A high-precision face recognition camera

Abstract

The present invention provides a high-precision face recognition camera, which includes: an image acquisition module for acquiring a face image within a target area and performing image enhancement processing on the face image to obtain a target face image; a detection module for determining the current position and current motion state of the user corresponding to each target face image after image enhancement processing of the same user; based on the current position and current motion state, determining whether each target face image meets the shooting requirements and retaining the corresponding image; a feature extraction module for analyzing the retained images and extracting feature information in the retained images; an identity determination module for matching the feature information with the reference image information stored in a preset face image library to determine the identity information of the target person in the retained target face image, thereby completing high-precision recognition of the face. This improves the accuracy of face image recognition and also improves the precision of the camera for face recognition.

Description

Technical Field

[0001] The present invention relates to the technical field of cameras, and particularly to a high-precision face recognition camera. Background Art

[0002] Currently, cameras only have the function of image acquisition. If the function of face recognition is to be realized, a background server needs to be set up. The background server compares and recognizes the face images of the photographed persons and outputs the recognition results. Due to its inconvenient movement, such a system is only applicable to specific places, such as inside a company or a security checkpoint, and has poor applicability. Generally, during recognition, only simple image comparison is used to obtain the recognition results. However, there may be situations where the recognition is not timely and the recognition accuracy is low.

[0003] Therefore, the present invention provides a high-precision face recognition camera, which is used to collect and process the face images of people in the target area, and through detecting the images, to ensure that the face images are clear enough, improve the accuracy of face image recognition, and at the same time quickly and accurately determine the identity information of the current person according to the image analysis results, enhance the practicability of face recognition, and also improve the accuracy of the camera for face recognition. Summary of the Invention

[0004] The present invention provides a high-precision face recognition camera, which is used to collect and process the face images of people in the target area, and through detecting the images, to ensure that the face images are clear enough, improve the accuracy of face image recognition, and at the same time quickly and accurately determine the identity information of the current person according to the image analysis results, enhance the practicability of face recognition, and also improve the accuracy of the camera for face recognition.

[0005] The present invention provides a high-precision face recognition camera, comprising:

[0006] An image acquisition module, configured to collect face images in the target area and perform image enhancement processing on the face images to obtain target face images;

[0007] A detection module, configured to determine the current position and current motion state of the user corresponding to each target face image after image enhancement processing for the same user;

[0008] Based on the current position and current motion state, determine whether each target face image meets the shooting requirements. If it meets the requirements, input the corresponding target face image into a general image detection model for detection. If it does not meet the requirements, adjust the acquisition accuracy of the camera for re-acquisition and processing, and input the re-acquired and processed target face images into a specific image detection model for detection;

[0009] When the detection result meets the detection requirements, the corresponding target face image is retained;

[0010] A feature extraction module, configured to analyze the retained target face image and extract the feature information in the retained target face image;

[0011] An identity determination module, configured to match the feature information with the reference image information stored in the preset face image library to determine the identity information of the target person in the retained target face image, and complete the high-precision recognition of the face.

[0012] Preferably, the image acquisition module includes:

[0013] A person detection unit, configured to monitor the target area in real time based on a preset camera, and determine whether there is a person in the target area based on the monitoring result;

[0014] An image acquisition unit, configured to, when there is a person in the target area, control the preset camera to collect face images of the person in the target area from different shooting angles to obtain multiple face images of the person;

[0015] An optimal image determination unit, configured to screen the multiple face images based on a preset screening requirement to determine the optimal face image, where the optimal face image is a frontal face image of the person in the target area;

[0016] A monitoring unit, configured to continue to monitor the target area in real time when there is no person in the target area.

[0017] Preferably, the image acquisition unit includes:

[0018] A distance determination subunit, configured to obtain the distance value between the person in the target area and the preset camera, and compare the distance value with a first preset threshold and a second preset threshold;

[0019] If the distance value is greater than the first preset threshold, control the preset camera to perform magnification processing on the image of the person in the target area;

[0020] If the distance value is greater than or equal to the second preset threshold and less than or equal to the first preset threshold, it is determined that the distance value between the person in the target area and the preset camera is appropriate, and face image acquisition is performed on the person in the target area based on the distance value;

[0021] If the distance value is less than the second preset threshold, control the preset camera to perform blurring processing on the area other than the face image of the person in the target area to obtain the final face image.

[0022] Preferably, the image acquisition module further includes:

[0023] An image segmentation unit, configured to obtain the face image and determine the brightness value of each pixel point in the face image;

[0024] An image segmentation unit, configured to segment the face image into N layers based on the brightness value, and respectively determine the brightness value of the pixel points in each layer;

[0025] A brightness enhancement unit, configured to compare the pixel point brightness value with a preset brightness value, determine the target pixel points in each layer where the pixel point brightness value is less than the preset brightness value, and perform enhancement processing on the brightness value of the target pixel points in each layer based on the preset processing method to obtain N initial enhanced layers;

[0026] A noise removal unit, configured to perform noise removal processing on the N initial enhanced layers to obtain N target enhanced layers;

[0027] An image fusion unit, configured to fuse the N target enhanced layers to obtain a target face image.

