Pole detection and autonomous gripping are fundamental but essential requirements for biped pole-climbing robots. This kind of robots is mainly designed for the truss-style environment, constructed of rigid poles. For each step of the robot climbing, the gripper must get the pose of the target pole and be aligned well with it. The pole detection is difficult due to the complex climbing environment. The poles may be cross-distribution and obstruction in the three-dimensional space. In this paper, an RGB-D sensor is applied to acquire the color images and point cloud of real scenes simultaneously. A noniterative algorithm is presented to detect the surrounding poles and to compute the pose of the gripper relative to the selected target pole. Experiments verify that the proposed algorithm is feasible and high-efficiency and a biped pole-climbing robot can autonomously grip the target pole utilizing this sensing system.