improving the accuracy of face determination, and then solving the technical problem of low accuracy of face recognition of the related technology.

[0048] Furthermore, the method also includes: detecting the influencing parameters included in the face shape detection environment to obtain parameter influence indicators, wherein the influencing parameters are used to characterize the parameters that will affect the quality of the collected image when the image is collected in the face shape detection environment, and the parameter influence indicators are used to characterize the degree of influence of the influencing parameters on the collected image; in response to the parameter influence indicator being greater than or equal to a preset indicator, determining that the environment type of the face shape detection environment is a preset type.

[0049] The above-mentioned influencing parameters may refer to parameters that may affect the quality of the collected image in the face shape detection environment, for example, light intensity, dust concentration, background complexity, etc., but are not limited thereto. The above-mentioned parameter influencing index may refer to the degree of influence of the above-mentioned influencing parameters on the above-mentioned collected image. The above-mentioned preset index may be pre-set and used to determine the degree of influence of the environmental type of the face shape detection environment.

[0050] In an optional embodiment, considering that the face shape detection process may be affected by a variety of environmental factors in the above-mentioned face shape detection environment, in order to better adapt to different environmental conditions and thus improve the accuracy of face shape detection in different environments, the detection system may detect the influencing parameters contained in the above-mentioned face shape detection environment to obtain the degree of influence of the above-mentioned influencing parameters on the collected image, that is, the parameter influence index. If the above-mentioned parameter influence index is greater than or equal to the preset index, the detection system may determine that the environmental type of the face shape detection environment is the preset type.

[0051] For example, the above-mentioned influencing parameters may be light intensity and dust concentration, wherein the preset index of light intensity may be 60%, and the preset index of dust concentration may be 40%. The detection system detects the above-mentioned influencing parameters, and obtains that the parameter influencing index of light intensity may be 80%, which is greater than the above-mentioned preset index, and at the same time, obtains that the parameter influencing index of dust concentration may be 50%, which is also greater than the above-mentioned preset index, then the environment type of the current face shape detection environment may be determined by the detection system as the preset type.

[0052] For another example, the above-mentioned influencing parameters may be light intensity, dust concentration, and background complexity, wherein the preset index of light intensity may be 55%, the preset index of dust concentration may be 30%, and the preset index of background complexity may be 70%. The detection system detects the above-mentioned influencing parameters, and obtains that the parameter influencing index of light intensity may be 55%, the parameter influencing index of dust concentration may be 50%, and the parameter influencing index of background complexity may be 20%. The detection system may also perform weighted fusion on the above-mentioned different parameter influencing indexes based on different weights to obtain a weighted parameter influencing index. The detection system may further compare the above-mentioned weighted parameter influencing index with the above-mentioned preset index. If the weighted parameter influencing index is greater than or equal to the preset index, the detection system may determine that the environmental type of the current face shape detection environment is the preset type. If the weighted parameter influencing index is less than the preset index, the detection system may determine that the environmental type of the current face shape detection environment is not the preset type.

[0053] It should be noted that the selection of the values ​​of the above-mentioned preset indicators of light intensity, dust concentration, background complexity, etc. is only for illustrative purposes and is not limited here.

[0054] Furthermore, the influencing parameters include at least: light intensity and dust concentration of the face detection environment. The influencing parameters contained in the face detection environment are detected to obtain parameter influence indicators, including: light intensity detection of the face detection environment to obtain a light intensity index; dust concentration detection of the face detection environment to obtain a dust concentration index; weighted fusion of the light intensity index and the dust concentration index to obtain a parameter influence index.

[0055] The light intensity detection may refer to the process of measuring the brightness of light in a specific area or point. The dust concentration detection may refer to the process of measuring the concentration of suspended particles in the air. The light intensity index may be used to reflect the current light intensity in the face shape detection environment. The dust concentration index may be used to reflect the current dust concentration in the face shape detection environment.

