Design And Implementation Of 3D Human Keypoint Detection Based On Video

With the improvement of computation ability,computer vision technology has also received more and more attention and development.As an important branch of computer vision,deep learning technology also gets more development and application.At the same time,the popularity of modern portable shooting equipment,such as smart phones,tablet computers and other electronic devices,as well as the improvement of network transmission speed,has greatly promoted the growth of image data.With the support of huge image data,deep learning technology has been booming,and deep learning has been more and more applied to the actual scene.As a practical problem,the detection of human keypoints is proposed to solve the problem of how to extract features from images and detect human posture information based on features.This paper investigates and analyses each technics in the process of human joint point 3D detection technology,especially the object detection network technology,human keypoint detection technology and depth prediction technology in the process of huma 3D kepoint detection.Firstly,this paper introduces how to collect and build the data set of related tasks.Base on this,the paper designs and implements a 3D human keypoint detection algorithm process with depth information,and optimizes the accuracy of each algorithm module in the process.Through the ablation experiment designed by the author,the effectiveness of the algorithm is verified by comparing with the experimental results of current network models commonly used in academia.On this basis,this paper designs and implements a 3D joint detection service system,which realizes the function of predicting 3D joint results from the input video to the final output.This paper gives a detailed description of the detection service system from the aspects of system requirements analysis,outline design,detailed design and system testingThe object detection algorithm described in this paper has been compared with the current classical algorithm in the detection accuracy,which is improved,and the overall accuracy has been significantly improved in the overall human keypoint detection process.The service system based on the algorithm has passed the system function test,and the application in the real scene has also been preliminarily verified.