Research of object tracking based on soft feature

Tracking and Extracting Action Trajectory of Athlete Based on Hierarchical Features

“…Based on hidden Markov model (HMM), Wu et al [17] proposed a trajectory tracking method for the stroking images of badminton players: Firstly, an HMM was trained with the stroking movements, and the distance between stroking arm targets was calculated in turn; Next, the trajectory features of the stroking actions were subject to dimensionality reduction, followed by fuzzy C-means (FCM) clustering; Finally, the trained HMM was employed to accurately predict the trajectory of the stroking actions. Jiang et al [18] designed a trajectory tracking strategy for the stroking actions of badminton players based on adaptive threshold segmentation: the stroking arm targets were extracted from the background through adaptive threshold segmentation, using the hexagonal vertebral body model, and used to predict the stroking action trajectory of the target player. With the aid of low angle camera, Zhang et al [19] presented a trajectory tracking approach for stroking actions of badminton players: First, the 3D information of feature points were acquired from the stroking arm, and the stable and unstable feature points were differentiated by the height of the arm; after that, the stable feature points were tracked, creating the stroking action trajectory.…”

A new spectrum optimization method for variable structure object tracking

“…Randomness and complexity of object motion exist in the real environment, so the theoretical and practical researches on these have great significance [1]. In object tracking, problems are found such as object lost in the condition of variable structure moving with soft feature tracking method [2]. In order to solve this problem, the soft feature theory of moving object and the method of feature extraction with spectrum optimization were put forward in this paper.…”