Face recognition method and system based on image analysis

Abstract

The invention discloses a face recognition method and system based on image analysis, and the method comprises the steps: determining an image set, and carrying out the preprocessing of the image set; analyzing the preprocessed image set, and determining parameter data of the shooting equipment; performing parameter adjustment on the image acquisition equipment according to the parameter data of the shooting equipment; performing feature collection on the preprocessed image to obtain reference features and reference feature positions; acquiring people in the image set in real time to obtain a to-be-verified image; performing feature collection on the to-be-verified image to obtain to-be-verified features and to-be-verified feature positions; and through similarity calculation, respectively comparing the reference features with the to-be-verified features and comparing the reference feature positions with the to-be-verified feature positions, and judging whether the features are the same face or not. According to the invention, a large amount of sample information is prevented from occupying storage space, consumption of computing resources is reduced, and fluency of face recognition is greatly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of face recognition, and in particular to a face recognition method and system based on image analysis. Background Art

[0002] Face recognition is a hot research topic in computer pattern recognition and biometric identification technology. It usually uses a camera or a camera to capture images or video streams containing faces, and automatically detects, tracks and identifies faces in the images.

[0003] Compared with biometric technologies such as fingerprint recognition, palm print recognition, and iris recognition, face recognition has the characteristics of being non-invasive, easy to capture, and accepted. However, the accuracy of face recognition will be affected by many factors, such as the posture of the camera (front or side), light (backlight, day or night), weather (rainy days, foggy days), and obstructions (hair, glasses, beards). Among them, the factor that has the greatest impact on the accuracy is the long vertical hair between the eyebrows and the corners of the eyes. How to accurately and quickly extract and identify facial information is an urgent problem to be solved in the landing and application of face recognition technology. Existing common processing methods for face recognition include wavelet analysis, histograms, PCA, etc., all of which require a large amount of sample information. Summary of the invention

[0004] In view of the above problems, the present invention is proposed to provide a face recognition method and system based on image analysis that overcomes the above problems or at least partially solves the above problems.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A face recognition method based on image analysis, the method comprising the following steps:

[0007] S1, determining an image set and preprocessing the image set;

[0008] S2, analyzing the preprocessed image set and determining parameter data of the shooting device;

[0009] S3, adjusting the parameters of the image acquisition device according to the parameter data of the shooting device, so that the parameter data of the image acquisition device is consistent with the parameter data of the shooting device;

[0010] S4. Perform feature acquisition on the preprocessed image by using a convolutional neural network algorithm to obtain reference features and reference feature positions, where the reference features include reference feature shapes and reference feature sizes;

[0011] S5. Using an image acquisition device to acquire real-time images of people in the image set to obtain images to be verified;

[0012] S6. Collect features of the image to be verified by using a convolutional neural network algorithm to obtain features to be verified and positions of features to be verified. The features to be verified include shapes of features to be verified and sizes of features to be verified.

[0013] S7. By similarity calculation, the reference feature and the feature to be verified, and the reference feature position and the feature position to be verified are compared to determine whether they are the same face.

[0014] Optionally, the method further includes: step S8, optimizing step S7 according to the judgment result;

[0015] Optimizing step S7 includes: when the face recognition result is incorrect, if the feature shape and the feature shape to be verified show that they are not the same face, then adjusting and optimizing the set first threshold; if the feature size and the feature size to be verified show that they are not the same face, then adjusting and optimizing the set second threshold.

[0016] Optionally, in step S1, preprocessing the image set includes removing interference from noise in the image, deleting invalid areas in the image, and enhancing features of the image.

[0017] Optionally, removing interference from noise in the image includes:

[0018] Use the mean of the local area to smooth the image;

[0019] Remove salt and pepper noise by taking the median of the local area;

[0020] Smooth the image using a Gaussian function.

[0021] Optionally, deleting invalid areas in the image includes deleting areas in the image that do not contain facial features by cropping the image.

[0022] Optionally, features of the enhanced image include:

[0023] Improve the contrast of the image to make it clearer;

[0024] Increase or decrease the brightness of the image;

[0025] Expands the range of gray levels in an image to enhance detail.

[0026] Optionally, in step S2, analyzing the preprocessed image includes reading EXIF ​​data of the preprocessed image.

[0027] Optionally, in step S4, collecting features of the preprocessed image by using a convolutional neural network algorithm includes:

[0028] Region segmentation: perform region segmentation on the preprocessed image and record the segmentation method and region segmentation size;

[0029] Regional feature extraction: extracting features from each region through a convolutional neural network algorithm, obtaining regional features, analyzing and measuring the regional features, and determining reference features, wherein the reference features include reference feature shapes and reference feature sizes;

[0030] Mark the feature position, construct a coordinate system for the area, determine the position of the reference feature in the coordinate system, and record the coordinate system construction method.