[0028] Preferably, the feature extraction module includes:

[0029] A segmentation model training unit, configured to construct an image main body segmentation model and obtain historical face images, where the historical face images are at least two;

[0030] A segmentation model training unit, configured to determine the main body area and the background area in the historical face image, determine the segmentation line between the main body area and the background area based on a preset segmentation annotation method, and perform segmentation annotation on the segmentation line;

[0031] The segmentation model training unit is further configured to determine the image direction of the historical face image based on the segmentation annotation result, and determine a segmentation scheme for the historical face image based on the picture direction;

[0032] The segmentation model training unit is further configured to train the image main body segmentation model based on the segmentation scheme to obtain a target image main body segmentation model;

[0033] A main body segmentation unit, configured to perform segmentation processing on the reserved target face image based on the target image main body segmentation model to obtain a target face main body image in the reserved target face image, and place the reserved target face main body image in a preset two-dimensional coordinate system;

[0034] A face size determination unit, configured to determine the image size information of the face in the reserved target face main image based on the preset two-dimensional coordinate system, and simultaneously determine the scaling ratio between the reserved target face image and the actual object size;

[0035] The face size determination unit is further configured to determine the actual size information of the face in the reserved target face main image based on the scaling ratio, where the actual size information includes the length and width of the face and the face shape information of the face;

[0036] A feature extraction unit, configured to obtain a historical face image dataset, and train a convolutional neural network based on the historical face image dataset to obtain a face recognition model;

[0037] The feature extraction unit is configured to process the target face main image based on the face recognition model, extract the target feature points of the face in the target face main image, and determine the position information of the target feature points on the face based on the actual size information of the face;

[0038] A feature information determination unit, configured to determine the feature information in the reserved target face image based on the actual size information of the face and the position information of the target feature points on the face.

[0039] Preferably, the identity determination module includes:

[0040] A data acquisition unit, configured to acquire the feature information, and match the feature information with the reference image information stored in the preset face image library respectively to obtain the matching degree between the feature information and the reference image information, where the reference image information is not unique;

[0041] An optimal reference image information determination unit, configured to sort the matching degrees in a descending order, and determine the reference image information with the largest matching degree as the target reference image information based on the sorting result;

[0042] An identity information determination unit, configured to determine the image identifier of the target reference image information, and match the target identity information from the preset identity information storage library based on the image identifier to obtain the identity information of the target person in the reserved target face image.

[0043] Preferably, the optimal reference image information determination unit includes:

[0044] An identity verification sub-unit, configured to acquire the matching degree between the feature information and the reference image information, and compare the matching degree with a preset matching degree;

[0045] If the matching degree is greater than or equal to the preset matching degree, it is determined that the reference image information matched with the feature information is qualified, and the identity information corresponding to the reference image information is determined;

[0046] Otherwise, it is determined that there is no reference image information in the preset face image library that matches the feature information, and the collected face image is transmitted to the management terminal and an alarm notification is made.

[0047] Preferably, the image acquisition module includes:

[0048] An image acquisition unit, configured to perform multiple face image acquisitions on the people inside the target area based on a preset camera, obtain face images, and determine the size information of the face images;

[0049] A calculation unit, configured to calculate the resolution of the preset camera according to the size information of the face image, and calculate the qualification rate of the face images acquired by the preset camera based on the resolution. The specific steps include:

[0050] A first calculation unit, configured to calculate the resolution of the preset camera according to the following formula:

[0051]

[0052] where α represents the resolution of the preset camera; μ represents the error factor, and its value range is (0.05, 0.15); β represents the total number of pixel points in the face image; δ represents the pixel coefficient, and its value is 1*10 6 ; G represents the ratio of the horizontal pixel dimension to the vertical pixel dimension in the face image; H represents the ratio of the vertical pixel dimension to the horizontal pixel dimension in the face image; L represents the length value of the face image; K represents the width value of the face image;

[0053] A second calculation unit, configured to calculate the qualification rate of the face images acquired by the preset camera according to the following formula:

[0054]

[0055] where η represents the qualification rate of the face images acquired by the preset camera, and its value range is (0, 1); ψ represents the number of face images that meet the preset clarity requirement in the multiple face image acquisitions; λ represents the total number of face images acquired in the multiple face image acquisitions;

[0056] A comparison unit, configured to compare the calculated qualification rate with the preset qualification rate;

[0057] If the qualified rate is greater than or equal to the preset qualified rate, it is determined that the acquisition of the face image of the person inside the target area by the preset camera meets the preset requirements;

[0058] Otherwise, it is determined that the acquisition of the face image of the person inside the target area by the preset camera does not meet the preset requirements, and the resolution of the preset camera is adjusted until the qualified rate is greater than or equal to the preset qualified rate.

