The human swing sport plays an important role in tennis, badminton, table tennis and other sports, and its training effect depends on accurate sports posture evaluation. Traditional sports training methods are difficult to acquire athletes' 3D posture data in real time and efficiently, while the development of computer vision technology provides a new solution for sports posture analysis. In this paper, a computer vision estimation algorithm for human 3D posture based on enhanced image details is proposed and applied to swing sports training. This paper investigates the key techniques of computer vision in human posture estimation and discusses the effect of enhanced image details on the accuracy of 3D posture estimation. In this paper, a human 3D pose estimation algorithm integrating image enhancement and deep learning is designed, including key aspects such as data preprocessing, model training and optimization. In this paper, a computer vision-assisted swing sports training system is constructed to realize real-time acquisition of sports data, posture analysis and feedback, and to improve the science and accuracy of sports training. The experimental results show that the method is better than the traditional method in terms of posture estimation accuracy and sports training assistance effect, which provides effective support for intelligent sports training.

Chaobing Yan, Ting Zheng. A study on assisted swing sports training based on computer vision estimation algorithm of human three-dimensional posture with enhanced image details. Academic Journal of Computing & Information Science (2025), Vol. 8, Issue 4: 33-39. https://doi.org/10.25236/AJCIS.2025.080404.

[1] Tran C , Trivedi M M .3-D Posture and Gesture Recognition for Interactivity in Smart Spaces[J]. IEEE Transactions on Industrial Informatics, 2012, 8(1):178-187.DOI:10.1109/TII.2011.2172450.

[2] Bilal S , Akmeliawati R , Shafie A A ,et al.Hidden Markov model for human to computer interaction: a study on human hand gesture recognition[J].Artificial Intelligence Review, 2013, 40(4):495-516. DOI:10.1007/s10462-011-9292-0.

[3] Tsai M J , Lung H Y .Two-phase optimized inverse kinematics for motion replication of real human models[J]. Journal of the Chinese Institute of Engineers, 2014, 37(7):899-914. DOI:10.1080/02533839.2014.904474.

[4] Murugan A S , Noh G , Jung H ,et al.Optimising computer vision-based ergonomic assessments: sensitivity to camera position and monocular 3D pose model[J].Ergonomics, 2025, 68(1):120-137. DOI:10.1080/00140139.2024.2304578.

[5] Bandouch J , Jenkins O C , Beetz M .A Self-Training Approach for Visual Tracking and Recognition of Complex Human Activity Patterns[J].International Journal of Computer Vision, 2012, 99(2):166-189. DOI:10.1007/s11263-012-0522-y.

[6] Aminian Modarres A F , Soryani M .Body posture graph: a new graph-based posture descriptor for human behaviour recognition[J].Iet Computer Vision, 2013, 7(6):488-499.DOI:10.1049/iet-cvi.2012.0121.

[7] Chwa D .Robust Distance-Based Tracking Control of Wheeled Mobile Robots Using Vision Sensors in the Presence of Kinematic Disturbances[J].IEEE Transactions on Industrial Electronics, 2016, 63(10):1. DOI:10.1109/TIE.2016.2590378.

[8] Trigueiros P , Ribeiro F , Reis L P .Generic System for Human-Computer Gesture Interaction: Applications on Sign Language Recognition and Robotic Soccer Refereeing[J].Journal of Intelligent & Robotic Systems, 2015, 80(3):573-594.DOI:10.1007/s10846-015-0192-4.

[9] Cho S Y, Wong J J. Human face recognition by adaptive processing of tree structures representation[J]. Neural Computing & Applications, 2008, 17(3):201-215.DOI:10.1007/s00521-007-0108-8.

[10] Hao Y, Guo S, Zhou X Y H .Underwater swimming fish mass estimation based on binocular vision[J]. Aquaculture International, 2024, 32(6):7973-7995.DOI:10.1007/s10499-024-01550-z.