A 3D imaging sensor for mobile manipulation

This paper aims to propose a robust invisible near-infrared (NIR) structured-light 3D sensor to identify the location of the picked-and-placed object for a mobile manipulator. The NIR 3D sensor consists of a gobo commercial projector and two off-the-shelf monochrome camera with NIR pass filter. Taking advantage of the NIR light, the negative influences of both object texture and ambient light are significantly reduced. The geometric feature-based pattern utilizes the customized design chessboard corner as the primitive of the one-shot pattern. The designed pattern has the ability to measure the target distance regardless of the target texture and occlusion. The prototype of proposed sensor is implemented and installed on the manipulator, and then experiments were conducted to evaluate the robustness and accuracy of the proposed 3D sensor using the proposed one-shot pattern. The results demonstrate that the efficiency of the developed NIR 3D sensor used for grasping manipulation.