Objects Position Detection And Pose Estimation For Mobile Robotics Transportation Based On RGB-D Data

To solve the object detection and localization problem in mobile robotics application,this paper optimizes and improves the robustness of the original algorithm based on RGB-D data for object detection.Experiments for manipulation with robotics arm on the mobile robot are applied to verify the performance of the optimized algorithm.Mobile robots show a potential practical value in the field of industrial,national defense,disaster search,rescue and so on.The algorithm proposed in this paper based on RGB-D data for objects detection and localization is mainly consisted of two parts:optimized dominant gradient orientation template algorithm for objects detection and iterative closest point algorithm based on 3d point clouds data for objects localization.The algorithm for objects detection is applied on the segmentation image based on depth image,which extracts dominant gradient orientation as features for template matching.Besides,the local region feature based algorithm is developed to improve the robustness in occlusion scenarios.For texture-less objects detection and small amount samples scenes,algorithm based on template matching still has a good performance.Position information can be calculated by ICP algorithm through matched feature points and depth data.Finally,the mobile robot can acquire position information for the specified object in real time,ensuring the manipulation of the robotics arm for the object.The Turtlebot2 mobile robot platform carried with ORBBEC Astra RGB-D camera and RealSense SR300 camera is built for experiments.The ORBBEC Astra RGB-D camera is used for visual navigation for mobile robots.And the RealSense SR300 camera is used for object detection and localization.The PhantomX Pincher Robot Arm is used for manipulation for objects,whose trajectory can be planned through serial communication by mobile robots.Experiments proves the good performance of the object detection algorithm proposed in this paper.The overall mobile robot system can realize the recognition and manipulation for specified objects under unfamiliar scenes,which has an important value in practical applications.