This paper presents a study on improving adaptability of humanoid climbing robot in a vertical environment. Although the robot is intended to use for rescue and load-carriage in mountain or hazard areas finally, this paper focuses on path planning algorithm for the robot on climbing wall as the initial phase of our development. Therefore, the first step is to acquire Kinect's depth map to extract accurate information about climbing holds on the vertical wall. Secondly, we propose an algorithm of path planning of humanoid robot using Kinect's data. The proposed algorithm ensures that climbing robot with the Kinect's depth camera finds the best route to climb up the wall. The presented algorithm is a form of graph algorithm clustering the data of climbing holds. The algorithm takes into account the specific abilities of robot. Finally, this algorithm is evaluated using a humanoid climbing robot system with a simple practical example, and its effectiveness is proved experimentally in a real environment.