Avoidance of near misses or collisions is required for almost all satellites on orbit, but it is of particular interest for manned missions and spacecraft at densely populated regions. In order to avoid these possible collisions, it is needed to determine a possible conjunction and its associated uncertainty. Two main constraints must be taken into account when a tool to forecast the collision risk of an object is being developed: the high number of objects in space and the accuracy of the catalogued object data. The number of objects on Earth orbit makes impossible to propagate all the catalogued objects, thus filtering and parallel processing techniques are presented. The accuracy of the catalogued object data and the propagation of the error over the time identify a position ellipsoid of error, whose behaviour has an important influence on some parameters on the filtering techniques and the way the collision probability is computed. Some collision probability methods are presented.