【摘要】 为解决传统视频侵权检测精度低、存证难等问题，提出一种基于区块链存证的多模态视频侵权检测技术。结合自适应阈值与结构相似性指数优化关键帧提取算法，并通过卷积神经网络与循环神经网络实现多模态特征表征。利用区块链与星际文件系统构建版权存证机制，提出基于加权融合的相似度计算模型以提升检测精度。设计智能合约实现链上自动化检测与处置。实验结果表明，该方法在UCF101数据集上检测准确率达95%，具有显著的鲁棒性与实时性。更多还原

【Abstract】 To address the issues of low accuracy and difficult evidence preservation in traditional video infringement detection, a multimodal video infringement detection technology based on blockchain evidence preservation was proposed. The keyframe extraction algorithm was optimized by combining adaptive thresholding and structural similarity index(SSIM), and multimodal feature representation was achieved by fusing convolutional neural networks(CNN) and recurrent neural networks(RNN). A copyright evidence preservation mechanism was constructed using blockchain and InterPlanetary file system(IPFS), and a similarity calculation model based on weighted fusion was proposed to improve detection precision. Smart contracts were designed to realize on-chain automated detection and handling. Experimental results demonstrate that the proposed method achieves an accuracy of 95% on the UCF101 dataset, exhibiting significant robustness and real-time performance.更多还原