In order to solve the low accuracy problem of route travel time estimation when using low-frequency floating car data, the characteristics of low-frequency floating car data are analyzed from the perspective of travel time distribution estimation. Potential errors are presented and discussed, each of them has an error correction model. Selecting Changshou Road of Shanghai for empirical analysis using floating car data from 1 500 taxies. Travel time under various correction methods would be directly estimated, which compares with estimation results provided by the vehicle license plate recognition device. The results reveal that travel time with all errors corrected improves the average estimation accuracy by 9.5%, and median, 25th and 75th percentile have higher matching with the license plate recognition method. The low-frequency floating car data error correction model can improve the accuracy of the travel time estimation, and provide an effective travel time information.