Dynamic tolerance setting method for plane sweep algorithm

In this paper, we analyze the characteristic of floating point numbers computation in scientific computing, and adopt the Gaussian error propagation rule to analyze the computational error. By this method, an appropriate double value (i.e. the dynamic tolerance) is computed in advance and set for the plane sweep algorithm to judge whether two floating point numbers equal. By this method, we can get the segments passing the event point correctly. The dynamic tolerance guarantees the correctness of the plane sweep algorithm despite the floating computational errors. The application based on the plane sweep algorithm indicates that the method works correctly and efficiently for general use.