Drive-by bridge inspection is a method that uses the responses of a test vehicle passing over a bridge to identify the bridge s properties, like road surface profile, modal parameters, damage, etc. It needs only accelerometers installed on the test vehicle, so it is time-and labor-saving and portable. Accurate identification of road surface profiles is essential for drive-by bridge inspection because it excites the vehicle responses the most and may also govern the identification accuracy of the other characteristics. In a laboratory experiment, identifying road profile and frequency utilizing drive-by bridge inspection was successful. However, the accuracy was a remaining open problem. This study aims to improve the accuracy of road profile identification from the responses of a test vehicle passing over a real bridge by correcting the inclinations of sensors on the vehicle. The initial inclinations were estimated by an existing method and their reliability was evaluated by the ratio of standard deviation and correlation coefficient between horizontal and vertical accelerations. From the result of a field experiment, an improvement of roadway profile identification was observed.