[0056] In an optional embodiment, considering that the factors that affect the accuracy of target person face detection in the underground mine environment are mainly light intensity and dust concentration, the detection system needs to focus on the detection of light intensity and dust concentration. Further considering that the degree of influence of light intensity and dust concentration on face detection may be different, it is not possible to simply determine the degree of influence of light intensity or dust concentration in the current environment separately. Therefore, the detection system can first perform light intensity detection on the above-mentioned face detection environment to obtain the above-mentioned light intensity index, and at the same time, perform dust concentration detection on the face detection environment to obtain the above-mentioned dust concentration index. Finally, the detection system can perform weighted fusion of the light intensity index and the dust concentration index to obtain the parameter influence index in the current environment, so as to refine the degree of influence of light intensity and dust concentration on face detection, thereby improving the accuracy of the detection system on target person face detection.

[0057] For example, the total weight can be set to 1, and the weight of the light intensity can be further set to A, and the weight of the dust concentration can be set to B, and A and B satisfy A+B=1. In order to obtain the face shape detection result more accurately, the detection system can perform light intensity detection on the face shape detection environment, and obtain the light intensity index C. At the same time, the detection system can perform dust concentration detection on the face shape detection environment, and obtain the above dust concentration index D. Based on the above-set weights, the light intensity index and the dust concentration index can be weighted fused to obtain the parameter influence index E as shown in the following formula:

[0058] E=A×C+B×D.

[0059] It should be noted that the selection of the above-mentioned light intensity weight, dust concentration weight, light intensity index, and dust concentration index is only for illustrative purposes and is not limited here.

[0060] Furthermore, the face image is recognized based on the face shape recognition model to obtain the face shape recognition result of the target person, including: extracting features from the face image to obtain multiple face image features; constructing an initial face shape based on the multiple face image features; matching the initial face shape with multiple preset face shapes respectively to determine the target face shape of the target person from the multiple preset face shapes; marking the target face shape based on the person identification of the target person to obtain the face shape recognition result.

[0061] The above-mentioned facial image features may refer to specific attributes or marks in the facial image that can be recognized and distinguished by the above-mentioned face shape recognition model. The above-mentioned initial face shape may be a basic understanding or classification framework of the above-mentioned face shape recognition model on the shape and features of the face before specific training or optimization. The above-mentioned preset face shape may be a series of face shape categories pre-defined in the above-mentioned face shape recognition model, which are used to constitute the basis for recognition and classification. For example, the above-mentioned preset face shape may include at least one or more of the following: round, oval, long, square, heart-shaped, etc., but not limited to this. Each face shape has its specific facial feature description, for example, a long face may have a long facial contour, while a square face has a wide and angular chin. The above-mentioned target face shape may refer to the face shape of a specific person that the face shape recognition model attempts to recognize and classify in a specific application scenario. The above-mentioned person identification may be an identification used to associate and index the identified facial features or face shape with a specific person, for example, it may be a person number, a biometric number, a radio frequency identification tag, etc., but not limited to this.

[0062] In an optional embodiment, considering that the facial features of different target persons may have similarities, for example, the facial features of target person A and target person B are both round, in order to quickly determine the facial shape of the target person, thereby improving the efficiency of facial recognition, the facial recognition model can pre-construct multiple preset face shapes based on more common face shapes, such as round, oval, long, square, heart-shaped, etc. When facial recognition is required, it is only necessary to match the facial features of the target person with the above-mentioned preset face shapes to quickly determine the target face shape corresponding to the target person. Therefore, in the facial recognition process, the facial recognition model can extract multiple facial image features from the face image, and based on the multiple facial image features, construct the initial face shape corresponding to the above-mentioned facial image features, and then, the facial recognition model can match the above-mentioned initial face shape with multiple preset face shapes respectively, thereby determining the target face shape of the target person from multiple preset face shapes. In order to accurately establish the corresponding relationship between the above-mentioned target face shape and the target person, the facial recognition model can mark the above-mentioned target face shape based on the person identification of the target person, thereby obtaining a facial recognition result that accurately corresponds to each target person.

