Motion estimation using a fuzzy neural network

This paper introduces an algorithm for two-dimensional (2-D) motion detection and velocity estimation, based on fuzzy logic. In the proposed algorithm, velocity vectors are calculated at each pixel of the image by comparing brightness patterns in consecutive image frames using fuzzy relations. A fuzzy control mechanism is employed to strengthen or weaken the results of the comparisons. The algorithm is implemented with a five-layer fuzzy neural network. Results of simulations show that the proposed algorithm performs well even in the presence of noise.