[0031] Optionally, in step S7:

[0032] The reference features and features to be verified are compared including:

[0033] Calculate the overlap between the reference feature shape and the feature shape to be verified, determine the shape overlap weight, compare the shape overlap weight with the set first threshold, if the shape overlap weight is greater than the set first threshold, the feature shape to be verified is different from the reference feature shape, if the shape overlap weight is less than or equal to the set first threshold, the feature shape to be verified is the same as the reference feature shape;

[0034] The reference feature size and the feature size to be verified are calculated for overlap, and a size overlap weight is determined; the size overlap weight is compared with a set second threshold value, and if the size overlap weight is greater than the set second threshold value, the feature size to be verified is different from the reference feature size; if the size overlap weight is less than or equal to the set second threshold value, the feature size to be verified is the same as the reference feature size;

[0035] The comparison between the reference feature position and the feature position to be verified includes:

[0036] If the reference feature position and the feature position to be verified are the same, it is necessary to compare the reference feature with the feature to be verified and the reference feature size with the feature size to determine whether they are the same face;

[0037] If the reference feature position and the feature position to be verified are not the same, they are not the same face.

[0038] The present invention also provides a face recognition system based on image analysis, which is used to execute any of the above-mentioned face recognition methods based on image analysis, including:

[0039] Image preprocessing module, used for image and real-time image denoising, cropping and enhancement to ensure that subsequent similarity calculations are more accurate;

[0040] An image acquisition module, used for real-time image acquisition of people in the image set;

[0041] A feature acquisition module is used to acquire features of preprocessed images and real-time images through a convolutional neural network algorithm;

[0042] A similarity calculation module is used to calculate the similarity between the reference feature and the feature to be verified, and between the reference feature position and the feature position to be verified, to determine whether they are the same face;

[0043] An optimization module is used to optimize the similarity calculation module according to the judgment result of the similarity calculation module to make face recognition more accurate;

[0044] The data storage and update module is used to add and delete facial information.

[0045] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0046] 1. The present invention extracts the features of facial images through a convolutional neural network algorithm and combines it with similarity calculation to achieve efficient and accurate recognition of faces. It is suitable for a variety of scenarios such as smart parks, office clocking in, identity recognition, security monitoring, etc. It has been widely used and greatly improves user safety.

[0047] 2. The present invention combines the convolutional neural network algorithm with the similarity calculation. Without using a large amount of sample information, the similarity calculation result can be used to fine-tune the set threshold, thereby avoiding the use of a large amount of sample information to occupy storage space, while reducing the consumption of computing resources and greatly improving the fluency of face recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic diagram of a face recognition method based on image analysis provided in an embodiment of the present application;

[0049] Figure 2 A flow chart of a face recognition system based on image analysis provided in an embodiment of the present application. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0051] Embodiment 1:

[0052] See also Figure 1This embodiment provides a face recognition method based on image analysis, the method comprising the following steps:

[0053] S1. Determine an image set and preprocess the image set.

[0054] Preprocessing of image sets includes removing noise interference in images, deleting invalid areas in images and enhancing image features to improve image quality and subsequent processing effects.

[0055] The interference of removing noise from the image includes:

[0056] Smooth the image using the mean of the local region.

[0057] Remove salt and pepper noise by taking the median of the local area.

[0058] Smooth the image using a Gaussian function.

[0059] Deleting invalid areas in the image includes deleting areas in the image that do not contain facial features by cropping the image.

[0060] Features of enhanced images include:

[0061] Histogram equalization improves the contrast of the image to make the image clearer.

[0062] Brightness adjustment, increase or decrease the brightness of the image.

[0063] Contrast stretching expands the range of gray levels in an image to enhance detail.

[0064] S2. Analyze the preprocessed image set and determine parameter data of the shooting device.

[0065] Analyzing the pre-processed image includes reading EXIF ​​data of the pre-processed image and determining parameter data of the shooting device.

[0066] S3. Adjust the parameters of the image acquisition device according to the parameter data of the shooting device, so that the parameter data of the image acquisition device is consistent with the parameter data of the shooting device.

[0067] S4. Perform feature collection on the preprocessed image through a convolutional neural network algorithm to obtain reference features and reference feature positions. The reference features include reference feature shapes and reference feature sizes.

[0068] The feature acquisition of the preprocessed image by the convolutional neural network algorithm includes:

[0069] Region segmentation: perform region segmentation on the preprocessed image and record the segmentation method and region segmentation size.

