Design And Implementation Of Human Motion Parameter Prediction Subsystem

With the development of artificial intelligence,the field of computer vision has been studied by more and more scholars.In the past few years,the research of 2D vision has undergone rapid development.Researchers begin to pay more attention to 3D vision,hoping that the machine can perceive the information of 3D space from 2D image.As an important task,3D human body reconstruction has become the current research hotspot.The goal of the project is to reconstruct the three-dimensional human motion based on the image.The goal of the paper is to predict the motion parameters of the human body in the image.The expected problem of this paper is how to extract the feature of image and predict the motion parameters of human body according to the feature.Due to the difficulty of 3D information collection and the diversity of human actions,it is still a challenge to predict human motion parameters in general outdoor scenes.In this paper,human motion parameters prediction algorithm(HMPP)is designed.HMPP is divided into thre stages:2D key point detection,3D key point detection and human motion parameter prediction.The training in the three stages is independent.In addition,projection supervision is added in the latter two stages,so that outdoor 2D human body data set can be used for training,not only relying on indoor 3D human body data set,but also improving the prediction ability of the algorithm in outdoor scene.The HMPP algorithm has achieved good evaluation results on the common data set human3.6m,with an increase of more than 10%compared with the baseline method.This system consists of three parts:data preprocessing module,algorithm prediction module and application display module.The data preprocessing module divides the input video into frames and processes it.The algorithm prediction module is based on the HMPP algorithm.The application display module uses the motion parameters to predict the motion parameters of the human body in the image 3D human body rendering and action redirection.After testing,the system runs stably and the application shows good results.Firstly,the paper introduces and analyzes the background and significance of key point detection and motion parameter prediction,and then introduces the related technologies in detail,then researches,designs and evaluates the HMPP algorithm,and then analyzes the demand,outline design,detailed design and system test of human motion parameter prediction subsystem.Finally,the work of this paper is presented It also makes a summary and prospects for the follow-up work.