A Complex Action Recognition Method Based on Simple Actions Exploring Privileged Information
An action recognition and action technology, which is applied in the field of knowledge-assisted learning of complex actions, can solve the problem that the intelligence cannot reach the user, and achieve the effect of improving the recognition rate.
Active Publication Date: 2021-08-10
SOUTH CHINA UNIV OF TECH
View PDF4 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0002] In recent years, video surveillance has been widely used, and the demand is still growing. However, the intelligence of existing video surveillance systems is far from meeting user expectations.
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
[0028] figure 1 is a flow chart of the complex action recognition method for exploring privileged information based on simple actions proposed by the present invention, such as figure 1 As shown, the method includes the following steps:
[0029] Step S1, using the improved dense trajectory (IDT feature) to perform feature extraction on videos of simple actions and complex actions to obtain a feature matrix.
[0030] In step S1, the improved dense trajectory (IDT feature) method has a very good effect in behavior recognition. In the improved dense trajectory, there are four descriptors: trajectory displacement vector (trajectory), gradient histogram (histogram of oriented gradients, HOG), optical flow histogram (histogram of optical flow, HOF) and X and Y directions. Optical flow information (motion boundary histograms, MBH). The Fisher vector is used for encoding, the size of K in the Gaussian mixture model (GMM: Gaussian mixture model) is set to 256, and the final dimensio...
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 discloses a complex action recognition method for exploring privileged information based on simple actions. The method includes the following steps: S1. Using an improved dense trajectory (IDT feature) to extract features from videos of simple actions and complex actions to obtain a feature matrix ; S2. Exploring effective information in simple actions as privileged information; S3. Building models through privileged information to help learn complex actions. The present invention analyzes complex actions, considers that complex actions can be composed of a series of simple actions, and proposes a brand-new framework to realize the recognition of complex actions by using a series of simple actions. The invention uses the privileged information matrix to encode the probability of occurrence of each simple action in the complex action to help the learning of the complex action, thereby improving the recognition rate of the complex action.
Description
technical field [0001] The invention relates to the technical field of artificial intelligence and video processing, in particular to a complex action recognition method for exploring privileged information based on simple actions, in particular to a learning technique based on privileged information that utilizes the knowledge of simple actions to help the learning of complex actions. Background technique [0002] In recent years, video surveillance has been widely used, and the demand is still growing. However, the intelligence of existing video surveillance systems is far from meeting users' expectations. The recognition of human behavior has important application value in video surveillance. It can be used as the basis for further high-level semantic analysis and understanding, making intelligent detection, alarm, and retrieval possible. With the development of information technology, related application requirements are becoming more and more diverse. Video-based behavi...
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 Patents(China)
IPC IPC(8): G06K9/00G06K9/46G06T7/246G06T7/269H04N19/136H04N19/139
CPCH04N19/136H04N19/139G06T7/246G06T7/269G06T2207/10016G06V40/20G06V20/42G06V20/46G06V10/50
Inventor 徐向民刘芳
Owner SOUTH CHINA UNIV OF TECH