Method for facilitating identity recognition of old people
By employing a layered processing and multi-dimensional verification approach, this technology addresses the issues of aging and occlusion of facial feature points in the elderly, achieving efficient and accurate identification of senior citizens and resolving the recognition difficulties caused by distortion and occlusion of facial feature points.
Patent Information
- Application Number
- CN202511658709.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-13
- Publication Date
- 2026-02-17
AI Technical Summary
Existing technologies for identifying the elderly suffer from low accuracy and high false positive rates due to aging physiological features and facial occlusion.
By dividing facial images into Class I and Class II, effective recognition regions are extracted respectively, center points are determined, and reference axes and symmetry axes are established. Sub-regions are divided horizontally, feature point matching and cleaning are performed, and the proportion of effective feature points is statistically analyzed to complete the recognition.
It improved the accuracy of elderly identification, reduced the probability of false positives, and achieved fast and efficient identification.
Smart Images
Figure CN121545200A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of identity recognition technology, and in particular to a method that facilitates identity recognition for the elderly. Background Technology
[0002] In recent years, cases of elderly people going missing have occurred frequently. The main rescue method is to extract images from cameras in the area where the elderly went missing and perform facial recognition to identify them. However, for the elderly, the natural aging of physiological characteristics can lead to distortion of the recognition target points. For example, loose facial skin, increased wrinkles, and the appearance of age spots can change the position and shape of key feature points, affecting the algorithm's ability to recognize faces. Facial fat loss or swelling, as well as the obstruction of masks, can reduce the number of feature points that can be collected in core recognition areas such as the nose. Changes in the eyes, such as cloudy eyes and drooping eyelids, can affect the accuracy of recognizing eye-related feature points, thereby increasing the probability of misjudgment.
[0003] Therefore, there is an urgent need for a method that facilitates the identification of the elderly and can identify them with high accuracy. Summary of the Invention
[0004] In view of this, this application provides a method to facilitate identification of the elderly, in order to address the shortcomings of existing technologies.
[0005] The first aspect of this application provides a method for facilitating identification of elderly individuals, including: Acquire facial images of elderly people and classify them into Class I facial images or Class II facial images based on the effective acquisition area. Class I facial images contain part of the facial area, and Class II facial images contain the entire facial area. Based on the facial contour and the contour of the region corresponding to the feature points, the effective recognition region in the first type of facial image is extracted and denoted as the first recognition region; and based on the facial contour, the effective recognition region in the second type of facial image is extracted and denoted as the second recognition region. Obtain the center point of the first recognition area, denoted as the first center point, and obtain the center point of the second recognition area, denoted as the second center point; The horizontal line passing through the first center point or the second center point is designated as the reference axis, and the vertical line passing through the first center point or the second center point is designated as the axis of symmetry. Based on the reference axis, the corresponding identification area is horizontally divided into multiple sub-regions of equal width, and a preset number of feature points are symmetrically selected in the regions on both sides of each sub-region about the axis of symmetry. In the first or second identification region, feature points that are symmetrical about the axis of symmetry in each sub-region are matched with each other, and the feature point pairs that fail to match are cleaned. When any feature point is designated as the first feature point and successfully matches the corresponding feature point of the elderly person's face in the database, and all feature points in the adjacent sub-regions that are closest to the first feature point are successfully matched, the first feature point is designated as a valid feature point. All feature points after cleaning are traversed, the proportion of valid feature points is calculated, and the identification of the elderly person is completed.
[0006] In one possible implementation of the first aspect, classifying the facial image into a first-class facial image or a second-class facial image based on the effective acquisition area includes: Determine whether the effective acquisition area completely covers the elderly person's face. If not, classify the acquired facial image into the first type of facial image; if so, classify the acquired facial image into the second type of facial image.
[0007] In one possible implementation of the first aspect, extracting the effective recognition region from the type of facial image based on the facial contour and the contour of the region corresponding to the feature points includes: The facial images of the aforementioned type are preprocessed, including grayscale conversion, noise reduction, and contrast enhancement. After preprocessing, facial images are subjected to facial contour detection, and multiple facial contour detection points are extracted. Obtain the smallest bounding rectangle containing all facial contour detection points, denoted as the first smallest bounding rectangle, and perform a cropping operation on the facial image of the first type based on the first smallest bounding rectangle; Threshold comparison is performed on a class of facial images after one cropping to extract the boundary contours of the occluded areas; The first minimum bounding rectangle is cut by a straight line that runs through the boundary contour of the occluded area and its upper vertex to obtain the recognition rectangle. Based on the recognition rectangle, a second cropping is performed on a type of facial image that has been cropped once; Edge adjustment is performed on a type of facial image after secondary cropping to obtain the corresponding effective recognition area.
