基于Kalman滤波的视觉预测目标跟踪及其应用

为了使得护理服务机器人实现一种平滑、柔顺的任务执行过程，建立了基于图像处理的视觉信息感知的几何模型，提出一种基于Kalman滤波预测估计的视觉信息感知方法，并据Kalman滤波理论建立了对包括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试验研究，试验结果表明所提出的方法可以较好的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问题，增强了人机对话和信息感知能力。

Kalman filter based vision prediction and object tracking method and application

In order to make healthcare service robots fluently and flexibly complete some tasks, geometry model has been set up to detect vision information by image processing, and a kind of vision information predict and process method which is based on kalman filter, has been put forward, moreover, a set of predictive and object tracking model has also been provided in light of kalman filter theory. some experiments have been done with these models in a service robot system, and the experiment results show that the object status can been accurately predicted by the provided method, furthermore, the vision information can be quickly detected by prediction and modification with the models and methods. On the other hand, the time lag for vision information feedback can be decreased by using the provided method in service robot system, so the communication between human and robot takes place more fluently.