[0059] Preferably, the detection module includes:

[0060] A prediction unit, configured to determine the current position of the user corresponding to the corresponding target face image based on the position conversion rule, and at the same time, obtain the first images at the previous and subsequent moments corresponding to the current moment, and predict the current motion state of the corresponding user;

[0061] A conversion unit, configured to construct a connection line between the positions corresponding to the previous and subsequent moments and the current position, obtain the position matrix W, and at the same time, based on the current motion state, obtain the motion matrix Y, where W and Y are matrices with 1 row and 3 columns;

[0062] An operation unit, configured to calculate the current satisfaction value Z of the corresponding target face image according to the position matrix W and the motion matrix Y;

[0063]

[0064] Among them, Y T represents the transpose of the motion matrix Y; σ1 represents the position standard deviation obtained based on each position element in the position matrix W; σ2 represents the motion standard deviation obtained based on each motion element in the motion matrix Y;

[0065] When the current satisfaction value Z is less than or equal to the preset satisfaction value, it is determined that the corresponding target face image meets the shooting requirements;

[0066] Otherwise, it is determined that the corresponding target face image does not meet the shooting requirements.

[0067] Other features and advantages of the present invention will be described in the subsequent specification, and part of them will become obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in the written specification, claims, and drawings.

[0068] Next, through the drawings and embodiments, the technical solutions of the present invention will be further described in detail. Description of the Drawings

[0069] The accompanying drawings are used to provide a further understanding of the present invention and form a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention. In the accompanying drawings:

[0070] Figure 1 It is a structural diagram of a high-precision face recognition camera in an embodiment of the present invention;

[0071] Figure 2 It is a structural diagram of an image acquisition module in a high-precision face recognition camera in an embodiment of the present invention;

[0072] Figure 3 It is a structural diagram of an identity determination module in a high-precision face recognition camera in an embodiment of the present invention. Detailed implementation manners

[0073] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention and are not used to limit the present invention.

[0074] Embodiment 1:

[0075] This embodiment provides a high-precision face recognition camera, as Figure 1 shown, including:

[0076] An image acquisition module, configured to acquire a face image in a target area and perform image enhancement processing on the face image to obtain a target face image;

[0077] A detection module, configured to determine the current position and current motion state of the user corresponding to each target face image after image enhancement processing for the same user;

[0078] Based on the current position and current motion state, determine whether each target face image meets the shooting requirements. If it meets, input the corresponding target face image into a general image detection model for detection. If it does not meet, adjust the acquisition accuracy of the camera for re-acquisition and processing, and input the re-acquired and processed target face image into a specific image detection model for detection;

[0079] When the detection result meets the detection requirements, retain the corresponding target face image;

[0080] A feature extraction module, configured to analyze the retained target face image and extract feature information in the retained target face image;

[0081] An identity determination module, configured to match the feature information with the reference image information stored in a preset face image library to determine the identity information of the target person in the retained target face image, and complete the high-precision recognition of the face.

[0082] In this embodiment, the target area refers to the monitoring area where the camera is installed, which can be the entrance of a community or an office place, etc.

[0083] In this embodiment, image enhancement processing refers to adjusting the resolution, color values, etc. of the image.

[0084] In this embodiment, the target face image refers to a clear face image obtained by adjusting various parameters in the collected original face image.

[0085] In this embodiment, the feature information refers to the facial feature information of the person contained in the face image, such as face shape, distribution of facial features, etc.

[0086] In this embodiment, the preset face image library is set in advance and stores multiple face images with known identity information in advance.

[0087] In this embodiment, the reference image information refers to the distinct feature information corresponding to the face images pre-stored in the preset face image library.

[0088] In this embodiment, the target person refers to the person contained in the target face image, which can be one or multiple.

[0089] In this embodiment, both the current position and the current motion state can be obtained based on the full-body image of the user, so as to determine the relevant information of the user.

[0090] In this embodiment, determining whether the shooting requirements are met is to avoid being too deviated in the motion state or motion position, such as moving to a position that is not completely within the shooting area.

[0091] In this embodiment, the set general image detection model and specific image detection model are both pre-trained, and in order to be able to process images that meet and do not meet the requirements, so as to ensure the clarity of the acquired images and indirectly ensure the high precision of the camera. The general one is for general recognition, and the specific one is for targeted recognition, such as image recognition for areas that are slightly off.

[0092] The beneficial effects of the above technical solutions are: by collecting and processing the face of the person in the target area, ensuring that the face image is clear enough, improving the accuracy of face image recognition, and at the same time quickly and accurately determining the identity information of the current person according to the image analysis results, enhancing the practicality of face recognition, and also improving the accuracy of the camera for face recognition.