[0008] In one possible implementation of the first aspect, extracting the effective recognition region from the two types of facial images based on facial contours includes: The two types of facial images are preprocessed, including grayscale conversion, noise reduction, and contrast enhancement. After preprocessing, the second type of facial images are used for facial contour detection, and multiple facial contour detection points are extracted. Obtain the minimum bounding rectangle containing all facial contour detection points, and crop the two types of facial images based on the minimum bounding rectangle; Edge adjustment is performed on the cropped Class II facial images to obtain the corresponding effective recognition area.
[0009] In one possible implementation of the first aspect, obtaining the center point of the first identification region includes: In the first recognition area, two feature points near the root of the nose of the elderly person's eyes are obtained and recorded as the second feature points, and the midpoint between the two second feature points is taken as the center point of the first recognition area.
[0010] In one possible implementation of the first aspect, obtaining the center point of the second identification region includes: In the second recognition area, the two outermost feature points of the elderly person's nasal wing are obtained and recorded as the third feature points, and the midpoint between the two third feature points is taken as the center point of the second recognition area.
[0011] In one possible implementation of the first aspect, when any feature point, designated as the first feature point, successfully matches a corresponding feature point on an elderly person's face in the database, and all feature points in adjacent sub-regions closest to the first feature point also successfully match, the first feature point is designated as a valid feature point, including: When the sub-region where the first feature point is located is at the top or bottom of the corresponding recognition region, if the first feature point is successfully matched with the corresponding feature point of the elderly face in the database, and the single feature point closest to the first feature point in the adjacent sub-region is successfully matched, the first feature point is recorded as a valid feature point. When the sub-region where the first feature point is located is in the middle of the corresponding recognition region, if the first feature point is successfully matched with the corresponding feature point of the elderly face in the database, and the two feature points closest to the first feature point in the adjacent sub-region are successfully matched, the first feature point is recorded as a valid feature point.
[0012] In one possible implementation of the first aspect, the proportion of statistically valid feature points includes: The proportion of effective feature points. The sum of effective feature points, This represents the sum of the feature points after cleaning.
[0013] In one possible implementation of the first aspect, identifying the elderly person includes: Determine whether the proportion of valid feature points exceeds a set threshold. If yes, the elderly person's identity is successfully identified; otherwise, the elderly person's identity is deemed to have failed to be identified.
[0014] One possible implementation of the first aspect includes: When matching between any two feature points, the similarity between the feature descriptors corresponding to the two feature points is calculated, and it is determined whether the similarity exceeds a set value. If it does, the corresponding feature point is determined to be successfully matched; otherwise, the corresponding feature point is determined to be unmatched.
[0015] Its beneficial effects are as follows: This invention discloses a method for facilitating elderly identification. First, facial images of the elderly are acquired and categorized into two types based on whether the effective acquisition area completely covers the face. Then, the effective recognition areas of both types of facial images are extracted, and the center point of each effective recognition area is determined. A reference axis and a symmetry axis are established based on the center point. Based on the constructed reference axis and symmetry axis, the effective recognition area is horizontally divided into multiple equal-width sub-regions, and feature points are symmetrically selected. The symmetrically selected feature points on both sides are first matched against each other. Feature points that fail to match are cleaned up. Then, the correlation between the spatial positions of multiple feature points is used to determine the effective feature points, reducing the interference of a single distorted feature point on the overall recognition result. Finally, the proportion of effective feature points is calculated to complete the identification of the elderly. This invention solves the problem of low facial recognition accuracy caused by physiological aging or facial occlusion in the elderly, enabling rapid and efficient identification of the elderly. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0017] Figure 1 This is a schematic diagram of a method for facilitating identity verification for the elderly, provided in an embodiment of this application. Detailed Implementation
[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0019] In this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.
