A real-time dense method to address the problem of mobile robot simultaneous localization and 3D mapping(3D SLAM) in complex indoor environment is proposed. In this approach, the environmental data is captured by using a RGB-D camera which is fixed on the robot. Combining with the local texture association, a hybrid algorithm model is established to ensure the pose estimation accuracy and concurrently decrease the failure rate during mapping by using the point cloud and image texture. By taking advantage of the keyframe selection mechanism, a visual-based loop detection algorithm and tree-based network optimizer(TORO) are used to achieve a global consistency map. Experimental results show the feasibility and effectiveness of the proposed algorithm in the indoor environment.