In order to solve the problem that the blind person is easily affected by the obstacles in the indoor environment, an indoor obstacle detection algorithm based on depth information is proposed. First, the Kinect 2.0 device is used to record the image of the indoor environment. Secondly, the obtained depth information is used to reconstruct the 3D scene, and the corresponding 3D point-cloud data is obtained. Then, Combined with random sampling consensus algorithm, the ground is extracted from the point-cloud. After extracting the ground, each obstacle is dispersed in the space, and the method of European clustering is used to segment the obstacles. The experimental results show that the proposed algorithms in this paper can effectively detect a variety of obstacles in the room, and give the blind precise and effective instructions. It can effectively solve the problem that the traditional method is easy to be affected by the light, and have poor detection effect for the obstacles with different depth, inclination and irregular shape, and the output information of traditional methods is monotonous.