[0063] For example, in order to improve the efficiency of face recognition and ensure that each target person can obtain a face recognition result that accurately corresponds to himself, the detection system can first perform preprocessing such as cropping, graying, and image enhancement on the above-mentioned face image to ensure that the face recognition model can accurately analyze the face image. Then, the detection system can input the preprocessed image into the trained face recognition model, and the face recognition model can match the initial face shape with multiple preset face shapes respectively, so as to determine the target face shape of the target person from multiple preset face shapes. Finally, the detection system can associate the above-mentioned target face shape with the target person ID (Identification) to obtain a face recognition result that accurately corresponds to each target person.

[0064] Furthermore, the initial face shape is matched with multiple preset face shapes respectively, and the target face shape of the target person is determined from the multiple preset face shapes, including: obtaining face shape matching values ​​corresponding to different preset face shapes, wherein the face shape matching values ​​are used to characterize the degree of matching between the initial face shape and the different preset face shapes; based on the face shape matching values, a first face shape is screened out from the multiple preset face shapes, wherein the face shape matching value corresponding to the first face shape is greater than or equal to the preset value; outputting the first face shape, and receiving a face shape selection result, wherein the face shape selection result includes a second face shape selected by the target person from the first face shape; and determining the second face shape as the target face shape.

[0065] The above-mentioned face shape matching value may refer to the value of the similarity between the current face shape and the preset face shape. The above-mentioned first face shape may refer to one or more face shapes among the preset face shapes that have a higher degree of matching with the current face shape. The above-mentioned preset value may be a face shape matching value used to filter out the above-mentioned first face shape from the above-mentioned multiple preset face shapes. The above-mentioned face shape selection result may be a face shape after further screening of the above-mentioned first face shape, and the matching value corresponding to the face shape. The above-mentioned second face shape may be a face shape among the first face shapes that meets the needs of the target person.

[0066] In an optional embodiment, considering that there may be multiple candidate face shapes similar to the target person's face shape in the above-mentioned preset face shapes, that is, face shapes that may meet the needs of the target person, in order to screen out the face shape with a higher degree of similarity to the target person's face shape from the above-mentioned candidate face shapes, thereby obtaining a more accurate face shape recognition result, it is necessary to further process the above-mentioned candidate face shapes. In order to improve the accuracy of face shape recognition, the face shape recognition model first matches the constructed initial face shape with different preset face shapes to determine the degree of match between the initial face shape and the different preset face shapes, and then determines the face shape matching values ​​corresponding to the different preset face shapes based on the degree of match between the above-mentioned initial face shape and the different preset face shapes. After determining the above-mentioned face shape judgment value, the face shape recognition model can screen out one or more face shapes with face shape matching values ​​greater than or equal to the preset value from multiple preset face shapes as the above-mentioned first face shape. Subsequently, the detection system can display the above-mentioned first face shape screened out by the face shape recognition model to the target person. The target person can determine the face shape that meets the target person's needs from the first face shape as the above-mentioned second face shape, and generate a face shape selection result based on the above-mentioned second face shape, and return it to the detection system. The detection system can determine the above-mentioned second face shape from the first face shape according to the face shape selection result returned by the user as the target face shape of the target person.

[0067] For example, the preset value may be 70%, and the preset face shapes may include but are not limited to round, oval, long, square, heart-shaped, etc. Assume that the matching degree of the circle is 85%, the matching degree of the oval is 80%, the matching degree of the long is 50%, the matching degree of the square is 75%, and the matching degree of the heart is 40%. Based on the matching degree of the above-mentioned various face shapes, the face shape recognition model may use the circle, oval, and square as the above-mentioned first face shape, and output the first face shape to the detection system. The detection system may display the above-mentioned first face shape to the target person, and the target person may determine that the above-mentioned second face shape is a circle according to the needs of the target person, and return the above-mentioned second face shape and the matching degree of the face shape to the detection system.

