Semantic Marking Method of Video Scene Based on 3D Convolutional Neural Network

Abstract

With the rapid development of multimedia technology, video data has become an important source big data. How to identify and segment different semantic objects in video scenes accurately has become a research hotspot in the field of computer vision. On the basis of studying the traditional video scene semantic segmentation method, this paper proposes a semantic marking method of video scene based on 3D convolutional neural network. This method uses 3D convolutional neural network to learn the high-level discriminant features of objects of different semantic categories to improve the accuracy of semantic segmentation. The experimental results show that the proposed method is comparable to the classical method in segmentation accuracy and can overcome the drag effect caused by the false optical flow field relations.