Abstract: In this paper, the structure and the training method of self-organizing neural network are summarized, the importance of the data association and limitations of some traditional data association methods of tracking multiple targets are analyzed, and the problem of how to solve the data association with this neural network in multiple target condition is studied. A new tracking algorithm for multiple targets tracking based on self-organizing neural network is proposed, which solves the problem of data association by means of a clustering function of self-organizing neural network and incorporating the Kalman filter data into the training method of self-organizing neural network. Simulation results show that the tracking algorithm works well in multiple targets condition.