Human body posture estimation method and system

Abstract

The invention provides a human body posture estimation method and system, and the method comprises the steps: inputting an image into a coder-decoder network which comprises four coding layers and four decoding layers for training, and outputting a semantic segmentation result; converting the pixel semantic probability graphs obtained in the first two coding layers into edge activation pixel graphs by utilizing an energy function, wherein the activation value of a pixel is greater than an activation value threshold, and the pixel is an edge pixel; gathering pixels belonging to the same instance based on semantic tags in the semantic segmentation result to obtain an instance segmentation result, wherein the instance segmentation result comprises a mask representing the instance to which each pixel belongs; and generating a human body skeleton confidence map by using a full convolutional network, outputting a skeleton component label to which each pixel in each instance belongs, regressing joint point positions through a full connection network, and connecting joint points in each instance to construct a skeleton structure of a human body to obtain human body posture information. According to the method and the system, the crossing phenomenon existing in posture estimation between people is improved.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a method and system for estimating human posture. Background technique [0002] Human Pose Estimation Technology (Human Pose Estimation Technology), also known as human pose recognition technology, aims to map the human body pixels in the image or video stream to the three-dimensional surface of the limbs, and can continuously, automatically and quickly extract the human body existing in the image or video stream. The position information of each joint part involves many computer vision tasks, such as target detection, pose estimation, segmentation, and so on. A large number of modeling can be carried out on the basis of the joint point position information obtained by this technology. Its application scenarios include not only human joint point positioning, such as graphics (Graphics), augmented display (Augmented Reality, AR), human-computer interaction (Human-Computer ...

AI Technical Summary

Problems solved by technology

[0010] In order to solve the technical problems in the prior art that the running speed of the top-down human pose estimation method is proportional to the number of people, which makes it inapplicable in complex scenes, and the bottom-up human pose estimation method is prone to wrong connection of joint points, this paper The invention proposes a method and system for estimating human body posture, which solves the above technical problems

Images