[0020] Example In existing technologies, when identifying elderly individuals, the natural aging of physiological characteristics can lead to distortion of the identification target points. For example, loose facial skin, increased wrinkles, and the appearance of age spots can change the position and shape of key feature points, affecting the algorithm's ability to recognize faces. Facial fat loss or swelling, as well as the obstruction caused by masks, can reduce the number of feature points that can be collected in core recognition areas such as the nose. Changes in the eyes, such as cloudy eyes and drooping eyelids, can affect the accuracy of identifying eye-related feature points, thereby increasing the probability of misjudgment.
[0021] Therefore, this application provides a method to facilitate identification for the elderly, such as... Figure 1 As shown, it includes: Acquire facial images of elderly people and classify them into Class I facial images or Class II facial images based on the effective acquisition area. Class I facial images contain part of the facial area, and Class II facial images contain the entire facial area. Based on the facial contour and the contour of the region corresponding to the feature points, the effective recognition region in the first type of facial image is extracted and denoted as the first recognition region; and based on the facial contour, the effective recognition region in the second type of facial image is extracted and denoted as the second recognition region. Obtain the center point of the first recognition area, denoted as the first center point, and obtain the center point of the second recognition area, denoted as the second center point; The horizontal line passing through the first center point or the second center point is designated as the reference axis, and the vertical line passing through the first center point or the second center point is designated as the axis of symmetry. Based on the reference axis, the corresponding identification area is horizontally divided into multiple sub-regions of equal width, and a preset number of feature points are symmetrically selected in the regions on both sides of each sub-region about the axis of symmetry. In the first or second identification region, feature points that are symmetrical about the axis of symmetry in each sub-region are matched with each other, and the feature point pairs that fail to match are cleaned. When any feature point is designated as the first feature point and successfully matches the corresponding feature point of the elderly person's face in the database, and all feature points in the adjacent sub-regions that are closest to the first feature point are successfully matched, the first feature point is designated as a valid feature point. All feature points after cleaning are traversed, the proportion of valid feature points is calculated, and the identification of the elderly person is completed.
[0022] This embodiment provides a method to facilitate identification for the elderly, and its working principle is as follows: Facial image classification: Based on whether the effective acquisition area completely covers the elderly person's face, the acquired facial images are divided into two categories: Category I facial images, where the effective acquisition area does not completely cover the elderly person's face but includes part of the facial area, and Category II facial images, where the effective acquisition area completely covers the elderly person's face and includes the entire facial area.
[0023] Effective region extraction: Different processes are used to extract effective recognition regions for different types of facial images. Specifically, the first recognition region is extracted for one type of facial image as follows: Preprocessing involves converting a type of facial image to grayscale, denoising, and enhancing contrast. Contour detection: Perform facial contour detection on the preprocessed image and extract multiple facial contour detection points; One cropping operation is performed to obtain the smallest bounding rectangle containing all facial contour detection points, and the image is cropped based on this rectangle. Extract the occlusion boundary by performing a threshold comparison on the image after one cropping to extract the boundary contour of the occlusion area; Secondary cropping involves cropping the first minimum bounding rectangle using a straight line that passes through the vertices of the occluded region's boundary contour to obtain a recognition rectangle. The image after the first cropping is then cropped again based on the recognition rectangle. Edge adjustment: The edges of the image after secondary cropping are adjusted to eliminate the influence of noise such as hair, and the first recognition area is obtained.
[0024] Extracting the second recognition region from the second type of facial images, specifically: Preprocessing involves grayscale conversion, noise reduction, and contrast enhancement of the two types of facial images; Contour detection: Perform facial contour detection on the preprocessed image and extract multiple facial contour detection points; Cropping involves obtaining the smallest bounding rectangle containing all facial contour detection points and cropping the image based on this rectangle. Edge adjustment: The cropped image is adjusted to obtain the second recognition region.
[0025] Identification of the center point of the region: In the first recognition area, two feature points near the root of the nose of the elderly person's eyes are obtained, and the midpoint between these two feature points is taken as the center point of the first recognition area. The second center point is obtained by acquiring the two feature points on the outermost side of the elderly person's nose in the second recognition area, and the midpoint between these two feature points is taken as the center point of the second recognition area.
[0026] Axis establishment and region division: The horizontal line passing through the first center point or the second center point is denoted as the reference axis, and the vertical line passing through the first center point or the second center point is denoted as the axis of symmetry. Based on the reference axis, the corresponding recognition area is horizontally divided into multiple sub-regions of equal width, and a preset number of feature points are symmetrically selected in the regions on both sides of each sub-region about the axis of symmetry.
