基于自适应Kalman滤波的机器人运动目标跟踪算法

针对足球机器人比赛时的模型变化及其环境噪声先验估计不准确的问题,提出一种基于自适应卡尔曼滤波的足球机器人视觉跟踪算法。该算法将一种基于减背景的运动目标识别的方法与自适应卡尔曼滤波跟踪模型进行结合,对背景进行实时更新,并通过形态学滤波去除残留的小区域,从而准确的识别运动目标,通过自适应的在线调整运动模型参数来保证模型预测值的准确性,进而提高了目标跟踪时的匹配效率,实现了目标的精准、迅速跟踪。通过实验证明,该算法是很有效的,具有推广价值。

Robot Moving Target Tracking Algorithm Based on Adaptive Kalman Filter

In view of the model change of soccer robot in games as well as inaccuracy of a priori estimate to the ambient noise, a kind of visual tracking algorithm is put forward for the soccer robot based on the Adaptive Kalman Filter (AKF). The algorithm combines a moving target recognition method based on background subtraction and the AKF tracking model together, makes real-time update to the backgrounds, removes the residual small areas through morphological filter, and thus accurately recognizes the moving targets. The adaptive online adjustment of motion model parameters is adopted to ensure accuracy of predicted values of the model, so that the matching efficiency at the target tracking is improved, and the accurate and rapid tracking of the target is achieved. It has been proved through experiment that this algorithm is efficient and worthy of popularization.