：A method of sample consensus based on Skinner operant conditioning(Skinner-Ransac) is proposed for the problem of fundamental matrix estimation. In this method, every sample is given with the weight, which is changed according to the feedback of every sampling results. For the problem of lacking prior knowledge, three iteration termination selections are proposed. Taking a set of simulated data and a set of real images as the experimental subject, four existing methods are
taken to compare with Skinner-Ransac. The results show that the performance of Skinner-Ransac is better than the other methods.