[0093] Embodiment 2:

[0094] Based on the above-mentioned Embodiment 1, this embodiment provides a high-precision face recognition camera, as Figure 2 shown, the image acquisition module includes:

[0095] A person detection unit for real-time monitoring of the target area based on a preset camera and judging whether there is a person in the target area based on the monitoring result;

[0096] An image acquisition unit for controlling the preset camera to perform image acquisition of the face of the person in the target area based on different shooting angles when there is a person in the target area, and obtaining multiple face images of the person;

[0097] An optimal image determination unit for screening the multiple face images based on preset screening requirements to determine the optimal face image, where the optimal face image is a frontal face image of the person in the target area;

[0098] A monitoring unit for continuing to perform real-time monitoring of the target area when there is no person in the target area.

[0099] In this embodiment, the preset camera is set in advance and is used to obtain the face image of the person in the target area.

[0100] In this embodiment, different shooting angles are set in advance. For example, it can be to control the preset camera to rotate the current shooting angle to obtain different shooting angles.

[0101] In this embodiment, the preset screening requirements are set in advance. For example, it can be to ensure that the frontal face of the person in the target area can be clearly and intuitively seen in the captured image.

[0102] The beneficial effects of the above technical solutions are: By using the preset camera to perform real-time monitoring of the target area, it is convenient to perform image acquisition in a timely manner when a person appears in the target area, improving the timeliness of image acquisition. At the same time, collecting face images of the person from different shooting angles is conducive to collecting the frontal face image of the person, thus providing a guarantee for improving the accuracy of face recognition.

[0103] Embodiment 3:

[0104] Based on the above-mentioned Embodiment 2, this embodiment provides a high-precision face recognition camera, and the image acquisition unit includes:

[0105] A distance determination sub-unit for obtaining the distance value between the person in the target area and the preset camera and comparing the distance value with a first preset threshold and a second preset threshold;

[0106] If the distance value is greater than the first preset threshold, control the preset camera to magnify the image of the person inside the target area;

[0107] If the distance value is greater than or equal to the second preset threshold and less than or equal to the first preset threshold, determine that the distance value between the person inside the target area and the preset camera is moderate, and collect the face image of the person in the target area based on the distance value;

[0108] If the distance value is less than the second preset threshold, control the preset camera to blur the area other than the face image of the person inside the target area to obtain the final face image.

[0109] In this embodiment, the distance value refers to the degree of proximity between the person captured in the image frame of the preset camera and the camera.

[0110] In this embodiment, the first preset threshold refers to the maximum distance allowed between the person in the target area and the preset camera.

[0111] In this embodiment, the second preset threshold refers to the minimum distance allowed between the person in the target area and the preset camera.

[0112] In this embodiment, the magnification process refers to the preset camera pulling in and magnifying the face image of the person in the image frame.

[0113] In this embodiment, the purpose of the blurring process is to filter out the image other than the face image of the person, making the face image more distinct and prominent.

[0114] The beneficial effects of the above technical solution are: By judging the distance relationship between the person in the target area and the preset camera, the face image of the person captured by the preset camera is made more reasonable and clear, improving the accuracy of image acquisition, and at the same time facilitating the improvement of the high-precision recognition of the camera.

[0115] Embodiment 4:

[0116] Based on the above Embodiment 1, this embodiment provides a high-precision face recognition camera, and the image acquisition module further includes:

[0117] An image segmentation unit, configured to obtain the face image and determine the brightness value of each pixel point in the face image;

[0118] An image segmentation unit, configured to segment the face image into N layers based on the brightness value and respectively determine the brightness value of the pixel points in each layer;

[0119] A brightness enhancement unit for comparing the pixel point brightness value with a preset brightness value, determining target pixel points in each layer where the pixel point brightness value is less than the preset brightness value, and enhancing the brightness values of the target pixel points in each layer based on the preset processing method to obtain N initial enhanced layers;

[0120] A noise removal unit for denoising the N initial enhanced layers to obtain N target enhanced layers;

[0121] An image fusion unit for fusing the N target enhanced layers to obtain a target face image.

[0122] In this embodiment, the preset brightness value is set in advance and is used to measure whether the brightness value of each pixel point in the layer meets the requirements of image analysis.

[0123] In this embodiment, the target pixel points refer to the pixel points in each layer where the pixel point brightness value is less than the preset brightness value.

[0124] In this embodiment, the preset processing method is set in advance and is used to determine the adjustment of the brightness value of the pixel points in the layer.

[0125] In this embodiment, fusing the N target enhanced layers means orderly fusing the N target enhanced layers according to the positional relationship of the layers in the original face image.

[0126] The beneficial effects of the above technical solution are: By splitting the face image into multiple layers and realizing the adjustment of the pixel points with unsatisfactory brightness values in each layer, it ensures that the obtained layers are clear enough, and at the same time improves the accuracy of determining the identity of the person inside the target area based on the face image, enhancing the accuracy of camera recognition.

