CN-121971037-A - Child morning check method and device

CN121971037ACN 121971037 ACN121971037 ACN 121971037ACN-121971037-A

Abstract

The application discloses a child morning inspection method and device, which are used for solving the technical problems of low efficiency and high missing inspection rate of the existing morning inspection method. The method comprises the steps of collecting a child face image, determining child identity information by means of a preset face recognition algorithm, outputting a voice prompt and an animation prompt according to a to-be-detected item, collecting a child morning detection image when a human body key point detection algorithm is used for judging that child actions are consistent with corresponding actions of the to-be-detected item, wherein the voice prompt and the animation prompt are related to the corresponding actions of the to-be-detected item, the to-be-detected item at least comprises any one of a back of a hand, a palm of the hand, an oral cavity and a mouth, detecting child health abnormality based on the child morning detection image, generating a morning detection analysis report, and storing the morning detection analysis report after binding the child identity information and the child morning detection image. According to the method, full-flow automation and multi-dimensional project detection of the child morning detection are realized, and high efficiency and low omission ratio are achieved.

Inventors

DING XIN
YAN LONG
XIA TIANYU

Assignees

神思电子技术股份有限公司

Dates

Publication Date: 20260505
Application Date: 20251225

Claims (10)

1. A method of morning inspection of a child, the method comprising: collecting a child face image, and determining child identity information by using a preset face recognition algorithm; According to a to-be-morning-detected item, outputting a voice prompt and an animation prompt, and when a human body key point detection algorithm is utilized to judge that the child motion is consistent with the corresponding motion of the to-be-morning-detected item, acquiring a child morning-detected image, wherein the voice prompt and the animation prompt are related to the corresponding motion of the to-be-morning-detected item, and the to-be-morning-detected item at least comprises any one of a back of a hand, a palm of the hand, an oral cavity and a mouth; and detecting children health abnormality based on the children morning check image, generating a morning check analysis report, binding the morning check analysis report with the child identity information and the children morning check image, and storing the morning check analysis report.
2. The method for morning check according to claim 1, wherein determining the child identity information comprises: Traversing a preset face library based on the child face image to determine child identity information, or acquiring the child identity information by reading a child identity card, wherein the child identity information at least comprises a child name; After obtaining the child identity information, the method further comprises: and broadcasting the child name.
3. The method of claim 1, wherein capturing a child morning check image comprises: positioning the positions of hands, mouth and mouth of the child through the key points of the human body; Based on the position of the hands of the child, the mouth area of the child and the position of the oral cavity of the child, capturing a back-of-hand image, a palm-of-hand image, a mouth static image and an oral cavity internal image, wherein the back-of-hand image, the palm-of-hand image, the mouth static image and the oral cavity internal image form the morning check image of the child; Skin abnormality detection is performed based on the back images, the palm images, and the mouth static images, and oral cavity abnormality detection is performed based on the oral cavity internal images.
4. A method of child morning inspection according to claim 3, wherein detecting a child health abnormality based on the child morning inspection image comprises: Preprocessing the child morning inspection image to obtain a local image, wherein the local image only comprises a child back of hand, a child palm of hand, a child mouth and/or a child oral cavity; Inputting the local image into a preset matching model to match the local image with a preset image library, and/or inputting the local image into a preset detection model to identify an abnormal position in the local image; and judging the child morning inspection result corresponding to the partial image according to the matching result and/or the detection result.
5. The method of claim 4, wherein matching the partial image with a library of pre-set images, comprises: performing feature extraction on the local image through a first feature extraction network in the preset matching model to obtain a first feature map; determining a reference image in the preset image library, and extracting features of the reference image through a second feature extraction network in the preset matching model to obtain a second feature image, wherein the first feature network and the second feature network are two weight sharing networks; And calculating the similarity of the first feature map and the second feature map through a similarity measurement network in the preset matching model so as to obtain a matching result between the second image and the preset image library.
6. The method of claim 5, wherein prior to feature extraction of the partial image by the first feature extraction network in the predetermined matching model, the method further comprises: dividing the image blocks of the local image according to the pixel sizes corresponding to the local image; Calculating effective values corresponding to each divided pixel block respectively, and sequencing the effective values, wherein the effective values are used for indicating the correlation degree of the pixel blocks and the palm of the child and/or the mouth of the child; and selecting 50% of image blocks before the effective value ordering to perform feature extraction.
7. The method of claim 4, wherein performing similarity calculation on the first feature map and the second feature map comprises: Determining the size of a search frame according to the pixel size corresponding to the first feature map, and determining an area to be matched in the first feature map based on the search frame; Taking the search frame as a search step length, traversing the second feature map from the upper left corner of the second feature map to obtain a plurality of search areas; Calculating the similarity between the region to be matched and the plurality of search regions, and determining one search region with the highest similarity as a matching region corresponding to the region to be matched; Moving the search box in the first feature map to update the region to be matched; Repeatedly executing the operation until each region to be matched in the first feature map finds a corresponding matching region in the second feature map; Summing the similarity between each region to be matched and the corresponding matching region, and taking the sum result as the similarity between the local image and the reference image; Traversing the preset image library to obtain a similarity matrix between the local image and the preset image library, wherein the similarity matrix is a row matrix, the number of elements in the similarity matrix is equal to the number of reference images in the preset image library, and the element values are used for indicating the similarity between the local image and the reference images; Determining a reference image corresponding to the maximum element value in the similarity matrix as a matching image of the local image; and determining a child morning check result corresponding to the partial image based on the child health condition corresponding to the matching image.
8. The method for early detection of children according to claim 7, wherein calculating the similarity between the region to be matched and the plurality of search regions comprises: determining feature matrixes respectively corresponding to the region to be matched and any search region; based on a preset weight, carrying out fusion processing on the feature matrix to obtain a fusion matrix; And inputting the fusion matrix into a Sigmoid activation function, and determining the similarity between the region to be matched and any search region based on determinant values corresponding to the output matrix.
9. A child morning check device, the device comprising: A processor; And a memory having stored thereon executable instructions that when executed cause the processor to perform a child morning check method according to any one of claims 1-8.
10. The child morning check equipment according to claim 9, wherein the equipment comprises a cradle head camera which automatically adjusts a shooting angle and supplements light according to the oral cavity position of the child.