[0070] Regional feature extraction, extracting features from each region through a convolutional neural network algorithm, obtaining regional features, analyzing and measuring the regional features, and determining reference features, wherein the reference features include reference feature shapes and reference feature sizes.

[0071] Mark the feature position, construct a coordinate system for the area, determine the position of the reference feature in the coordinate system, and record the coordinate system construction method.

[0072] S5. Use an image acquisition device to acquire real-time images of people in the image set to obtain images to be verified.

[0073] S6. Perform feature collection on the image to be verified by using a convolutional neural network algorithm to obtain features to be verified and positions of features to be verified. The features to be verified include shapes of features to be verified and sizes of features to be verified.

[0074] S7. By similarity calculation, the reference feature and the feature to be verified, and the reference feature position and the feature position to be verified are compared to determine whether they are the same face.

[0075] The reference features and features to be verified are compared including:

[0076] The overlap between the reference feature shape and the feature shape to be verified is calculated to determine the shape overlap weight, and the shape overlap weight is compared with the set first threshold. If the shape overlap weight is greater than the set first threshold, the feature shape to be verified is different from the reference feature shape. If the shape overlap weight is less than or equal to the set first threshold, the feature shape to be verified is the same as the reference feature shape.

[0077] The overlap between the reference feature size and the feature size to be verified is calculated to determine the size overlap weight; the size overlap weight is compared with the set second threshold value. If the size overlap weight is greater than the set second threshold value, the feature size to be verified is different from the reference feature size. If the size overlap weight is less than or equal to the set second threshold value, the feature size to be verified is the same as the reference feature size.

[0078] The comparison between the reference feature position and the feature position to be verified includes:

[0079] If the reference feature position and the feature position to be verified are the same, it is necessary to compare the reference feature with the feature to be verified and the reference feature size with the feature size to determine whether they are the same face.

[0080] If the reference feature position and the feature position to be verified are not the same, they are not the same face.

[0081] S8. Optimize step S7 according to the judgment result.

[0082] Optimizing step S7 includes: when the face recognition result is incorrect, if the feature shape and the feature shape to be verified show that they are not the same face, then adjusting and optimizing the set first threshold; if the feature size and the feature size to be verified show that they are not the same face, then adjusting and optimizing the set second threshold.

[0083] This embodiment extracts the features of facial images through a convolutional neural network algorithm and combines it with similarity calculation to achieve efficient and accurate recognition of faces. It is suitable for various scenarios such as smart parks, office clocking in, identity recognition, security monitoring, etc. It has been widely used and greatly improved user safety.

[0084] This embodiment combines the convolutional neural network algorithm with the similarity calculation. Without using a large amount of sample information, the similarity calculation result can be used to fine-tune the set threshold, thereby avoiding the use of a large amount of sample information to occupy storage space, while reducing the consumption of computing resources and greatly improving the fluency of face recognition.

[0085] Embodiment 2:

[0086] As attached Figure 2 As shown, Embodiment 2 discloses a face recognition system based on image analysis, which is used to execute the above-mentioned face recognition method based on image analysis, including:

[0087] Image preprocessing module, used for image and real-time image denoising, cropping and enhancement to ensure that subsequent similarity calculation is more accurate.

[0088] The image acquisition module is used to acquire real-time images of people in the image set.

[0089] The feature acquisition module is used to acquire features of preprocessed images and real-time images through a convolutional neural network algorithm.

[0090] The similarity calculation module is used to calculate the similarity between the reference feature and the feature to be verified, and the reference feature position and the feature position to be verified to determine whether they are the same face.

[0091] The optimization module is used to optimize the similarity calculation module according to the judgment result of the similarity calculation module to make face recognition more accurate.

[0092] The data storage and update module is used to add and delete facial information.

[0093] Practical application scenarios:

[0094] Smart park scenario: In a smart park environment, the data storage and update module will regularly collect employee resignation or employment information in the smart park, and regularly update facial information, such as "delete resignation facial information" and "add employment facial information".

[0095] Security monitoring scenario: In security monitoring, the data storage and update module will collect facial images of sensitive users, such as "objects in need of protection", etc. When the similarity calculation module determines the identity of the sensitive user, it can provide more accurate and personalized service solutions.