[0027] Feature point processing: Feature point cleaning involves matching feature points that are symmetrical about the axis of symmetry in each sub-region within the first or second recognition region, and cleaning up the feature point pairs that fail to match. Valid feature point determination: When the sub-region containing the first feature point (any feature point to be determined) is located at the top or bottom of the corresponding recognition region, if the first feature point successfully matches the corresponding feature point of the elderly face in the database, and the single feature point closest to the first feature point in the adjacent sub-region successfully matches, the first feature point is recorded as a valid feature point; when the sub-region containing the first feature point is located in the middle of the corresponding recognition region, if the first feature point successfully matches the corresponding feature point of the elderly face in the database, and the two feature points closest to the first feature point in the adjacent sub-region successfully match, the first feature point is recorded as a valid feature point.
[0028] Identity verification result determination: The percentage of effective feature points is calculated using the following formula: The proportion of effective feature points. The sum of effective feature points, This represents the sum of the feature points after cleaning. Determine whether the proportion of valid feature points exceeds a set threshold. If yes, the elderly person's identity is successfully identified; otherwise, the elderly person's identity is deemed to have failed to be identified.
[0029] It should be noted that when matching between any two feature points, the similarity between the feature descriptors corresponding to the two feature points is calculated, and it is determined whether the similarity exceeds a set value. If it does, the corresponding feature point is considered to have matched successfully; otherwise, the corresponding feature point is considered to have failed to match.
[0030] The reason this embodiment can quickly and efficiently identify elderly individuals lies in its design of a layered processing, multi-dimensional verification, and targeted optimization technical solution that addresses the core pain points of facial recognition for the elderly (physiological aging leading to feature point distortion, facial occlusion reducing effective features, and changes in the eyes causing recognition errors). This is achieved through the following key methods: Classify and process data based on facial integrity for precise adaptation to occlusion scenarios: Elderly people often have their faces obscured by wearing masks, and traditional methods are weak in recognizing certain facial images. This embodiment solves this problem by using image classification and differential extraction of effective regions.
[0031] First, determine whether the effective acquisition area completely covers the face, and divide the image into two categories to avoid feature omissions caused by using a uniform standard to process occluded and unoccluded images. For category one images, additional steps are added, including threshold comparison to extract occlusion boundaries, cropping rectangles along occlusion vertices, and secondary cropping, to ensure that the effective recognition area only contains complete local parts that can be used for recognition, reducing the interference of occlusion on feature extraction. For category two images, cropping is performed directly based on the face outline to retain the core area of the entire face and avoid redundant background affecting the accuracy of feature point localization.
[0032] Optimize the effective identification area and enhance the focus on key features: The facial contours of elderly people may be blurred due to loose skin and loss of fat. Traditional methods are prone to extracting non-core areas. This embodiment improves the accuracy of area extraction through multi-step preprocessing and precise contour positioning.
[0033] Preprocessing eliminates noise interference, and all images are converted to grayscale, denoised, and contrast enhanced to solve the problem of difficult feature point localization on the face of the elderly due to age spots or uneven skin color. Contour-based minimum bounding box cropping detects facial contour points and finds the minimum bounding box that contains all facial contour points, ensuring that the extracted area exactly covers the facial core and reducing the impact of invalid areas on subsequent feature extraction. Edge adjustment optimizes the integrity of the cropped area by adjusting its edges to ensure clear boundaries of the effectively identifiable area and eliminate interference from irrelevant noise.
[0034] Scientifically define the center point and axis to ensure symmetry in the selection of feature points: In one type of facial image, the midpoint between two feature points near the root of the nose is taken as the center point. Since the eyes are usually a relatively intact area in occluded scenes, and eye features (such as eye distance) are relatively less affected by aging, taking the center point can ensure that subsequent region division is carried out around the core stable features. The second type of facial image uses the midpoint between the two outermost feature points of the nasal wing as the center point. The position of the nasal wing is relatively stable (less affected by skin laxity) and is located in the center of the face, which can evenly divide the entire face area and avoid the area division deviation caused by facial edge aging and deformation. With the center point as the core, a reference axis is set horizontally (to ensure that the sub-regions are of equal width horizontally) and a symmetry axis is set vertically (to ensure that the feature points are symmetrical left and right). Feature points are symmetrically selected on both sides of the symmetry axis of each sub-region. The reason for this design is to reduce feature distortion caused by unilateral aging (such as deeper wrinkles on one side) by comparing the left and right feature points, while increasing the correlation of feature points and providing a basis for subsequent matching verification.
