Abstract: A method of sample consenus algorithm based on Skinner probabilistic automata (Skinner-Ransac) was proposed to solve the data association problem in monocular vision simultaneous location and mapping (SLAM) for mobile robots. Combined with the priori information of extended kalman filter (EKF) motion model, to assign weight for each sample in the set of image matching points samples, and update the weight for each sample based on the current sample results, a new iteration terminal condition was put forward for the lack of the priori knowledge. A set of open image data were taken as test samples. Results show that Skinner-Ransac algorithm is efficient and reliable, and SLAM’s pose estimation accuracy can be achieved for the need of mobile robot autonomous navigation.