[0096] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. A face recognition method based on image analysis, characterized in that: The method comprises the following steps: S1, determining an image set and preprocessing the image set; S2, analyzing the preprocessed image set and determining parameter data of the shooting device; S3, adjusting the parameters of the image acquisition device according to the parameter data of the shooting device, so that the parameter data of the image acquisition device is consistent with the parameter data of the shooting device; S4. Perform feature acquisition on the preprocessed image by using a convolutional neural network algorithm to obtain reference features and reference feature positions, where the reference features include reference feature shapes and reference feature sizes; S5. Using an image acquisition device to acquire real-time images of people in the image set to obtain images to be verified; S6. Perform feature collection on the image to be verified by using a convolutional neural network algorithm to obtain features to be verified and positions of features to be verified. The features to be verified include shapes of features to be verified and sizes of features to be verified. S7. By similarity calculation, the reference feature and the feature to be verified, and the reference feature position and the feature position to be verified are compared to determine whether they are the same face.

2. A face recognition method based on image analysis as claimed in claim 1, characterized in that: The method further includes: step S8, optimizing step S7 according to the judgment result; Optimizing step S7 includes: when the face recognition result is incorrect, if the feature shape and the feature shape to be verified show that they are not the same face, then adjusting and optimizing the set first threshold; if the feature size and the feature size to be verified show that they are not the same face, then adjusting and optimizing the set second threshold.

3. A face recognition method based on image analysis as claimed in claim 1, characterized in that: In step S1, preprocessing the image set includes removing the interference of noise in the image, deleting the invalid area in the image and enhancing the characteristics of the image.

4. A face recognition method based on image analysis as claimed in claim 3, characterized in that: The interference of removing noise from the image includes: Use the mean of the local area to smooth the image; Remove salt and pepper noise by taking the median of the local area; Smooth the image using a Gaussian function.

5. A face recognition method based on image analysis as claimed in claim 3, characterized in that: Deleting invalid areas in the image includes deleting areas in the image that do not contain facial features by cropping the image.

6. A face recognition method based on image analysis as claimed in claim 3, characterized in that: Features of enhanced images include: Improve the contrast of the image to make it clearer; Increase or decrease the brightness of the image; Expands the range of gray levels in an image to enhance detail.

7. A face recognition method based on image analysis as claimed in claim 1, characterized in that: In step S2, analyzing the preprocessed image includes reading EXIF ​​data of the preprocessed image.

8. The face recognition method based on image analysis as claimed in claim 1, characterized in that: In step S4, feature acquisition of the preprocessed image by a convolutional neural network algorithm includes: Region segmentation: perform region segmentation on the preprocessed image and record the segmentation method and region segmentation size; Regional feature extraction: extracting features from each region through a convolutional neural network algorithm, obtaining regional features, analyzing and measuring the regional features, and determining reference features, wherein the reference features include reference feature shapes and reference feature sizes; Mark the feature position, construct a coordinate system for the area, determine the position of the reference feature in the coordinate system, and record the coordinate system construction method.

9. The face recognition method based on image analysis as claimed in claim 1, characterized in that: In step S7: The reference features and features to be verified are compared including: Calculate the overlap between the reference feature shape and the feature shape to be verified, determine the shape overlap weight, compare the shape overlap weight with the set first threshold, if the shape overlap weight is greater than the set first threshold, the feature shape to be verified is different from the reference feature shape, if the shape overlap weight is less than or equal to the set first threshold, the feature shape to be verified is the same as the reference feature shape; The reference feature size and the feature size to be verified are calculated for overlap, and a size overlap weight is determined; the size overlap weight is compared with a set second threshold value, and if the size overlap weight is greater than the set second threshold value, the feature size to be verified is different from the reference feature size; if the size overlap weight is less than or equal to the set second threshold value, the feature size to be verified is the same as the reference feature size; The comparison between the reference feature position and the feature position to be verified includes: If the reference feature position and the feature position to be verified are the same, it is necessary to compare the reference feature with the feature to be verified and the reference feature size with the feature size to determine whether they are the same face; If the reference feature position and the feature position to be verified are not the same, they are not the same face.

10. A face recognition system based on image analysis, used to execute the face recognition method based on image analysis according to any one of claims 1 to 9, characterized in that: include: Image preprocessing module, used for image and real-time image denoising, cropping and enhancement to ensure that subsequent similarity calculations are more accurate; An image acquisition module, used for real-time image acquisition of people in the image set; A feature acquisition module is used to acquire features of preprocessed images and real-time images through a convolutional neural network algorithm; A similarity calculation module is used to calculate the similarity between the reference feature and the feature to be verified, and between the reference feature position and the feature position to be verified, to determine whether they are the same face; An optimization module is used to optimize the similarity calculation module according to the judgment result of the similarity calculation module to make face recognition more accurate; The data storage and update module is used to add and delete facial information.

Images