[0068] It should be noted that the selection of all the above-mentioned preset values, circular matching degree, elliptical matching degree, elongated matching degree, etc., is only for illustrative purposes. The staff can set them according to actual needs and are not limited here.

[0069] Furthermore, the method also includes: obtaining multiple training images matching different preset face shapes, wherein the number of multiple training images corresponding to different preset face shapes is the same, and the number of multiple training images is greater than or equal to a preset number; training the initial recognition model based on multiple training images according to a preset number of training times, a preset learning rate and a preset number of samples to obtain a face shape recognition model, wherein the preset number of training times is greater than or equal to a first value, the preset learning rate is a second value, the preset learning rate decreases as the number of training times increases according to a preset value, the preset number of samples is used to characterize the number of samples for updating the model weights of the initial recognition model, and the preset number of samples is greater than or equal to a third value.

[0070] The above-mentioned training images may refer to images in the data set used to train the above-mentioned face recognition model. The above-mentioned preset number may be the number of the above-mentioned multiple training images preset in order to achieve the expected face recognition model training goal. The above-mentioned preset training times may refer to the preset number of iterations for training the above-mentioned face recognition model on the complete training data set. The above-mentioned preset learning rate may be a preset value for controlling the update speed of the parameters of the above-mentioned face recognition model in each iteration. The above-mentioned preset number of samples may be a preset number of samples for updating the model weights of the initial recognition model. The above-mentioned initial recognition model may refer to the initial state of the above-mentioned face recognition model before starting training. The above-mentioned first value may be a value for describing the minimum value of the above-mentioned preset number of training times. The above-mentioned second value may be a value for describing the above-mentioned preset learning rate. The above-mentioned third value may be a value for describing the minimum value of the above-mentioned preset number of samples.

[0071] In an optional embodiment, considering that the above-mentioned face recognition model needs to be able to accurately and efficiently learn and generalize to unseen face shapes, the detection system can obtain multiple training images matching different preset face shapes, and then the detection system can train the initial recognition model based on the multiple training images according to the above-mentioned preset number of training times, the above-mentioned preset learning rate and the above-mentioned preset number of samples to obtain the above-mentioned face recognition model. It should be noted that in order to ensure the accuracy of the trained face recognition model in recognizing different preset face shapes, the number of multiple training images corresponding to different preset face shapes can be set to be the same, and the number of multiple training images can be set to be greater than or equal to the preset number. In addition, in order to improve the training effect of the above-mentioned face recognition model, the above-mentioned preset number of training times can be set to be greater than or equal to the first value, the above-mentioned preset learning rate can be set to be a second value, the above-mentioned preset learning rate can be set to decrease according to the preset number of training times, and the above-mentioned preset number of samples can be set to be greater than or equal to the third value.

[0072] For example, the preset number may be 10,000, the preset number of training times may be 20 rounds, the preset learning rate may be 0.01, the preset number of samples may be 32, the preset face shape may be a square face, a round face, or an oval face, the first value may be 16, the second value may be 0.01, and the third value may be 30. In order to accurately and efficiently train the face shape recognition model, the detection system may obtain a plurality of training images matching the different preset face shapes, wherein the number of training images for a square face is 20,000, the number of training images for a round face is 15,000, and the number of training images for an oval face is 11,000, and the number of training images for the different face shapes is greater than the preset number. Then, the detection system may train the initial recognition model based on a plurality of training images according to the preset number of training times, the preset learning rate, and the preset number of samples.

[0073] Specifically, the preset number of training times is greater than the first value, the preset learning rate is the second value, and the preset number of samples is greater than the third value. Based on the preset number of training times, the initial training model needs to traverse the entire data set at least 20 times. Based on the preset learning rate, the detection system can use a dynamic learning rate strategy. After a certain number of epochs, the learning rate decreases according to a preset attenuation strategy. Based on the preset number of samples, in each round of training, the face recognition model will process 32 images at the same time and then update the weights.

[0074] It should be noted that the above-mentioned preset numbers, preset training times and other values ​​are only for illustrative purposes. The staff can set them according to actual needs and are not limited here.

