Indoor Object Recognition And 6 DOF Pose Estimation Using Pcland ROS

Nowadays, with the rapid development of AI(Artificial Intelligence) and Smart Home technology, more and more service robots appeared in public. Indoor object recognition and pose estimation will be essential to the service robots. Therefore, we select the KUKA youBot from Guangdong University of Technology Robotics laboratory as a research object, by means of ROS(Robot Operating System) and PCL(Point Cloud Library), we focus on the problem of Indoor object recognition and 6DOF pose estimation to make the service robots pick and place autonomously.In the first place, we take an brief introduction of the youBot platform and the Kinect, as well as the ROS and PCL, then we will apply some filters on the raw data from the Kinect to get ready for the next segmentation step.Secondly, as the indoor environment has many plane features, RANSAC algorithm will be used to extract the plane elements and then get the segmented clusters. With the different view of the same object, here we could use the global geometrical feature CVFH(Camera view feature histogram) as the feature descriptors to build a model database, then use the open source HDF5 and kd-tree build the index, and at the end, FLANN will be used to search the nearest neighbor with respect to the object in the scene and the one in the database.Thirdly, only if we align the target in scene and the one in the database can we get the position and orientation between them. Whereas, the CVFH descriptor is invariant to the camera roll angles, we could only get 5DOF pose. The CRH is the solution which will calculate the roll angle, so we will get 6DOF pose at last.Moreover, the alignment result could be refined use the optional ICP, and the Global Hypothesis Verification will reject the false positives, the two methods talked above are not mandatory.Lastly, after Calibration the Kinect and the robot, besides the pose got from the pipeline above, we could use the robotics inverse kinematics to calculate the joint angle and reach the target, which had been done in RVIZ in ROS, the pipeline is practicable and will be have a good future.