基于计算机视觉的艺术体操轨迹跟踪研究

摘要

为了解决传统基于卡尔曼滤波算法进行艺术体操轨迹跟踪时存在的跟踪漂移以及跟踪效率低等问题,研究基于计算机视觉的艺术体操轨迹跟踪方法,通过ViBe运动目标检索算法对图像的颜色以及深度信息建模,基于图像颜色以及深度的波动检测出视频中的运动目标,采用KCF算法实现运动目标的初步跟踪,在该方法的基础上,通过改进KCF算法解决运动目标被遮挡出现的跟踪漂移问题,提高运动目标跟踪的精度和稳定性.通过Hermite插值运算运动目标质心,基于时刻t的运动模糊方向获取瞬时质心轨迹,得到最佳的运动目标质心轨迹,采用曲线拟合措施获取精确的运动目标质心轨迹.实验结果说明,所提方法可准确跟踪艺术体操运动轨迹,具有较高的跟踪效率和稳定性.%Since the traditional trajectory tracking of rhythmic gymnastics based on Kalman filtering algorithm exists tracking drifting and has low tracking efficiency,a trajectory tracking method of artistic gymnastics based on computer vision is studied,in which the ViBe moving target retrieval algorithm is used to model the color and depth information of the image,the moving target in video is detected according to the variation of the image color and depth,and KCF algorithm is adopted to realize the preliminary tracking of moving target. Based on this method,the KCF algorithm is improved to solve the problem of tracking drifting while the moving target is occluded,and improve the tracking accuracy and stability of the moving target. The Hermite interpolation is used to calculate the centroid of moving target to get the instantaneous centroid trajectory of motion blur direction at t moment,which can acquire the optimal centroid trajectory of moving target. The curve fitting method is employed to obtain the accurate centroid trajectory of moving target. The experimental results show that the method can track the trajectory of artistic gymnastics accurately,and has high tracking efficiency and high stability.