救援机器人的目标检测与跟踪系统研究

Research on Target Detection and Tracking System of Rescue Robot

目标检测与跟踪系统是救援机器人实现自主搜救作业的基础。针对Robo Cup Rescue救援比赛项目,救援机器人通过Kinect摄像头获取环境视觉信息,采用与目标模板的SURF特征匹配实现视频帧中的目标检测,同时结合RANSAC算法对错配点进行剔除从而提高检测精度;使用TLD跟踪算法对检测到的目标进行跟踪,同时结合目标深度信息,推算出其世界坐标,驱动机器人实现目标的跟踪。实验结果表明,该方法可快速检测到目标并驱动机器人跟踪。

This paper is used for RoboCup Rescue Competition,the environmental visual information of a Rescue robot is obtained through a Kinect,target detection in video frames is achieved by using the SURF feature matching with the target template,while combined with the RANSAC algorithm to eliminate the mismatch points for improving the detection accuracy. TLD tracking algorithm is used to track the detected target,while the world coordinate of the detected target is calculated by combining the depth information,which helps the robot to track the target.