融合卡尔曼滤波的网球运动目标检测研究

Research on Tennis Moving Target Detection Based on Kalman Filter

网球运动视频中的网球目标在画面中所占像素较少,且球速较快,传统目标检测算法对网球运动目标进行检测时,会出现目标漏检和错检的情况。提出一种融合卡尔曼滤波的网球运动目标检测方法,即EKF-GMM(Extended Kalman Filter-Mixed Gaussian Model),在混合高斯模型目标检测算法的基础上,通过建立二阶扩展卡尔曼方程对目标检测结果进行跟踪,并对检测结果进行预测和修正,根据目标与卡尔曼预测值的对应关系,消除漏检和错检目标干扰,得到网球运动目标位置参数。实验结果表明,与GMM目标检测算法相比,该方法在F1-measure指标上提升了8.18%。

This paper proposes a tennis moving target detection method based on Kalman filtering,namely EKF-GMM(Extended Kalman Filter-Mixed Gaussian Model).Based on the mixed Gaussian model target detection algorithm,the target detection results are tracked by establishing a second-order extended Kalman equation,and the detection results are predicted and corrected.According to the corresponding relationship between the target and the Kalman prediction value,the interference of missed and misdetected targets is eliminated,obtain the target position parameters of tennis movement.Experimental results show that compared to the GMM target detection algorithm,this method improves the F1-measure index by 8.18%.