Description

Child morning check method and device Technical Field The application relates to the technical field of image processing, in particular to a method and equipment for morning inspection of children. Background Children in kindergarten and primary school are older and are easy to develop aggregation diseases, such as hand-foot-and-mouth diseases, and in order to avoid the diseases, a morning check mode is generally adopted to find out sick children in advance. However, the existing morning check method mainly uses manual observation by kindergarten teachers and primary school teachers, the accuracy of the morning check result cannot be guaranteed, and the morning check is generally carried out before a class, so that a long time is required to be delayed, more manpower is required, and the working pressure of the teacher is increased. The specific disadvantages are as follows: 1) Relying on manual operation, the traditional morning check is manually completed by a health care teacher, hands, oral cavity, body temperature and the like of children need to be checked in sequence, each child takes more than 30 seconds, queuing congestion is easy to cause in peak hours of a kindergarten, and large-scale kindergarten or children groups such as more than 500 people are difficult to deal with. 2) The detection subjectivity is strong, the omission ratio is high, the manual detection depends on experience judgment, the detection is easy to be omitted for symptoms such as slight rash and early herpes, and meanwhile, when the children cry and are not matched, the parts such as the oral cavity and the hands are difficult to observe, so that the detection accuracy is reduced. 3) The method has the advantages that the method lacks standardized record and trace, paper record or simple electronic registration is difficult to correlate detection images and results, abnormal conditions cannot be traced accurately, data are scattered, and long-term trace of the child health record cannot be formed. 4) The existing equipment has single function, namely the partial semi-automatic morning inspection equipment can only measure the body temperature or single-part image, lacks face recognition identity verification and multi-part linkage acquisition, such as palm, back of hand, oral cavity and the like detection, still needs manual assistance, and cannot realize full-flow automation. Disclosure of Invention The embodiment of the application provides a method and equipment for detecting the morning hours of children, which are used for solving at least one of the technical problems. The method comprises the steps of collecting a child face image, determining child identity information by means of a preset face recognition algorithm, collecting a child morning inspection image when a child action is judged to be consistent with a corresponding action of the child morning inspection item by means of a human body key point detection algorithm according to the child morning inspection item, wherein the child morning inspection image is collected when the child action is judged to be consistent with the corresponding action of the child morning inspection item by means of the human body key point detection algorithm, the child morning inspection item at least comprises any one of a back of a hand, a palm of a hand, an oral cavity and a mouth, detecting child health abnormality based on the child morning inspection image, generating a morning inspection analysis report, and storing the morning inspection analysis report after binding the child identity information and the child morning inspection image. In one possible implementation manner of the application, the method for determining the child identity information comprises the steps of traversing a preset face library based on the child face image to determine the child identity information, or obtaining the child identity information by reading a child identity card, wherein the child identity information at least comprises a child name, and broadcasting the child name after the child identity information is obtained. In one possible implementation mode of the application, a child morning inspection image is acquired, wherein the child morning inspection image is formed by positioning a child hand position, a child mouth area and a child oral cavity position through human body key points, intercepting a hand back image, a hand palm image, a mouth static image and an oral cavity internal image based on the child hand position, the child mouth area and the child oral cavity position, wherein the hand back image, the hand palm image, the mouth static image and the oral cavity internal image form the child morning inspection image, and detecting skin abnormality based on the hand back image, the hand palm image and the mouth static image and detecting oral cavity abnormality based on the oral cavity internal image. In one possible implementation manner of the application, the detection of t