Algorithm for Human Action Recognition Based on Images and Visual Data Using OpenPose

ТАТУ хабарлари · 2025-yil

Annotatsiya

Human action recognition is a fundamental task in the field of computer vision and has become increasingly important in applications such as human-computer interaction, intelligent surveillance systems, virtual and augmented reality, and smart transportation.With the rapid advancement of deep learning techniques and cutting-edge algorithms, the effectiveness and accuracy of action recognition systems have significantly improved in recent years.In this study, we propose a real-time human action prediction model based on skeletal keypoints extracted from static RGB images using the OpenPose framework.The model processes spatial configurations of human joints to identify and classify actions with high precision.By leveraging Part Affinity Fields (PAFs) and confidence maps, our system successfully estimates human poses and tracks body movements efficiently.Experimental results demonstrate that the proposed approach outperforms several traditional methods, including DeeperCut (58.1%) and Convolutional Pose Machines (CPM) (55.2%), achieving a Percentage of Correct Keypoints (PCK) of 85.2%.This high accuracy indicates the model's robustness and its potential for real-world applications that require efficient and accurate motion analysis in real-time scenarios.The results highlight the advantages of combining deep learning with skeletal keypoint estimation for advanced and scalable human action recognition systems.

Maqola ma’lumotlari
JurnalТАТУ хабарлари
Nashr sanasi2025-04-11
DOI10.61663/251tuitmct6

ТАТУ хабарлари jurnalidan boshqa maqolalar

ТАТУ хабарлари — barcha maqolalar