Video semantic analysis method

Abstract

The present invention provides a video semantic analysis method. The method comprises the following steps: S1, performing preprocessing of a video training set, and constructing a sparse linear decoder; S2, adding topological property constraints to build a topology linear decoder, and performing image blocking processing of the video training set to train the topology linear decoder; S3, taking the parameters of a trained topology linear decoder as initial parameters of a convolution layer in a convolution nerve network; and S4, building a key frame set to perform fin adjustment of the convolution nerve network based on the video training set through adoption of a multifold cross validation mode, and building a general feature extractor on the base of the video data, and inputting the features extracted on the training set and a test set to support the video semantic classification in a vector machine. The model training method provided by the invention has video-type data samples capable of responding to various contents so as to improve the accuracy and the robustness of the model.

Description

technical field [0001] The invention relates to the technical field of video semantic detection, in particular to a video semantic analysis method. Background technique [0002] In order to realize the detection of video semantic concepts, the convolutional neural network model is used to extract the features of the key frame set of the video. The experiment proves that it is different from other manual design feature extraction methods. Distributed features are extracted, that is, the obtained features are data-driven and can be adapted to a wider range of fields. However, the convolutional neural network is a supervised learning model, that is, when training the convolutional neural network model, the training data set and the label corresponding to the training data set are required, and the convergence of the convolutional neural network also requires a large number of samples. Continuous iteration, for tasks such as classification and detection of massive video data, i...

AI Technical Summary

Problems solved by technology

However, the convolutional neural network is a supervised learning model, that is, when training the convolutional neural network model, the training data set and the label corresponding to the training data set are required, and the convergence of the convolutional neural network also requires a large number of samples. Continuous iteration, for tasks such as classification and detection of massive video data, it is impossible to obtain the corresponding label of each video

[0003] Aiming at using a convolutional neural network model with supervised training characteristics on video data, although the predecessors proposed an unsupervised pre-training method based on unsupervised training, it solved the problem of slow convergence of traditional convolutional neural networks; Compared with image data, video data will have the same target rotation, scaling, translation and other phenomena in the content, which requires the feature extractor used to be able to capture more complex and invariant features, so how good The extraction of features with strong invariance has become the problem to be solved

Images