[0035] Feature point cleaning and association verification to remove distorted features: Facial feature points in the elderly are prone to shape deformation (such as wrinkles covering feature points) or positional shift (such as sagging skin causing drooping corners of the eyes) due to aging. Traditional methods of single matching are prone to misjudgment. This embodiment uses a two-layer screening mechanism to ensure the quality of effective feature points.
[0036] The first layer of symmetry matching and cleaning removes invalid feature point pairs: In each sub-region, symmetrical feature points on both sides of the axis of symmetry are matched with each other (e.g., the features of the left and right corners of the eyes are compared). If the matching fails (the similarity does not reach the set value), the pair of feature points is cleaned up. This step can remove unilaterally distorted feature points caused by unilateral aging or acquisition noise (e.g., the left corner of the eye is blurred due to occlusion and cannot be matched with the symmetrical point on the right), reducing the interference of erroneous features on subsequent recognition. The second layer of correlation verification determines the validity of feature points and enhances matching reliability: different verification rules are set for the differences in the location of sub-regions. If the feature point is at the top or bottom of the recognition area (such as the forehead and chin, where aging deformation is more obvious), it needs to match the database and the nearest feature point in the adjacent sub-region to be considered valid. If the feature point is in the middle of the recognition area (such as the eye and nose areas, where the features are more stable), it needs to match the database and the nearest two feature points in the adjacent sub-region to be considered valid. This differentiated rule avoids the problem of too few valid feature points caused by strict verification of unstable edge areas, and reduces the impact of aging-induced distortion of a single feature point by associating multiple feature points (rather than a single feature point), thereby improving the reliability of valid feature points.
[0037] Based on the determination of the proportion of effective feature points, the dependence on a single feature is reduced: Traditional methods often use a fixed number of feature points as the criterion for judgment. However, due to occlusion or aging, the number of effective feature points on the face of the elderly is reduced, which can easily lead to misjudgment and recognition failure due to insufficient number. This embodiment achieves more flexible and accurate result judgment by statistically analyzing the proportion of effective feature points.
[0038] The percentage of effective feature points is calculated quantitatively. For example, if only 50 feature points are retained after screening due to occlusion, aging, and symmetry of an elderly person's face, only 80% of them need to be effective feature points to determine successful recognition. The threshold is flexibly adapted to elderly scenarios. The percentage threshold can be adjusted according to the overall aging level of the elderly population and the accuracy of the data collection device, balancing accuracy and inclusiveness, and reducing misjudgment of severely aged elderly people caused by a uniform threshold.
[0039] Unified feature point matching standards ensure consistency across different scenarios: The facial features of the elderly vary under different collection environments. Traditional methods are prone to errors due to inconsistent matching standards. This embodiment ensures matching consistency by standardizing similarity.
[0040] All feature point matching uses feature descriptor similarity as the core standard (e.g., extracting feature descriptors using algorithms such as SIFT and ORB, and calculating Euclidean distance or cosine similarity), and sets a uniform similarity threshold. Regardless of the collection scenario, a match is considered successful as long as the similarity exceeds the threshold, avoiding matching deviations caused by manually set different standards. At the same time, feature descriptors can capture the feature details of the elderly's face, further improving matching accuracy.
[0041] This embodiment optimizes the entire process from image processing, region extraction, feature selection, matching verification, and result determination, taking into account the physiological characteristics and recognition pain points of the elderly's face. Through three core logics—precisely focusing on effective information, verifying feature reliability from multiple dimensions, and flexibly adapting to elderly scenarios—it ultimately achieves fast, efficient, and highly accurate identification of the elderly.
[0042] In some embodiments, classifying the facial images into Class I or Class II facial images based on the effective acquisition area includes: Determine whether the effective acquisition area completely covers the elderly person's face. If not, classify the acquired facial image into the first type of facial image; if so, classify the acquired facial image into the second type of facial image.
