Action recognition method based on local marginal maximization
An action recognition and action technology, applied in the field of computer vision, can solve the problems of ignoring the information maintenance and protection of discrimination
Inactive Publication Date: 2021-02-23
SUN YAT SEN UNIV
View PDF6 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0005] In order to solve the current problem of maintaining and protecting the discriminant information in the original action behavior data through the method of projection dimensionality reduction in the face of noise interference in video action data, the present invention proposes an action recognition based on local margin maximization method to improve the accuracy of action behavior recognition
Method used
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
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment 1
[0077] Such as figure 1 The flow chart of the action recognition method based on local margin maximization is shown, see figure 1 , the method includes the following steps:
[0078] S1. The action video data set is divided into an action video training set and an action video data set to be identified;
[0079] S2. Represent each action video data in the action video training set and the action video data set to be identified as a third-order video sequence tensor, and obtain the action video training set tensor and the action video data set tensor to be identified;
[0080] S3. Based on the tensor distance formula, the action video training set is divided into several discriminative parts;
[0081] S4. Extracting the similarity coefficient and dissimilarity coefficient of each discriminative part in the high-dimensional space;
[0082] S5. Using a multi-linear projection method, the action video training set tensor, the extracted similarity coefficient and non-similarity c...
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 PUM
Login to View More Abstract
The invention provides an action recognition method based on local marginal maximization, relates to the technical field of computer vision, and solves the problem that the maintenance and protectionof discrimination information in original action behavior data are ignored through a projection dimension reduction method at present when facing noise interference of video action data. The method comprises steps: representing each action video data in the action video training set and the to-be-identified action video data set by using the tensor, so as to fully consider the spatial informationof the video data; and keeping the similarity and non-similarity coefficients extracted from the original action video data set in the low-dimensional space, so that the local margins of the low-dimensional discriminant local are maximized, the discriminant information carried by each local can be better protected, and then by maximizing the local margins, the recognition accuracy of the action data points in the discriminative local part is improved.
Description
technical field [0001] The present invention relates to the technical field of computer vision, and more specifically, relates to an action recognition method based on local margin maximization. Background technique [0002] In recent years, with the sharp increase in the number of videos available, the processing and recognition of video data has attracted great attention in the fields of video retrieval and video summarization. Among them, the recognition of action videos plays an important role in the field of computer vision, such as video surveillance, content-based video retrieval, virtual reality, and human-computer interaction. [0003] Action video recognition can be viewed as classifying video sequences, and designing a robust method for recognizing complex action behaviors remains a challenging problem for real-world applications. Multiple linear algebra is a very powerful mathematical tool for analyzing motion video data. Since the change of action behavior ove...
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 Application Information
Patent Timeline
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/00G06K9/62G06N20/20
CPCG06N20/20G06V20/42G06F18/213G06F18/24147G06F18/214
Inventor 张舒婕马争鸣刘洁
Owner SUN YAT SEN UNIV