Target Tracking Technology Of Mobile Robot Based On RGBD Camera

Mobile robots show a potential practical value and broad application prospects in the field of people’s livelihood and national defense such as intelligent surveillance, unmanned vehicles, disaster search and rescue, environmental monitoring, military surveillance, interplanetary probe and so on. With the development of computer visual equipment and image processing algorithms, the visual sensor for mobile robots to complete a variety of tasks has gradually become a hot topic of scholars. Thesis presents a method of mobile robot carrying with a visual sensor to detect the target and determining the position of the object and other information from the measurement data against the tasks of tracking the moving target for a mobile robot with visual sensors. The main tasks are as follows:Build a structures platform of iRobot Create mobile robot with Kinect sensor. This platform uses PC as the main controller, Kinect sensor together with a PTZ as the visual subsystem to auxiliary control. The subsystem of iRobot Create mobile robots is used to do the underlying driving work. Power the PC, iRobot Create and Kinect respectively.To recognize the object, the method this thesis proposed comprises three steps. Firstly, do some pretreatment operations such as threshold segmentation and morphological operations in the depth image to remove the background so that we can attain the interesting foreground area including the target. Secondly, map the depth image to color image through correspondence between synchronized depth images and RGB images. Finally, use SURF algorithm and RANSAC algorithm in RGB images to identify the moving object.Take the depth information and color information of the identified moving target as a t arget model, with a particle filter algorithm to predict the location of the goals,to effectively avoid other interference with simial color during the tracking process.Experiments show that the system of mobile robot that established by this article is lightweight and reliable and the image processing algorithms have a great advantage in stability, instantaneity, reliability and robustness. The system has a good perform in tracking a moving target. The technology this paper studied has great scientific value and use value.