[0127] Embodiment 5:

[0128] Based on the above Embodiment 1, this embodiment provides a high-precision face recognition camera, and the feature extraction module includes:

[0129] A segmentation model training unit for constructing an image main body segmentation model and obtaining historical face images, where the historical face images are at least two;

[0130] A segmentation model training unit for determining the main body area and the background area in the historical face images, determining the segmentation line between the main body area and the background area based on a preset segmentation annotation method, and performing segmentation annotation on the segmentation line;

[0131] The segmentation model training unit is further configured to determine the image orientation of the historical face image based on the segmentation annotation result, and determine a segmentation scheme for the historical face image based on the picture orientation;

[0132] The segmentation model training unit is further configured to train the image body segmentation model based on the segmentation scheme to obtain a target image body segmentation model;

[0133] The body segmentation unit is configured to perform segmentation processing on the reserved target face image based on the target image body segmentation model to obtain a target face body image in the reserved target face image, and place the reserved target face body image in a preset two-dimensional coordinate system;

[0134] The face size determination unit is configured to determine the image size information of the face in the reserved target face body image based on the preset two-dimensional coordinate system, and simultaneously determine the scaling ratio between the reserved target face image and the actual object size;

[0135] The face size determination unit is further configured to determine the actual size information of the face in the reserved target face body image based on the scaling ratio, where the actual size information includes the length, width of the face, and the face shape information of the face;

[0136] The feature extraction unit is configured to obtain a historical face image dataset, and train a convolutional neural network based on the historical face image dataset to obtain a face recognition model;

[0137] The feature extraction unit is configured to process the target face body image based on the face recognition model, extract target feature points of the face in the target face body image, and determine the position information of the target feature points on the face based on the actual size information of the face;

[0138] The feature information determination unit is configured to determine the feature information in the reserved target face image based on the actual size information of the face and the position information of the target feature points on the face.

[0139] In this embodiment, the historical face image refers to a face image captured by a preset camera before, and an image for accurately recognizing the historical face image.

[0140] In this embodiment, the main body area refers to an area image in the captured face image that only contains a face.

[0141] In this embodiment, the background area refers to other partial images in the captured face image except the face image.

[0142] In this embodiment, the preset segmentation annotation method is set in advance and is used to determine the segmentation line between the face region and other non-face regions in the face image.

[0143] In this embodiment, the segmentation annotation can be to mark the current segmentation line with a red line or other methods to facilitate accurate segmentation of the face image.

[0144] In this embodiment, the image direction refers to the direction information of the captured face image. Among them, the direction information includes vertical pictures and horizontal pictures, and different segmentation methods are used for different directions.

[0145] In this embodiment, the target image body segmentation model refers to the segmentation model obtained after training the image body segmentation model through historical face image pairs.

[0146] In this embodiment, the preset two-dimensional coordinate system is set in advance and is used to determine the size information of the face region in the face image, so as to accurately confirm the size of the actual face.

[0147] In this embodiment, the scaling ratio refers to the scaling multiple performed by the camera to ensure that the captured face image is clear enough to accurately see the person's face due to the difference in the distance between the person and the camera.

[0148] In this embodiment, the face shape information can be a round face, an oval face, a square face, etc.

[0149] In this embodiment, the historical face image dataset refers to the images that have been accurately recognized by the camera and the face information data contained in the images.

[0150] In this embodiment, the target feature points can be the information with distinct features on the face. For example, they can be moles, etc.

[0151] The beneficial effects of the above technical solution are as follows: By segmenting the obtained target face image, only the face region is analyzed, improving the efficiency of image analysis. Secondly, by placing the segmented face image in the coordinate system, the size information of the face is accurately obtained, and the specific position of the feature points in the face is determined according to the face size information, ensuring that the extracted feature information is accurate and effective, providing convenience for improving the high-precision recognition of the camera and reducing the recognition error rate.

[0152] Embodiment 6:

[0153] Based on the above Embodiment 1, this embodiment provides a high-precision face recognition camera, as Figure 3 shown, the identity determination module includes:

[0154] A data acquisition unit, configured to acquire the feature information, and respectively match the feature information with the reference image information stored in the preset face image library to obtain a matching degree between the feature information and the reference image information, where the reference image information is not unique;

[0155] An optimal reference image information determination unit, configured to sort the matching degrees in a descending order, and determine the reference image information with the largest matching degree as the target reference image information based on the sorting result;

[0156] An identity information determination unit, configured to determine an image identifier of the target reference image information, and match the target identity information from a preset identity information storage library based on the image identifier to obtain the identity information of the target person in the retained target face image.

[0157] In this embodiment, the matching degree refers to the similarity between the feature information and the reference image information. The larger the matching degree, the more similar the two are.

[0158] In this embodiment, the target reference image information refers to the reference image consistent with the feature information, that is, the face image stored in the preset face image library.

[0159] In this embodiment, the image identifier is used to mark the identity of the current image, is a label used to distinguish it from other images, and one reference image corresponds to one identifier.

[0160] In this embodiment, the preset identity information storage library is set in advance and stores the personal identity information corresponding to each reference image information in the preset face image library.

[0161] In this embodiment, the target identity information refers to the identity information of the person corresponding to the captured face image, and the identity information may be the name, age, height, etc. of the person.