[0075] Furthermore, the face shape recognition model is a deep neural network model, the size of the convolution kernel included in the face shape recognition model is less than or equal to the fourth value, and the number of network structure layers of the face shape recognition model is greater than or equal to the fifth value.

[0076] The convolution kernel may be a small multidimensional matrix that slides over the input data and extracts features from the input data through a convolution operation. The fourth value may be a value that describes the maximum value of the size of the convolution kernel. The network structure layer may be a hierarchical structure of computing units that describes data from input to output. The fifth value may be a value that describes the minimum value of the number of network structure layers.

[0077] In an optional embodiment, considering that the data involved in the face recognition needs contain rich and complex visual information, such as facial contours, bone structure, distribution of facial features, etc., these features are often nonlinear and interrelated. Therefore, the above-mentioned face recognition model adopts a deep neural network model, which can learn more robust and generalized feature representations from these large-scale data, thereby improving the accuracy and robustness of the face recognition model. It should be noted that in order to improve the recognition accuracy of the above-mentioned face recognition model, the face recognition model can use a deep neural network model. The deep neural network model used by the above-mentioned face recognition model contains a convolution kernel size that is less than or equal to the above-mentioned fourth value, and the number of network structure layers of the above-mentioned deep neural network model is greater than or equal to the above-mentioned fifth value. For example, the above-mentioned fourth value can be 3×3, the above-mentioned fifth value can be 50, the corresponding above-mentioned convolution kernel size can be 3×3, and the number of the above-mentioned network structure layers can be 60.

[0078] It should be noted that the fourth value, the fifth value, the convolution kernel size, the number of network structure layers and other values ​​mentioned above are only exemplary displays. The staff can set them according to actual needs and are not limited here.

[0079] For ease of understanding, Figure 2 is a detailed flow chart of a method for face recognition according to an embodiment of the present invention. Figure 2As shown, the method can be composed of three parts, namely: training set construction, deep learning model training, and rapid face shape determination. Among them, the training set construction includes the following steps: determining the total number of face shape categories of mine workers, which is used to determine the face shape determination range of the present invention; face image acquisition, the detection system can be based on the face shape categories involved in the above-mentioned determination of the total number of face shape categories of personnel, and collect them in the underground mine staff clock-in sign-in system; face images correspond to face shapes, and the detection system can be based on the above-mentioned determination of the total number of face shape categories of personnel and the face image acquisition results, and one-to-one correspondence is performed according to the ID information of the personnel as an index, and finally the training set of the above-mentioned deep learning model is formed. The deep learning model training includes the following two steps: model training data selection, the detection system can use the training set in the above training set construction to select data, ensure that the number of people and face images under the above preset face shape is evenly distributed, and the total scale of the above selected model training data can be above 100,000; deep neural network model design, the detection system can carry out model training based on the above model training data selection, and the above model training process can follow the basic principles of convolution kernel size not higher than 3*3, total number of layers more than 50, training rounds not less than 20 epochs, batch-size (batch size) size not less than 32, initial learning rate 0.01 and dynamic reduction. Rapid face shape determination includes the following two steps: face shape prediction of unknown face images, the detection system can predict the face shape of unknown face images based on the model obtained by deep learning model training; face shape information determination of mine workers, the detection system can set a reasonable threshold based on the prediction results in the face shape prediction of the above unknown face images, and determine the specific face shape information of the mine workers according to the comparison results of the face shape prediction value and the above threshold.

[0080] It should be noted that the above-mentioned values ​​such as the total size of training data, convolution kernel size, number of network structure layers, and training rounds are only for illustrative purposes. The staff can set them according to actual needs and are not limited here.

[0081] The embodiment of the present application also provides a person face recognition device, which should be noted that the device can be used to execute the above-mentioned person face recognition method. The specific implementation method and application scenario are the same as those of the above-mentioned embodiment, and will not be repeated here. Figure 3 is a schematic diagram of a person face recognition device according to an embodiment of the present invention. Figure 3 As shown, the device comprises:

[0082] The first acquisition module 302 is used to acquire the environment information of the target person's environment in a historical time period, wherein the environment information at least includes the number of times different environments appear.

