基于Kalman滤波的视觉预测目标跟踪及其应用

Kalman filter based vision predicting and object tracking method and its application

建立了基于图像处理的视觉信息感知几何模型,提出了一种基于Kalman滤波预测估计的视觉信息感知方法,并据Kalman滤波理论建立了对机器人终端手抓等目标状态进行预测跟踪的模型。同时,利用建立的模型在护理服务机器人系统中进行了实验研究。实验结果表明,所提出的方法可以较好地对目标状态进行预测估计,通过"预测-校正"可以实现一种快速准确的视觉信息感知。该方法应用在护理服务机器人系统中可以减少视觉通道反馈信息的时间滞后,增强人机对话和信息感知能力。

In order to make healthcare service robots fluently and flexibly complete some tasks, a geometry model was set up to detect vision information by image processing. A kind of vision information method for predicting and processing based on Kalman filter was put forward. Moreover, a set of predicting and object tracking model was also provided in light of Kalman filter theory. Some experiments were carried out with these models in a service robot system, the experiment results show that the object status can be accurately predicted by provided method. Furthermore, the vision information can be quickly apperceived by prediction and correction of the models. On the other hand, the time lag for vision information feedback can be decreased with the provided method in the service robot system, so the communication between human and robot takes place more fluently.