[0162] The beneficial effects of the above technical solution are: By matching the feature information with the reference information image in the preset face image library, it is possible to quickly and accurately confirm the identity information of the person included in the captured face image, improving the accuracy and efficiency of camera recognition.

[0163] Embodiment 7:

[0164] Based on the above Embodiment 6, this embodiment provides a high-precision face recognition camera, characterized in that the optimal reference image information determination unit includes:

[0165] An identity verification subunit, configured to acquire the matching degree between the feature information and the reference image information, and compare the matching degree with a preset matching degree;

[0166] If the matching degree is greater than or equal to the preset matching degree, it is determined that the reference image information matched with the feature information is qualified, and the identity information corresponding to the reference image information is determined;

[0167] Otherwise, it is determined that there is no reference image information in the preset face image library that matches the feature information, and the collected face image is transmitted to the management terminal and an alarm notification is sent.

[0168] In this embodiment, the preset matching degree is set in advance and is used to measure whether the similarity between the feature information and the reference image information can meet the consistent requirements. The preset matching degree can be adjusted, for example, it can be 99%.

[0169] The beneficial effects of the above technical solution are: by verifying the matching degree, it is ensured that the analyzed identity information of the person is accurate enough, and when there is no current person's identity information, an alarm reminder is sent to the management terminal, which improves the practicability of high-precision identification of the camera and also enhances the security.

[0170] Embodiment 8:

[0171] Based on the above Embodiment 1, this embodiment provides a high-precision face recognition camera, and the image acquisition module includes:

[0172] An image acquisition unit, configured to perform multiple face image acquisitions on the person inside the target area based on a preset camera, obtain a face image, and determine the size information of the face image;

[0173] A calculation unit, configured to calculate the resolution of the preset camera according to the size information of the face image, and calculate the qualified rate of the face image collected by the preset camera based on the resolution. The specific steps include:

[0174] A first calculation unit, configured to calculate the resolution of the preset camera according to the following formula:

[0175]

[0176] where α represents the resolution of the preset camera; μ represents the error factor, and its value range is (0.05, 0.15); β represents the total number of pixel points in the face image; δ represents the pixel coefficient, and its value is 1*10 6 ; G represents the ratio of the horizontal pixel dimension to the vertical pixel dimension in the face image; H represents the ratio of the vertical pixel dimension to the horizontal pixel dimension in the face image; L represents the length value of the face image; K represents the width value of the face image;

[0177] A second calculation unit for calculating the qualification rate of the face images captured by the preset camera according to the following formula:

[0178]

[0179] where η represents the qualification rate of the face images captured by the preset camera, and its value range is (0, 1); ψ represents the number of face images that meet the preset clarity requirement in the multiple face image captures; λ represents the total number of face images captured in the multiple face image captures;

[0180] A comparison unit for comparing the calculated qualification rate with a preset qualification rate;

[0181] If the qualification rate is greater than or equal to the preset qualification rate, it is determined that the capture of the face images of the internal characters in the target area by the preset camera meets the preset requirements;

[0182] Otherwise, it is determined that the capture of the face images of the internal characters in the target area by the preset camera does not meet the preset requirements, and the resolution of the preset camera is adjusted until the qualification rate is greater than or equal to the preset qualification rate.

[0183] In this embodiment, the pixel coefficient refers to determining the pixel value of the entire image by multiplying the number of pixel points.

[0184] In this embodiment, the preset clarity requirement is set in advance, which is used to measure whether the clarity of the face image can meet the requirements for face image analysis and can be adjusted.

[0185] In this embodiment, the preset qualification rate is set in advance, which is used to measure the working performance of the preset camera.

[0186] In this embodiment, the preset requirements are set in advance.

[0187] The beneficial effects of the above technical solution are: by calculating the resolution of the preset camera and calculating the qualification rate of the target face images captured by the preset camera according to the resolution. When calculating the resolution, by calculating the resolution of the camera according to the pixels of the camera and the size information of the image to be captured, it is ensured that the calculated resolution is accurate and reliable. When calculating the qualification rate, by dividing the number of qualified images by the total number of images, and being affected by the camera resolution at the same time, the calculated qualification rate is accurate and reliable, improving the accuracy of image recognition, and also facilitating the real-time adjustment of the camera resolution according to the calculation results, providing a guarantee for achieving high-precision recognition of the camera.

[0188] Embodiment 9:

[0189] Based on the embodiment, the detection module includes:

[0190] A prediction unit, configured to determine the current position of the corresponding target face image for the user based on a position conversion rule, and at the same time, obtain first images at the previous and subsequent moments corresponding to the current moment, and predict the current motion state of the corresponding user;

[0191] A conversion unit, configured to construct a connection line between the positions corresponding to the previous and subsequent moments and the current position, obtain a position matrix W, and at the same time, based on the current motion state, obtain a motion matrix Y, where W and Y are matrices with 1 row and 3 columns;

[0192] An operation unit, configured to calculate the current satisfaction value Z of the corresponding target face image according to the position matrix W and the motion matrix Y;

[0193]

[0194] where, Y T represents the transpose of the motion matrix Y; σ1 represents the position standard deviation obtained based on each position element in the position matrix W; σ2 represents the motion standard deviation obtained based on each motion element in the motion matrix Y;

[0195] When the current satisfaction value Z is less than or equal to a preset satisfaction value, it is determined that the corresponding target face image meets the shooting requirement;

[0196] Otherwise, it is determined that the corresponding target face image does not meet the shooting requirement.

