Abstract: Aimed at the problem that fall detection can not be detected due to obstacles, a fall detection model based on mobile robot is proposed. Based on the robot attitude control technology, under the constraint of robot joint, the data set of NAO robot attitude was collected by using double upper computers. Based on the fusion of direction histogram and gray level co-occurrence matrix, a dual feature fall detection model was established. Based on ROS mobile robot, the fall detection model of NAO robot in different scenes was realized. The experimental results show that, based on big data, the fall detection of dual feature fusion improves the accuracy of fall detection compared with the fall detection of single feature. This algorithm is suitable for practical engineering and fall detection of the elderly.