Efficient motion characterization method and device based on motion boundary small displacement

A motion boundary, small displacement technology, applied in the fields of visual perception and artificial intelligence, can solve the problems of time-consuming and high computational complexity, and achieve the effect of ensuring real-time performance, high computational complexity, and small number of parameters
CN112446245APending Publication Date: 2021-03-05PEKING UNIV SHENZHEN GRADUATE SCHOOL
0 Cites 3 Cited by

Patent Information

Authority / Receiving Office
CN · China
Current Assignee / Owner
PEKING UNIV SHENZHEN GRADUATE SCHOOL
Publication Date
2021-03-05

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention relates to an efficient motion characterization method and device based on motion boundary small displacement. The method comprises the following steps: step 1, extracting original images of adjacent N frames in a video sequence; 2, processing adjacent N frames of original images by using a convolutional neural network to obtain a corresponding shallow feature map; 3, performing difference calculation on the shallow feature maps of all the two adjacent frames of the adjacent N frames to obtain a difference map of all the two adjacent frames in the feature space; 4, carrying out difference accumulation on difference graphs of all two adjacent frames in the feature space along the channel dimension; and step 5, encoding the difference accumulation result according to an encoding scheme so as to obtain the efficient motion representation provided by the invention. Compared with some methods depending on optical flow as motion characterization, the method does not need to perform complex optical flow calculation in advance, and can model the small displacement of the motion boundary by calculating the difference on the shallow feature space, thereby greatly reducing the complexity of motion characterization calculation.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The present invention relates to visual perception and artificial intelligence technology, in particular to an efficient motion characterization method and device based on small motion boundary displacements, which can effectively construct A small displacement of the mode motion boundary is used as a motion characterization. Background technique

[0002] Motion representations have been widely adopted in computer vision research in recent years, especially for video understanding tasks. The current mainstream video-based deep learning tasks, such as: action recognition, video description, video prediction, etc., in addition to the original color 3-channel RGB image as input to provide appearance information, motion representation is also required as one of the input modalities. To provide timing-related short-range motion information as a learning aid. The modeling of motion representations has gradually become an important research direction in the...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More