[0083] The environment determination module 304 is used to determine the face shape detection environment of the target person based on the environment information, wherein the number of times the face shape detection environment appears in the environment information is greater than or equal to a preset number of times.

[0084] The second acquisition module 306 is used to acquire the face image of the target person acquired by the image acquisition device in response to the environment type of the face detection environment being a preset type, wherein the image acquisition device is deployed in the face detection environment and the target person uses the image acquisition device at a frequency greater than or equal to a preset frequency.

[0085] The face recognition module 308 is used to recognize the face image based on the face recognition model to obtain the face recognition result of the target person.

[0086] Furthermore, the device also includes: an indicator determination module, which is used to detect the influencing parameters contained in the face shape detection environment to obtain parameter influence indicators, wherein the influencing parameters are used to characterize the parameters that will affect the quality of the collected image when the image is collected in the face shape detection environment, and the parameter influence indicators are used to characterize the degree of influence of the influencing parameters on the collected image; a preset environment determination module, which is used to determine that the environment type of the face shape detection environment is a preset type in response to the parameter influence indicator being greater than or equal to the preset indicator.

[0087] Furthermore, the influencing parameters include at least: the light intensity and dust concentration of the face detection environment, and the index determination module is also used to: perform light intensity detection on the face detection environment to obtain a light intensity index; perform dust concentration detection on the face detection environment to obtain a dust concentration index; perform weighted fusion of the light intensity index and the dust concentration index to obtain a parameter influencing index.

[0088] Furthermore, the face shape recognition module is also used to: extract features from facial images to obtain multiple facial image features; construct an initial face shape based on the multiple facial image features; match the initial face shape with multiple preset face shapes respectively, and determine the target face shape of the target person from the multiple preset face shapes; mark the target face shape based on the person identification of the target person to obtain a face shape recognition result.

[0089] Furthermore, the face shape recognition module is also used to: obtain face shape matching values ​​corresponding to different preset face shapes, wherein the face shape matching values ​​are used to characterize the degree of matching between the initial face shape and the different preset face shapes; screen out a first face shape from multiple preset face shapes based on the face shape matching values, wherein the face shape matching value corresponding to the first face shape is greater than or equal to the preset value; output the first face shape, and receive a face shape selection result, wherein the face shape selection result includes a second face shape selected by the target person from the first face shape; and determine that the second face shape is the target face shape.

[0090] Furthermore, the device also includes: a third acquisition module, used to acquire multiple training images matching different preset face shapes, wherein the number of multiple training images corresponding to different preset face shapes is the same, and the number of multiple training images is greater than or equal to the preset number; a fourth acquisition module, used to train the initial recognition model based on multiple training images according to a preset number of training times, a preset learning rate and a preset number of samples to obtain a face shape recognition model, wherein the preset number of training times is greater than or equal to a first value, the preset learning rate is a second value, the preset learning rate decreases as the number of training times increases according to a preset value, the preset number of samples is used to characterize the number of samples for updating the model weights of the initial recognition model, and the preset number of samples is greater than or equal to a third value.

[0091] Furthermore, the face shape recognition model is a deep neural network model, the size of the convolution kernel included in the face shape recognition model is less than or equal to the fourth value, and the number of network structure layers of the face shape recognition model is greater than or equal to the fifth value.

[0092] An embodiment of the present application further provides an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods in various embodiments of the present invention when running.

[0093] An embodiment of the present application further provides a computer-readable storage medium, which includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in various embodiments of the present invention.

[0094] An embodiment of the present application further provides a computer program product, including a computer program, which implements the methods in various embodiments of the present invention when executed by a processor.

[0095] An embodiment of the present application further provides a computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium is used to store a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0096] The embodiments of the present application further provide a computer program, which implements the methods in the above-mentioned embodiments of the present invention when executed by a processor.

