摘要
为了提高包装图像被恶意篡改盗用的检测效率,开发了基于Faster R-CNN的图像篡改检测系统。利用Python语言构建逻辑架构;利用Faster R-CNN算法实现图像识别与分类;基于Flask框架、Gunicorn框架以及Nginx服务器实现网络架构实现图像传输。通过CASIA v1.0 TIDE数据集验证系统的检测效果。结果表明,与传统的ELA,NOI1和CFA1等图像篡改检测方法相比,基于Faster R-CNN的篡改检测算法的总体精度F 1和AUC值都具有明显优势,总体精度F 1值为0.344,平均AUC值为0.816。
In order to improve the detection efficiency of packaging images being maliciously tampered with and stolen,an image tampering detection system based on Faster R-CNN is developed.This paper builds logical architecture using Python language;uses Faster R-CNN algorithm to achieve image recognition and classification.Based on Flask framework,Gunicorn framework,and Nginx server,network architecture is implemented to achieve image transmission.The detection performance of the system is verified by using the CASIA v1.0 TIDE dataset.The results show that compared with traditional image tampering detection methods such as ELA,NOI1 and CFA1,the Faster R-CNN based tampering detection algorithm has significant advantages in overall accuracy F 1 and AUC values,with an overall accuracy F 1 value of 0.344 and an average AUC value of 0.816.
作者
陶瑶
Tao Yao(College of Art and Design,Maanshan Teacher′s College,Ma′anshan,Anhui 243041,China)
出处
《黑龙江工业学院学报(综合版)》
2025年第4期123-127,共5页
Journal of Heilongjiang University of Technology(Comprehensive Edition)
基金
2023年度安徽省高校自然科学研究重点项目(项目编号:2023AH052471)
安徽省高校教学研究重点项目(项目编号:2022kcsz279)。
作者简介
陶瑶,硕士,讲师,马鞍山师范高等专科学校艺术设计学院。研究方向:艺术设计。
