The general goal for the pre-crash system is to detect all situations correctly in which an accident cannot be avoided any longer. The system decides for a critical situation if an accident will occur immediately which can not be avoided any longer by braking or steering. In these cases a safety action will be triggered which can be non-reversible like an airbag or active engine hoods. For that reason the detection performance and the measurement accuracy of the considered multi sensor system are of high interest and importance for the complete pre-crash system. To fulfill the strong requirements a multi sensor system is considered which consists of a 24 GHz radar sensor and a Photonic Mixing Device (PMD) camera. The automotive radar system has several advantages like high measurement accuracy and long-range target detection. The PMD camera has the advantage that it can recognize the dimensions of targets combined with the related position and the possibility to resolve in angle. Based on a deterministic model a new model has been developed which takes the sensor accuracies into account. With this probabilistic model a decision for a crash is always given together with a decision safety. Additional to the two deterministic alternatives of crash or no crash a third alternative is introduced where no assignment can be made with the desired false alarm probability. First the sensor data is gathered. The environment sensor data is interpreted for the detection of targets and the car data is used for the calculation of a crash area. These two datasets are combined to the pre-crash system. The paper describes the multisensor system combined with the new mathematical algorithm.