Quadruple convolutional neural network video fingerprint algorithm

Abstract

The invention provides a quadruple convolutional neural network video fingerprint algorithm. The algorithm comprises the steps of establishing a projection excitation network; constructing a quadrupleconvolutional neural network video fingerprint algorithm according to the projection excitation network; through video data selection, inputting the constructed quadruple video sequence into a quadruple convolutional neural network to carry out training and performance testing of the quadruple convolutional neural network. According to the invention, mapping from original video data to discrete binary codes can be realized in an end-to-end manner; simplified algorithm complexity, during training, tetrad loss and quantization error loss are used for jointly optimizing network parameters; on one hand, the intra-class variance is reduced and the inter-class variance is increased due to the loss of the tetrad; and on the other hand, the quantization error loss can reduce the loss of semanticsimilar information in a real value feature binarization process, the precision ratio and recall ratio of the method in the aspect of video copy detection are obviously improved, and the obtained video fingerprint can maintain relatively high robustness and uniqueness while meeting compactness.

Description

technical field [0001] The invention relates to the technical field of multimedia information security, in particular to a quadruplet convolutional neural network video fingerprint algorithm. Background technique [0002] With the popularity of the Internet, computers are undergoing a networked revolution. All kinds of multimedia information technologies related to it have sprung up like mushrooms after rain, and multimedia data can be used and disseminated quickly and conveniently. In this process, while massive video data enriches human life and increases human knowledge, some illegal content contained in it will directly damage the personal rights and interests of copyright owners during the transmission process, and seriously affect the healthy development of society. In order to increase the restraint and supervision of digital media, in recent years, the state has issued relevant laws and regulations to effectively protect video copyright and monitor video content. I...

AI Technical Summary

Problems solved by technology

[0005] Generally speaking, the existing technology has the following problems: First, most video copy detection technologies based on deep learning only use neural networks for feature learning, and traditional methods are still used for quantization of real-valued feature sequences, which makes the overall algorithm The complexity is high, which is not conducive to execution on large-scale data sets; second, although some existing end-to-end deep video fingerprint algorithms incorporate the quantization coding step into the training of the feature extraction network, whether it is Neither the supervised learning method with a single sample and label nor the unsupervised learning method composed of two-tuple or three-tuple sample pairs can make the model fully learn both robustness and uniqueness when the number of samples is limited. The video features directly affect the quality of the final generated fingerprint code; third, for the feature extraction method using the three-dimensional convolutional neural network, the feed-forward network with fewer layers and simple structure is not enough to mine complex video structures. semantic similarity information

Images