[0097] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0098] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0099] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0100] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0101] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0102] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for face recognition, characterized in that: include: Obtaining environmental information of the target person's environment in a historical period of time, wherein the environmental information at least includes the number of times different environments appear; Based on the environmental information, determining a face shape detection environment of the target person, wherein the number of times the face shape detection environment appears in the environmental information is greater than or equal to a preset number; In response to the environment type of the face shape detection environment being a preset type, acquiring a face image of the target person acquired by an image acquisition device, wherein the image acquisition device is deployed in the face shape detection environment, and a frequency of the target person using the image acquisition device is greater than or equal to a preset frequency; The face image is recognized based on a face recognition model to obtain a face recognition result of the target person.

2. The method according to claim 1, characterized in that The method further comprises: Detecting the influencing parameters included in the face shape detection environment to obtain parameter influencing indicators, wherein the influencing parameters are used to characterize the parameters that will affect the quality of the collected image when collecting the image in the face shape detection environment, and the parameter influencing indicators are used to characterize the degree of influence of the influencing parameters on the collected image; In response to the parameter influence index being greater than or equal to a preset index, determining that the environment type of the face shape detection environment is the preset type.

3. The method according to claim 2, characterized in that The influencing parameters at least include: light intensity and dust concentration of the face shape detection environment. The influencing parameters included in the face shape detection environment are detected to obtain parameter influencing indicators, including: Performing light intensity detection on the face shape detection environment to obtain a light intensity index; Performing dust concentration detection on the face shape detection environment to obtain a dust concentration index; The light intensity index and the dust concentration index are weightedly fused to obtain the parameter influence index.

4. The method according to claim 1, characterized in that: The face image is recognized based on the face recognition model to obtain the face recognition result of the target person, including: Extracting features from the facial image to obtain multiple facial image features; constructing an initial face shape based on the multiple facial image features; Matching the initial face shape with a plurality of preset face shapes respectively, and determining a target face shape of the target person from the plurality of preset face shapes; The target face shape is marked based on the person identification of the target person to obtain the face shape recognition result.

5. The method according to claim 4, characterized in that Matching the initial face shape with a plurality of preset face shapes respectively, and determining a target face shape of the target person from the plurality of preset face shapes, comprises: Obtaining face shape matching values ​​corresponding to different preset face shapes, wherein the face shape matching values ​​are used to represent the matching degree between the initial face shape and the different preset face shapes; Filtering a first face shape from the plurality of preset face shapes based on the face shape matching value, wherein the face shape matching value corresponding to the first face shape is greater than or equal to the preset value; Outputting the first face shape, and receiving a face shape selection result, wherein the face shape selection result includes a second face shape selected by the target person from the first face shape; The second face shape is determined as the target face shape.

6. The method according to claim 1, characterized in that The method further comprises: Acquire a plurality of training images matching different preset face shapes, wherein the number of the plurality of training images corresponding to the different preset face shapes is the same, and the number of the plurality of training images is greater than or equal to a preset number; The face recognition model is obtained by training the initial recognition model based on the multiple training images according to a preset number of training times, a preset learning rate and a preset number of samples, wherein the preset number of training times is greater than or equal to a first value, the preset learning rate is a second value, the preset learning rate decreases as the number of training times increases according to a preset value, the preset number of samples is used to characterize the number of samples for updating the model weights of the initial recognition model, and the preset number of samples is greater than or equal to a third value.

7. The method according to claim 1, characterized in that The face shape recognition model is a deep neural network model, the size of the convolution kernel included in the face shape recognition model is less than or equal to a fourth value, and the number of network structure layers of the face shape recognition model is greater than or equal to a fifth value.

8. An electronic device, characterized in that: include: A memory storing an executable program; A processor, configured to run the program, wherein the program executes the method according to any one of claims 1 to 7 when running.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored executable program, wherein when the executable program is executed, the device where the storage medium is located is controlled to execute the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that The invention comprises a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images