Face recognition method for an advertisement screen based on autonomous learning
A face recognition technology on the screen, applied in character and pattern recognition, data processing applications, marketing, etc., can solve the problems of not actually watching advertisements, difficult for advertisers to put into use on a large scale, and affecting the effect of advertisements.
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0051] A face recognition method for advertising screens based on autonomous learning, mainly comprising the following steps:
[0052] Step S1: Initial classification and merging:
[0053] Track the continuous frames taken. If the face center distance between two consecutive frames is less than a given threshold, it is determined that the two faces belong to the same person; classify and mark the face image frames and extract feature values; The tracked face features are merged under the same face ID;
[0054] Step S2: Secondary merge:
[0055] Learn the sample eigenvalues accumulated in the cycle time, perform secondary screening on the eigenvalues belonging to the same face angle interval, and merge the eigenvalues of the same face ID; use the face recognition algorithm to judge the similarity of face features Degree, if the similarity is greater than the set value, the recognized face features will be merged under the same face ID.
[0056] The continuous frames ca...
Embodiment 2
[0059] This embodiment is optimized on the basis of embodiment 1, such as Figure 5 As shown, in the step S1, if there is a human face that meets the standard in the picture, the face feature value is obtained, and if the center distance between the acquired picture and the human face in the previous frame is less than a given threshold, the human face in the previous frame is used. Face ID, otherwise generate a new face ID; then obtain the angle classification features of the face in the picture. If there are multiple people in the picture acquired in step S1, the picture is divided into multiple areas according to the number of faces, and a single person is processed in each area.
[0060] Such as Image 6 As shown, in the step S2, the feature list under the same face angle classification is taken out, and the elements in the list are compared in pairs. If the similarity is greater than a given threshold, then it is judged whether any of the two face IDs has a virtual paren...
Embodiment 3
[0064] This embodiment is optimized on the basis of embodiment 1 or 2, as image 3 , Figure 4 As shown, in the step S1, when the angle of the human face is in the front face, left side face, right side face, head buried, and head up, the face image frame is classified and marked and the feature value is extracted; the front face is It means that the horizontal angle interval of the human face is (-10, 10) and the pitch angle interval is (-12, 6); the left face refers to that the horizontal angle interval of the human face is (-30, 10) and the pitch angle interval is It is (-12, 6); the right side face refers to the horizontal angle interval of the human face is (10, 30) and the pitch angle interval is (-12, 6); the buried head refers to the horizontal angle interval of the human face is (-10, 10) and the pitch angle interval is (6, 30); the head-up means that the horizontal angle interval of the face is (-10, 10) and the pitch angle interval is (-30, -12).
[0065] The coun...
PUM
Abstract
Description
Claims
Application Information
- R&D Engineer
- R&D Manager
- IP Professional
- Industry Leading Data Capabilities
- Powerful AI technology
- Patent DNA Extraction
Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.
© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com