In order to improve the accuracy of the drowsiness detection method using the data of the drive recorder, we examined the data preprocessing method. In a drive recorder with a lane departure warning (LDW) function, lane line detection information is acquired by that function, and the accuracy of lane line detection is important when estimating drowsiness during driving based on the lateral position information in the driving lane. Lane changes and junction passage are cited as disturbance factors that deteriorate accuracy, and these were extracted as low-reliability sections from the characteristics of changes in the lateral position in the lane. When the results were compared with the recorded video, it was confirmed that they were effectively removed. In addition, it was shown that it is useful for improving the estimation accuracy by comparing the drowsiness estimation results.