A kind of video object classification method and device

Abstract

This application discloses a method and device for classifying video objects. In the method, key frames in the target video are extracted, and the image features of the key frames are extracted using a preset convolutional neural network to obtain multi-layer feature maps. The feature maps of each layer are The included local features are weighted and fused to obtain the fused features, the fused features are subjected to dimensionality reduction processing to obtain the dimensionality reduction results, and the dimensionality reduction results are clustered to obtain the classification results of all pixels contained in the key frame. After the fusion features are obtained, the dimensionality reduction processing is performed on the fusion features to obtain the dimensionality reduction results, which avoids the use of the fc-lstm structure for classification, and can effectively learn the time and space dependencies between video data, so that the dimensionality reduction results carry It has the characteristics of space and timing, and it can't produce effective classification for complex scenes with complex movements of multiple objects. At the same time, based on the algorithm characteristics of the clustering method, the pixel-level segmentation of a single frame image can be achieved to obtain the classification result.

Description

technical field [0001] The present application relates to the field of video object classification, in particular to a video object classification method and device. Background technique [0002] At present, it is very important to classify the objects in the video scene. In order to realize the classification of the objects in the video data, it is necessary to be able to link the context on the basis of processing the spatial dimension information of a single frame of pictures. However, the classification method in the prior art can easily lose the spatial correlation in a single frame of pictures because it can only map one-dimensional vectors, and it cannot be connected with the context, so it can only be used for simple and strong linear features. However, it cannot produce effective classification for complex scenes with complex movements of multiple objects. Contents of the invention [0003] The present application provides a video object classification method and...

AI Technical Summary

Problems solved by technology

[0003] This application provides a video object classification method and device to achieve effective classification for complex scenes with complex movements of multiple objects

Images