A collision prediction method for autonomous vehicles is mandatory. A na?ve approach can be used to calculate the distance between our vessel and each x-object every second. However, if there are a vast number of moving x-objects, it will take time to calculate all of them one by one. In this paper, we propose a hybrid approach to use a TPR-Tree index structure to narrow down the collision prediction only to the nearby x-objects. Therefore, it will minimize the cost and time required to predict collision.