The feature extraction from athlete action images is a research hotspot. To accurately evaluate athlete actions, it is necessary to partition the original image into refined blocks, and extract different levels of image features. However, the traditional feature extraction algorithms can only roughly divide action images into several classes, failing to acquire the accurate feature sets of the actions. This leads to relatively poor performance of feature extraction from action images. To overcome the defect of the traditional methods, this paper puts forward a feature extraction method for the action images of badminton players based on hierarchical features. The underlying image features were analyzed based on the techniques of badminton players, and mapped to the feature space of the corresponding dimension. Simulation results show that the proposed method can accurately extract the features from athlete action images. 1. Introduction 2. Principle of Feature Extraction from Athlete Action Images 3. Feature Extraction Based on Hierarchical Features 4. Tracking Stroking Action Trajectory of Badminton Players 5. Simulation and Results Analysis 6. Conclusions References
© Copyright 2020 Ingénierie des Systèmes d`Information. International Information and Engineering Technology Association. Alle Rechte vorbehalten.