Photos taken at the traffic accident scene preserve the situation of the scene and information of the accident better than any other types of records. However, due to its lacking of the geometric information among objects on the photo, police officers still need to measure the geometric data of every evidence at the scene to make an after-accident situation map. This takes not only more police hours, but also causes more human errors. This study is trying to apply the close-range photogrammety techniques to grab the geometric information on a picture to improve the proficiency and accuracy of reconstructing an accident scene. Based on the perspective grid method, it can be done by simply using an image processing software and digital images of the scene. A sensitivity study of the type of lens of the camera, the height and angle of camera while shooting, the difference of elevation of the surface, and the possible human errors of locating an object on a digital image was conducted and presented in this article.The result shows that an accident scene can be measured by using this modified perspective grid method and achieve the required precision without any extra special equipment or altering any regular procedures to the police work.