| Service robot is a kind of intelligent robots to serve human daily family life. It has helpage, disabled care, household, entertainment, education and many other features. In particular, that the aging trend of human society is intensifying cased that family-oriented applications service robots attract more and more people’s attention. Detecting human targets in the indoor environment and guiding service robot to track become a hot spot of robotics research in recent years. Visual system because of large amount of information has been gradually applied to the research on service robot, but mainly concentrated in monocular vision navigation. The monocular vision has relatively limited range of observation and just tracks objects in the image. The location of object can not be obtained in the objective world. Based on binocular stereo vision system, this paper senses environmental information by cameras to detect and identify human targets and service robots in the environment and obtains three-dimensional information of two goals by computer vision measurement techniques. Around human target tracking method for service robot, this research plans reasonably based on demand and focuses on the following four aspects:1. Make the checkerboard calibration board. Based on TSAI calibration algorithm, we use MATLAB 2012 develop the binocular stereo vision calibration system, and have done the calibration experiments. The binocular stereo vision calibration system has been calibrated according to the calibrated the left and right cameras’ parameters.2. With the statistical learning methods of detection, the application of Haar and Adaboost algorithm in human body detection has been researched. We make full use of the characteristics of Haar- like local details and learning classification ability of Adaboost to train human positive and negative samples to get classifier with the human body characteristics. The characteristics are used to test the human, and the effect of huamn detection and tracing is verified by experiment.3. CamShift algorithm fused of the Kalman filtering improves detection and tracking method of service robot. Target location is obtained by CamShift algorithm after iterative convergence. The iteration result as measuring values is passed to the Kalman filter and used to predict next position of service robot. The algorithm has judged the serious hiding condition and has made targeted treatment. It largely has improved the effect of detection and tracking.4. The human tracking system experimental platform of service robot has been built by using Visual Studio 2013 programming tool in the Windows environment and configuration of OpenCV 2.4.9. According to the above research and design of calibration of binocular vision, the human detection and tracking and service robot detection and tracking, we have completed the camera calibration, image acquisition, processing and space motion location. The experiment shows that the method has implemented the service robot’s tracking for pedestrian in the indoor environment. |
