In this paper, we propose a alarm system for pedestrian protection. We usually do not know that pedestrians may or may not be in dangerous situation, and to know whether pedestrians are in dangerous situation or not. In this paper, we construct collision probability system between vehicle and pedestrian. By using monte carlo simulation, we calculate the collision probability, and it is hard to know collision probability of all area, we recover collision probability of all area using neural networks. And, the collision probabilities are different according to tendency of pedestrian movement, we understand the tendency of pedestrian movement using interacting multiple model tracking method. Computer simulation will be show the validity of our proposed method.