Face and license plate desensitization method and device based on fisheye image
CN121121827APending Publication Date: 2025-12-12WUHAN KOTEI INFORMATICS
-1 Cites -1 Cited by
Patent Information
- Application Number
- CN202511427973.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-30
- Publication Date
- 2025-12-12
AI Technical Summary
Technical Problem
Existing technologies struggle to achieve high accuracy and irreversibility in efficiently and irreversibly detecting and desensitizing face and license plate regions in fisheye images.
Method used
An improved YOLOV11 neural network is used to identify face and license plate regions in fisheye images by enhancing the resolution of the backbone network and fusing neck features, and then performing irreversible pixel averaging.
Benefits of technology
It improves the accuracy of face and license plate detection and ensures that the desensitized images cannot be reversed, thus enhancing security.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN121121827A_ABST
Abstract
The invention relates to a face and license plate desensitization method and device based on a fisheye image. The method comprises the following steps: acquiring a mark data set of a face and a license plate and a fisheye image to be desensitized; based on the annotation data set, training a YOLOV11 neural network after structure optimization; the structural optimization comprises resolution enhancement of the backbone network and feature fusion enhancement of the neck; recognizing a face and a license plate in the fisheye image to be desensitized through the trained YOLOV11 neural network; and carrying out irreversible desensitization processing on the identified pixel areas of the human face and the license plate to obtain a desensitized fisheye image. According to the method, efficient and low-cost irreversible desensitization of the face and the license plate of the fisheye image is directly realized through the improved YOLOV11 neural network.
Need to check novelty before this filing date? Find Prior Art