3D Reconstruction And Obstacle Avoidance For Mobile Robot Based On RGB-D Sensor

With the rapid development of robot technology,mobile robot is widely used in industry and service trades because of the large tasking area,wide range of application coverage and moving mobility etc.In the field of mobile robots,how to detect the environment around the robot really,effectively and quickly is an important prerequisite for its follow-up study.That is to say,first we need to be aware of the surrounding scene actively.In dynamic scenes,real-time 3D reconstruction information of the environment is particularly critical.Robot can make behavior decision based on the scene information,and interact with the surrounding environment.In this paper,robot can acquire surrounding environment information based on the sensor system and plan the path,so as to achieve the expect destination without collision.During the research,using a RGB-D sensor for the images around the robot,we can reconstruct the depth image to 3D point cloud using the imaging principle.According to point cloud we fit an equation of the ground,we can find the points which belong to the obstacle,and build obstacle grid map.To identify the property of the obstacle,in this paper,we detect the moving object in the scene using the idea of robust principal component analysis(RPCA).In this algorithm,we use Lanczos algorithm to accelerate the segmentation of moving object from a batch of images.We can get a set of basis vectors.In this paper,improving an incremental RPCA algorithm to make detection online,we can update the basis vectors after a new image joining in and detect the moving object in real time.If the result of recognition is acquired,the robot makes behavioral decision based on the result of the detection of the obstacle and the moving object.When detecting the moving objects,robot stops to wait.When detecting none moving object,the robot make path planning based on the obstacle grid map using improved algorithm of vector field histogram.In the final of the research,we build an experimental platform based on Pioneer robot to verify the accuracy of moving objects detection and feasibility of obstacle avoidance.Through the method of theoretical analysis and experimental verification,we expounded the process of robot perception of the surrounding environment,and study on moving object detection theory based on RPCA and obstacle avoidance algorithm based on VFH.This paper provides some new ideas about intelligent applications of mobile robot in unknown environment.