[0043] In some embodiments, extracting the effective recognition region from the type of facial image based on the facial contour and the contour of the region corresponding to the feature points includes: The facial images of the aforementioned type are preprocessed, including grayscale conversion, noise reduction, and contrast enhancement. After preprocessing, facial images are subjected to facial contour detection, and multiple facial contour detection points are extracted. Obtain the smallest bounding rectangle containing all facial contour detection points, denoted as the first smallest bounding rectangle, and perform a cropping operation on the facial image of the first type based on the first smallest bounding rectangle; Threshold comparison is performed on a class of facial images after one cropping to extract the boundary contours of the occluded areas; The first minimum bounding rectangle is cut by a straight line that runs through the boundary contour of the occluded area and its upper vertex to obtain the recognition rectangle. Based on the recognition rectangle, a second cropping is performed on a type of facial image that has been cropped once; Edge adjustment is performed on a type of facial image after secondary cropping to obtain the corresponding effective recognition area.
[0044] In some embodiments, extracting the effective recognition region from the two types of facial images based on facial contours includes: The two types of facial images are preprocessed, including grayscale conversion, noise reduction, and contrast enhancement. After preprocessing, the second type of facial images are used for facial contour detection, and multiple facial contour detection points are extracted. Obtain the minimum bounding rectangle containing all facial contour detection points, and crop the two types of facial images based on the minimum bounding rectangle; Edge adjustment is performed on the cropped Class II facial images to obtain the corresponding effective recognition area.
[0045] In some embodiments, obtaining the center point of the first identification region includes: In the first recognition area, two feature points near the root of the nose of the elderly person's eyes are obtained and recorded as the second feature points, and the midpoint between the two second feature points is taken as the center point of the first recognition area.
[0046] In some embodiments, obtaining the center point of the second identification region includes: In the second recognition area, the two outermost feature points of the elderly person's nasal wing are obtained and recorded as the third feature points, and the midpoint between the two third feature points is taken as the center point of the second recognition area.
[0047] In some embodiments, when any feature point, designated as the first feature point, successfully matches a corresponding feature point on an elderly person's face in the database, and all feature points in adjacent sub-regions closest to the first feature point also successfully match, the first feature point is designated as a valid feature point, including: When the sub-region where the first feature point is located is at the top or bottom of the corresponding recognition region, if the first feature point is successfully matched with the corresponding feature point of the elderly face in the database, and the single feature point closest to the first feature point in the adjacent sub-region is successfully matched, the first feature point is recorded as a valid feature point. When the sub-region where the first feature point is located is in the middle of the corresponding recognition region, if the first feature point is successfully matched with the corresponding feature point of the elderly face in the database, and the two feature points closest to the first feature point in the adjacent sub-region are successfully matched, the first feature point is recorded as a valid feature point.
[0048] In some embodiments, the percentage of statistically valid feature points includes: The proportion of effective feature points. The sum of effective feature points, This represents the sum of the feature points after cleaning.
[0049] In some embodiments, identifying an elderly person includes: Determine whether the proportion of valid feature points exceeds a set threshold. If yes, the elderly person's identity is successfully identified; otherwise, the elderly person's identity is deemed to have failed to be identified.
[0050] In some embodiments, including: When matching between any two feature points, the similarity between the feature descriptors corresponding to the two feature points is calculated, and it is determined whether the similarity exceeds a set value. If it does, the corresponding feature point is determined to be successfully matched; otherwise, the corresponding feature point is determined to be unmatched.
[0051] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computing software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0052] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.
[0053] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.
Claims
1. A method for facilitating identity recognition of an elderly person, characterized by, The method comprises the following steps: acquiring a face image of the old person, and dividing the face image into a first type of face image or a second type of face image based on an effective acquisition region, wherein the first type of face image contains a partial face region, and the second type of face image contains a complete face region; extracting an effective recognition region in the first type of face image based on a face contour and a contour of a region corresponding to a feature point, and recording the effective recognition region as a first recognition region, and extracting an effective recognition region in the second type of face image based on the face contour, and recording the effective recognition region as a second recognition region; acquiring a center point of the first recognition region, and recording the center point as a first center point, and acquiring a center point of the second recognition region, and recording the center point as a second center point; taking a transverse line passing through the first center point or the second center point as a reference axis, and taking a longitudinal line passing through the first center point or the second center point as a symmetry axis; dividing the corresponding recognition region into a plurality of equal-width sub-regions in the transverse direction based on the reference axis, and selecting a preset number of feature points on both sides of the symmetry axis in each sub-region symmetrically; in the first recognition region or the second recognition region, matching the feature points symmetrically about the symmetry axis in each sub-region with each other, and cleaning up the feature point pairs that fail to match; when any feature point is recorded as a first feature point and the corresponding feature point in the database matches the face of the old person successfully, and all the feature points in the adjacent sub-regions closest to the first feature point are matched successfully, the first feature point is recorded as an effective feature point, all the cleaned feature points are traversed, the proportion of the effective feature points is counted, and the identity recognition of the old person is completed.
