基于RANSAC和Kalman滤波的足球机器人球速估计算法

Ball velocity estimation method based on RANSAC and Kalman filter for soccer robots

针对中型组足球机器人如何有效地估计足球速度的问题,提出了一种基于Kalman滤波和RANSAC算法的新方法。首先对存储的若干帧足球位置信息作Kalman滤波,接着利用这些足球位置信息,建立若干个可能的足球速度模型并运用随机采样一致(RANSAC)算法选出最优的速度模型作为速度值。实验结果验证了该算法的有效性,同时由于RANSAC算法可以有效地去除外点的干扰,因此当足球位置信息具有较大噪声时,该方法可以较准确地估计足球的速度,较以往球速估计的算法具有更高的鲁棒性。

A new method was proposed to estimate the ball velocity in the Robot World Cup （ RoboCup） Middle Size League （ MSL） more effectively. The method was based on Kalman filter and RANSAC algorithm. Firstly several frames of the ball positions were stored and then Kalman filter was used to optimize the positions. Hundreds of models were built and RANSAC algorithm was employed to calculate the best model as the final ball velocity. The experimental results show that the proposed algorithm is effective. Furthermore, RANSAC algorithm can eliminate the outlier effectively, so when there are lots of noises in the ball information, the ball velocity can still be estimated with high accuracy by using the presented algorithm, and higher robustness can be achieved being compared with the other existing methods.