[0197] In this embodiment, the position conversion rule is, for example, to determine the position through the first image of the current position of the user.

[0198] In this embodiment, for example: the position matrix W = (010), the motion matrix Y = (110), and the corresponding σ1 2 =(0 - 0) 2 +(1 - 0) 2 +(0 - 0) 2 = 1, σ2 2 =(1 - 1) 2 +(1 - 1) 2 +(0 - 1) 2 = 1, and the corresponding

[0199] where, the matrix is a value determined by three moments: the previous moment, the current moment, and the next moment respectively.

[0200] The beneficial effects of the above technical solution are as follows: By obtaining two matrices composed of the current positions and states at three adjacent moments, and then calculating and obtaining the corresponding current satisfaction value according to the formula, and determining whether the shooting requirements are met through judgment, which facilitates subsequent high-precision recognition of images and provides an indirect basis for ensuring the high precision of the camera.

[0201] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these changes and modifications.

Claims

1. A high-precision face recognition camera, characterized in that, Including: An image acquisition module, configured to acquire a face image within a target area, and perform image enhancement processing on the face image to obtain a target face image; A detection module, configured to determine the current position and current motion state of the user corresponding to each target face image after image enhancement processing for the same user; Based on the current position and current motion state, determine whether each target face image meets the shooting requirements. If it meets the requirements, input the corresponding target face image into a general image detection model for detection. If it does not meet the requirements, adjust the acquisition accuracy of the camera for re-acquisition and processing, and detect the target face image after re-acquisition and processing; When the detection result meets the detection requirements, retain the corresponding target face image; A feature extraction module, configured to analyze the retained target face image and extract feature information in the retained target face image; An identity determination module, configured to match the feature information with reference image information stored in a preset face image library to determine the identity information of the target person in the retained target face image, and complete high-precision recognition of the face; Among them, the feature extraction module includes: A segmentation model training unit, configured to construct an image main body segmentation model and obtain historical face images, where the historical face images are at least two; The segmentation model training unit, configured to determine the main body area and background area in the historical face images, and determine the segmentation line between the main body area and the background area based on a preset segmentation annotation method, and perform segmentation annotation on the segmentation line; The segmentation model training unit is further configured to determine the image direction of the historical face images based on the segmentation annotation result, and determine a segmentation scheme for the historical face images based on the image direction; The segmentation model training unit is further configured to train the image main body segmentation model based on the segmentation scheme to obtain a target image main body segmentation model; A main body segmentation unit, configured to perform segmentation processing on the retained target face image based on the target image main body segmentation model to obtain a target face main body image in the retained target face image, and place the retained target face main body image in a preset two-dimensional coordinate system; A face size determination unit, configured to determine the image size information of the face in the retained target face main body image based on the preset two-dimensional coordinate system, and at the same time determine the scaling ratio between the retained target face image and the actual object size; The face size determination unit is further configured to determine the actual size information of the face in the retained target face main body image based on the scaling ratio, where the actual size information includes the length, width of the face, and the face shape information of the face; A feature extraction unit, configured to obtain a historical face image data set, and train a convolutional neural network based on the historical face image data set to obtain a face recognition model; A feature extraction unit, configured to process the target face main image based on the face recognition model, extract target feature points of the face in the target face main image, and determine the position information of the target feature points on the face based on the actual size information of the face; A feature information determination unit, configured to determine the feature information in the retained target face image based on the actual size information of the face and the position information of the target feature points on the face; Among them, the image direction refers to the direction information of the captured face image. Among them, the direction information includes vertical pictures and horizontal pictures, and different segmentation methods are used for different directions.

2. The high-precision face recognition camera according to claim 1, wherein An image acquisition module, including: A person detection unit, configured to monitor the target area in real time based on a preset camera, and determine whether there is a person in the target area based on the monitoring result; An image acquisition unit, configured to, when there is a person in the target area, control the preset camera to acquire images of the faces of the people in the target area from different shooting angles, and obtain multiple face images of the people; An optimal image determination unit, configured to screen the multiple face images based on a preset screening requirement to determine an optimal face image, where the optimal face image is a frontal face image of a person in the target area; A monitoring unit, configured to continue to monitor the target area in real time when there is no person in the target area.

