A Method Of Human Recognition And Tracking With Kinect For Human-robot Collaboration

With the development of computer vision, the Human-robot Collaboration (HRC)technology has been one of the hot research topics. HRC refers that humans and robotscooperate on a common task, in the same working environment. This requires that robots cancommunicate with humans through actions, gestures, voice or other methods, to ensurehuman security issue during collaboration time. The human security issue is a hard work inthe HRC technology field.To solve this issue, the thesis presents a novel method of human recognition and trackingbased on the Kinect. In the thesis, it applies SURF (Speeded-Up Robust Features) algorithm,which has faster calculating speed, higher integrated performance and stronger robustness inrotation and lightness invariant. The thesis encompasses three parts: firstly, collect three-dimensional data from target human and the scene by Kinect, including color image data flowand depth image data flow. Combine the data information and filter out the background of theoriginal image, and obtain a new image only with the body area; Secondly, utilize the SURFto detect the new image to recognize the target human, and so increase the computation speedand the calculation; thirdly, use the UKF(Unscented Kalman Filter) algorithm to track thetarget.At last, the experiment shows, in the complex environment, the method can recognizethe target human rapidly and accurately, when the target is hidden by others or the lightnesschanges. In the lab, it achieves the goal of identifying and tracking one person.