Examinee posture segmentation and smoothing method

Abstract

The invention discloses an examinee posture segmentation and smoothing method, and the method mainly comprises the steps: obtaining posture data, obtaining data of all human body postures in a frame of image, and carrying out initialization; determining the centroid of each posture, and segmenting postures belonging to the same person in the video into a sequence by combining the position information of the human body, so as to achieve the purpose of human body tracking; and detecting the segmented attitude, the change characteristics of the attitude and the rationality of the attitude, eliminating the abnormal attitude and the deviated attitude in the attitude, and smoothing the corrected attitude sequence by adopting a sliding window smoothing method. According to the method, the human body posture in the video can be tracked, the abnormal posture data is screened out, the posture precision in the video is improved, and the posture jitter between the front frame and the rear frame isreduced, so that the influence of abnormal posture change on subsequent use is avoided.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a method for segmenting and smoothing an examinee's posture. Background technique [0002] With the development of deep learning technology and computer vision technology, human body pose estimation technology has become increasingly mature, and the accuracy of pose estimation is also increasing year by year, which has become a relatively active research field in recent years. Pose estimation has a wide range of application scenarios, such as human-computer interaction, intelligent monitoring and so on. These applications are based on the results of pose estimation, and analyze human behavior through pose features. The purpose of behavior analysis is to describe, recognize, and understand human movements, interactions between people and between people and the environment. The behavior analysis process includes a large amount of data analysis and image analysi...

AI Technical Summary

Problems solved by technology

[0003] There are many existing attitude estimation methods, but the accuracy of the attitude cannot be guaranteed in a scene with many people and a complex environment. Some errors are often caused by algorithms or observations, which affect the results of the output attitude and cannot correctly reflect the real attitude of the human body. Therefore, the smoothing technology Commonly used in the post-processing stage of pose estimation

[0004] However, the previous attitude smoothing technology is mainly aimed at simple scenes or scenes with sparse personnel, and handles the joint points independently. There are also some smoothing solutions based on wearable devices, which are not suitable for crowded and complex people such as examination rooms. scene

After the skeleton is extracted, the traditional smoothing technology only smoothes each joint point in the two-dimensional position of the image, which will ignore some abnormal postures of the human body and lack common sense restrictions. A person's posture; another technology uses wearable devices to perform real-time smoothing in the scene of a small number of people, which obviously cannot be used in the environment of the examination room

Images