Abstract: Mobile phone network data has been wildly used in origin-destination (OD)survey.To extract OD data from cellular data,there are two key steps:identifying travel endpoints;and matching them with traffic analysis zones (TAZs).The assumptions about cellular signal coverage in existing studies are quite different from the reality,which leads to complex analysis processes with low accuracy.Two improved methods were proposed to address such issues, based on the analyses about the features and application conditions of traditional TAZ and cell clustered TAZ.For tradi-tional TAZ,the matching accuracy of travel endpoint location is increased by identifying the maximum coverage areas of base stations,and by excluding those areas that are inaccessible using the maximum travel distances.This method is found to be effective for improving the matching accuracy of a portion of travel origin/destination points.For cell clustered TAZ,the cluster process is improved by removing less important steps.Then it is found that the method becomes simp-ler,but the matching accuracy remains to be very similar.