In this paper, a novel fuzzy logic data association algorithm is proposed for online visual multi-object tracking. Firstly, in the proposed algorithm, in order to incorporate expert experience into the data association for the improvement of performance in multi-object tracking, a fuzzy inference system based on knowledge is designed by using a set of fuzzy if-then rules. Given the error and change of error of motion, shape and appearance models in the last prediction, these rules are used to determine the fuzzy membership degrees that can be used to substitute the association probabilities between the objects and the measurements (or detection responses). Secondly, in order to deal with the fragmented trajectories caused by long-term occlusions, a track-to-track association approach based on the fuzzy synthetic function is proposed, which can effectively stitch track fragments (tracklets). Because of this, the proposed algorithm has the advantage that it does require no assumption of statistical models of measurement noise and of object dynamics. The experiment results on several public data sets show the efficiency and the ability to minimize the number of fragment tracks of the proposed algorithm.