Calibration of an eye-in-hand system using SoftKinetic DepthSense and Katana robotic arm

In a typical eye-in-hand system, a video camera is mounted on end effector of a robot arm. Visual images from the camera are then used for controlling the robot arm to accomplish its task. Since distance measurement cannot be obtained from a single image frame, many frames need to be analyzed by complex algorithms to retrieve the desired measurement. This is imperative for tasks that are critical on accurate distance information, such as collision avoidance and grasping. In this paper, we bypass these difficulties by using a depth camera, instead of a video camera, so that distance images can be directly acquired. More precisely, we install a depth camera DepthSense DS325 on a Katana robotic arm. We describe our implementation and to confirm effectiveness of the eye-in-hand system, a sample application performing object model construction is presented.