2. The method of claim 1, wherein, The method for dividing the face image into the first type of face image or the second type of face image based on the effective acquisition region comprises the following steps: determining whether the effective acquisition region completely covers the face of the old person, if not, dividing the acquired face image into the first type of face image, and if yes, dividing the acquired face image into the second type of face image.
3. The method of claim 1, wherein, The method for extracting the effective recognition region in the first type of face image based on the face contour and the contour of the region corresponding to the feature point comprises the following steps: preprocessing the first type of face image, including grayscale processing, noise removal and contrast enhancement processing; detecting the face contour of the first type of face image after the preprocessing, and extracting a plurality of face contour detection points; acquiring a minimum circumscribed rectangle containing all the face contour detection points, and recording the minimum circumscribed rectangle as a first minimum circumscribed rectangle, and performing first cutting on the first type of face image based on the first minimum circumscribed rectangle; performing threshold comparison on the first type of face image after the first cutting, and extracting the boundary contour of the occlusion region; cutting the first minimum circumscribed rectangle by a straight line passing through the top point of the boundary contour of the occlusion region, and obtaining a recognition rectangle; performing second cutting on the first type of face image after the first cutting based on the recognition rectangle; performing edge adjustment on the first type of face image after the second cutting, and obtaining the corresponding effective recognition region.
4. The method of claim 1, wherein, The method for extracting the effective recognition region in the second type of face image based on the face contour comprises the following steps: preprocessing the second type of face image, including grayscale processing, noise removal and contrast enhancement processing; The preprocessed second type face image is subjected to face contour detection to extract a plurality of face contour detection points; A minimum circumscribed rectangle frame containing all the face contour detection points is obtained, and the second type face image is cropped based on the minimum circumscribed rectangle frame; The cropped second type face image is subjected to edge adjustment to obtain a corresponding effective recognition region.
5. The method of claim 1, wherein, Obtaining the center point of the first recognition region includes: Two feature points near the root of the nose of the old person in the first recognition region are obtained and recorded as second feature points, and a midpoint between the two second feature points is taken as the center point of the first recognition region.
6. The method of claim 1, wherein, Obtaining the center point of the second recognition region includes: Two feature points at the outermost side of the alae nasi of the old person in the second recognition region are obtained and recorded as third feature points, and a midpoint between the two third feature points is taken as the center point of the second recognition region.
7. The method of claim 1, wherein the method is characterized by, When any feature point is recorded as a first feature point and the corresponding feature point of the face of the old person in the database is successfully matched, and all feature points in the adjacent sub-region closest to the first feature point are successfully matched, the first feature point is recorded as an effective feature point. When the sub-region where the first feature point is located is at the uppermost end or the lowermost end of the corresponding recognition region, if the first feature point is successfully matched with the corresponding feature point of the face of the old person in the database, and a single feature point in the adjacent sub-region closest to the first feature point is successfully matched, the first feature point is recorded as an effective feature point. When the sub-region where the first feature point is located is in the middle of the corresponding recognition region, if the first feature point is successfully matched with the corresponding feature point of the face of the old person in the database, and two feature points in the adjacent sub-region closest to the first feature point are successfully matched, the first feature point is recorded as an effective feature point.
8. The method of claim 1, wherein, Statistical proportion of effective feature points includes: a proportion of effective feature points, a total of effective feature points, a total of feature points after cleaning.
9. The method of claim 1, wherein, Completing the identity recognition of the old person includes: Determining whether the proportion of effective feature points exceeds a set threshold value, if yes, determining that the identity recognition of the old person is successful, and if no, determining that the identity recognition of the old person fails.
10. The method of claim 1-9, wherein, Including: When any two feature points are matched, the similarity between the feature descriptors corresponding to the two feature points is calculated, and it is determined whether the similarity exceeds a set value, if yes, it is determined that the corresponding feature points are successfully matched, and if no, it is determined that the corresponding feature points are unsuccessfully matched.