Development And Application Of Function Software For Mobile Service Robot Based On RGB-D Sensor

Human target tracking, interaction and recognition are the basic functions of mobile service robot, which has wide applications at public services such as airport baggage handling, restaurant service and tour guide service. Human target tracking service of mobile robot based on RGB-D sensor has a significant technical advantage. In this dissertation, based on the need of human target tracking under the complex dynamic environment, design and implementation of function software for high reliable tracking service, interaction and recognition is studied.The hardware and software development platform of mobile service robot system are Turtlebot and Robot Operating System(ROS). The robot perceives environment by RGB-D sensor and laser radar sensor. OpenNI software interface and point cloud Library are also used to develop the function software of tracking service.With the summary of related research work and technology of mobile service robot is given, the function requirements of robot under complex dynamic environment in public places are analyzed, and the designed system framework and function software architecture are proposed. There are three modules in the function software, human target feature extraction and detection, human target tracking and interactive obstacle avoidance. Firstly, after separating human target from background by skeleton extraction and extracting color histogram feature, the confidence of HOG feature is detected. According to the requirements of target tracking, target tracking method of multiple features fusion is proposed, which fuses target motion trend estimation, statistical characteristics of the target area color distribution and morphological characteristics of the target area. It improves the reliability of tracking in the presence of dynamic interference case. Secondly, dynamic window method based on RGB-D sensor and laser sensor and fuzzy PID method for adjusting angle and speed are presented. Then human-computer interface based on network and gesture is developed, which improves the intelligence level of mobile service robot. Finally, experiments of motion tracking, obstacle avoidance and human-computer interation are conducted for evaluation, validating the system's functions and reliability.