人脸识别方法、装置、电子设备及计算机可读存储介质
By verifying facial features in multiple image frames and combining information from the previous frame to confirm the target face, the accuracy problem of traditional face recognition technology in complex environments is solved, achieving highly accurate face recognition and tracking.
CN117115892BActive Publication Date: 2026-07-17BEIJING BAIDU NETCOM SCI & TECH CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING BAIDU NETCOM SCI & TECH CO LTD
- Filing Date
- 2023-08-31
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Traditional facial recognition technology lacks accuracy in complex environments, making it difficult to effectively identify and track facial features.
Method used
By acquiring multiple image frames, a target detection network is used to identify face prediction boxes, and the predictions are verified across multiple image frames based on the reference target face features. The accuracy of the target face image is confirmed by combining the information from the previous image frame.
Benefits of technology
It improves the accuracy of facial recognition, reduces false positives and false negatives, and ensures high-accuracy facial tracking in dynamic environments.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN117115892B_ABST
Abstract
本公开提供了一种人脸识别方法，涉及人工智能技术领域，具体为计算机视觉、图像处理、深度学习等技术领域，可应用于智慧城市等场景。具体实现方案为：获取多个图像帧，针对每一图像帧，确定当前的图像帧中包含的待测人脸图像；针对每一图像帧，基于参考目标人脸特征，从当前的图像帧中包含的待测人脸图像中确定出目标待测人脸图像；根据当前的图像帧中的目标待测人脸图像的图像信息和前一图像帧确定出的目标人脸图像的图像信息验证目标待测人脸图像，并在验证通过的情况下，确定目标待测人脸图像为当前的图像帧的目标人脸图像。通过这种方式，能够识别准确度较高的人脸识别。
Need to check novelty before this filing date? Find Prior Art