基于卷积神经网络的视频流隐藏信息检测方法

为解决现有视频流隐藏信息检测中,人工检测特征设计难度不断加大的问题,提出一种基于卷积神经网络的视频流隐藏信息检测方法。在神经网络中构建残差学习单元,避免深层次卷积神经网络在训练时的梯度消失,利用深层神经网络自动从数据中挖掘检测特征,在此基础上引入量化截断操作,增加检测模型多样性,提升检测性能。使用FFmpeg与x264编码标准CIF序列生成的视频进行实验,实验结果表明,该方法相比现有方法具有更高的检测准确率。

Video stream hidden information detection algorithm based on convolutional neural network

To solve the problem that the artificial video feature design is more difficult to detect in the hidden information detection of existing video streams,a method of video stream hidden information detection based on convolutional neural network was proposed.A residual learning unit in the neural network was constructed to avoid the gradient disappearance of the deep convolutional neural network during training,the neural network was used to automatically extract the detection features from the data.On this basis,the quantitative truncation operation was introduced to increase the model diversity,which improved detection performance.Experiments were carried out using the video generated by FFmpeg and x264 encoding standard CIF sequences.Experimental results show that the proposed method has higher detection accuracy than the existing methods.