3. The high-precision face recognition camera according to claim 2, wherein, The image acquisition unit includes: A distance determination subunit, configured to obtain the distance value between the person in the target area and the preset camera, and compare the distance value with a first preset threshold and a second preset threshold; If the distance value is greater than the first preset threshold, control the preset camera to magnify the image of the person in the target area; If the distance value is greater than or equal to the second preset threshold and less than or equal to the first preset threshold, determine that the distance value between the person in the target area and the preset camera is appropriate, and acquire a face image of the person in the target area based on the distance value; If the distance value is less than the second preset threshold, control the preset camera to blur the area other than the face image of the person in the target area to obtain a final face image.

4. The high-precision face recognition camera according to claim 1, characterized in that, The image acquisition module further includes: An image segmentation unit, configured to acquire the face image and determine the brightness value of each pixel point in the face image; The image segmentation unit, configured to segment the face image into N layers based on the brightness value, and respectively determine the brightness value of the pixel points in each layer; A brightness enhancement unit, configured to compare the pixel point brightness value with a preset brightness value, determine the target pixel points in each layer whose pixel point brightness value is less than the preset brightness value, and enhance the brightness value of the target pixel points in each layer based on a preset processing method to obtain N initial enhanced layers; A noise removal unit, configured to perform noise removal processing on the N initial enhanced layers to obtain N target enhanced layers; An image fusion unit for fusing the N target enhanced image layers to obtain a target face image.

5. The high-precision face recognition camera according to claim 1, characterized in that, An identity determination module, comprising: A data acquisition unit for acquiring the feature information and respectively matching the feature information with the reference image information stored in the preset face image library to obtain the matching degree between the feature information and the reference image information, wherein the reference image information is not unique; An optimal reference image information determination unit for sorting the matching degrees in a descending order and determining the reference image information with the largest matching degree as the target reference image information based on the sorting result; An identity information determination unit for determining the image identifier of the target reference image information and matching the target identity information from the preset identity information storage library based on the image identifier to obtain the identity information of the target person in the retained target face image.

6. The high-precision face recognition camera according to claim 5, characterized in that, The optimal reference image information determination unit, comprising: An identity verification sub-unit for acquiring the matching degree between the feature information and the reference image information and comparing the matching degree with a preset matching degree; If the matching degree is greater than or equal to the preset matching degree, determining that the reference image information matched to the feature information is qualified and determining the identity information corresponding to the reference image information; Otherwise, determining that there is no reference image information in the preset face image library that matches the feature information, and transmitting the captured face image to the management terminal and giving an alarm notification.

7. The high-precision face recognition camera according to claim 1, characterized in that, An image acquisition module, comprising: An image acquisition unit for collecting multiple face images of the people inside the target area based on a preset camera to obtain face images and determining the size information of the face images; A calculation unit for calculating the resolution of the preset camera according to the size information of the face images and calculating the qualified rate of the face images captured by the preset camera based on the resolution. The specific steps include: A first calculation unit for calculating the resolution of the preset camera according to the following formula: ； Among them, represents the resolution of the preset camera; represents the error factor, and its value range is (0.05, 0.15); represents the total number of pixel points in the face image; represents the pixel coefficient, and its value is ; represents the ratio of the horizontal pixel dimension to the vertical pixel dimension in the face image; represents the ratio of the vertical pixel dimension to the horizontal pixel dimension in the face image; represents the length value of the face image; represents the width value of the face image; A second calculation unit for calculating the qualified rate of the face images captured by the preset camera according to the following formula: ； Among them, represents the qualified rate of the face images collected by the preset camera, and its value range is (0, 1); represents the number of face images that meet the preset clarity requirements in the multiple face image collections; represents the total number of face images collected in the multiple face image collections; A comparison unit for comparing the calculated qualified rate with a preset qualified rate; If the qualified rate is greater than or equal to the preset qualified rate, determining that the acquisition of the face images of the people inside the target area by the preset camera meets the preset requirements; Otherwise, determining that the acquisition of the face images of the people inside the target area by the preset camera does not meet the preset requirements, and adjusting the resolution of the preset camera until the qualified rate is greater than or equal to the preset qualified rate.

8. The high-precision face recognition camera according to claim 1, wherein The detection module, comprising: A prediction unit for determining the current position of the user corresponding to the target face image based on a position conversion rule, and at the same time, acquiring the first images at the previous and subsequent moments corresponding to the current moment to predict the current motion state of the corresponding user; A conversion unit for constructing a connection line between the corresponding positions at the previous and subsequent moments and the current position, acquiring a position matrix W, and at the same time, acquiring a motion matrix Y based on the current motion state, wherein W and Y are matrices with 1 row and 3 columns; An operation unit is used to calculate the current satisfaction value Z of the corresponding target face image according to the position matrix W and the motion matrix Y; ； Among them, represents the transpose of the motion matrix Y; represents the position standard deviation obtained based on each position element in the position matrix W; 2 represents the motion standard deviation obtained based on each motion element in the motion matrix Y; When the current satisfaction value Z is less than or equal to the preset satisfaction value, it is determined that the corresponding target face image meets the shooting requirements; Otherwise, it is determined that the corresponding target face image does not meet the shooting requirements.

Images