The Ground Mobile Robot Based On RGB-D Sensor Target Detection And Tracking

With the rapid development of robot technology,people’s demand for mobile robots will increase.Whether in military needs or in civilian use,ground mobile robots will play an increasingly important role,and intelligence will be the development trend of ground mobile robots.Based on the ROS framework,the Xtion pro live RGB-D sensor is the robot core sensor and Turtlebot2 is the experimental platform.It is related to the visual detection and tracking of the ground mobile robot in the indoor environment,which involves the extraction of the target feature,the machine Learning,exercise control and so on.Specific work is as follows :First of al,by reviewing the relevant papers and books,the current situation and development trend of mobile robot’s domestic development are introduced,including the most advanced robot and the main sensor and some problems of the robot.Secondly,the hardware and software needed for the experiment are explained,including ROS(robot operating system),PCL(point cloud library),Xtion pro live(RGB-D sensor).The method of target detection is introduced in detail,that is,the HOD algorithm is used to extract the positive and negative samples of the human body,and the obtained feature data are classified and stored.The SVM classifier is created,the feature data set of the positive sample is read,and the detection model is obtained by machine learning.Then the negative sample is detected by the obtained detection model,and the detector with high accuracy is obtained.Then,combined with the two-dimensional target RGB color histogram to identify the tracking target,Finally,the detected target location is sent to the control node.In the control aspect,the motion parameters of the mobile robot are obtained experimentally,and then the speed and angular velocity of the mobile robot are calculated,and the speed of the left and right is calculated to achieve the purpose of accurate tracking.The experimental part introduces the software system platform and the mobile robot platform,and tests the single target detection and tracking and multi-target detection and tracking respectively.The real-time images and data are detected and analyzed,and the results are analyzed.The main innovation of this paper is(1)The depth map feature extraction and machine learning to complete the task of human detection.(2)The visual control and robot mobile platform to complete the task of tracking the target.