Human body posture classification method based on computer vision

Abstract

The invention relates to the technical field of computer vision human body recognition and classification, in particular to a human body posture classification method based on computer vision. The invention comprises the following steps: segmenting an obtained coordinate from a small numerical value to a large numerical value; marking once every 5 cm in length; obtaining the value of the mark; collecting the motion frame number of the human body motion obtained by the color-depth sensor light projector according to the marked numerical value; acquiring the video collected on the action frame number, sorting and collecting the video, and labeling the video; after the labels are labeled, putting the labels into the table; and performing number classification on different actions by using table screening. The invention is high in precision and very clear, and no fuzzy concept is generated; a hyperlink is inserted into the label, and the video of the action can be directly seen when the action label is clicked; and finally, proofreading is carried out, so that the proofreading is very convenient.

Description

technical field [0001] The invention relates to the technical field of computer vision human body recognition and classification, in particular to a computer vision-based human body posture classification method. Background technique [0002] Human posture based on computer vision: Extracting effective motion features from video sequences is an important part of human action recognition, which directly affects the accuracy and robustness of action recognition, and the same feature can describe different types of human actions Capabilities are not the same. Therefore, we often choose different types of features according to the video quality and application scenarios, which are related to the specific application scenarios and the action categories that researchers care about. For example: in the case of the distant view, the trajectory analysis of the target can be used; in the case of the close view, it is necessary to use the information extracted from the image sequence ...

AI Technical Summary

Problems solved by technology

[0003] The existing computer vision-based human posture classification methods can all be classified very well when they are classified. They will use machines to model human movements in three models, and then establish coordinates. After the coordinates are established , when performing the extraction action, the extraction is not performed according to a certain length of the coordinates, because when the human body is performing an action, there will be a range of activities. We habitually name and extract without length, so it leads to classification. Sometimes, there will be ambiguity, resulting in low accuracy and clarity

Images