Abstract: In this paper, we use smart phone to process the video in real-time for vehicle detection and distance measurement. Two steps of vehicle detection method is being applied, which are finding hypothesis region and verification of the region. The method uses two steps of extracting the threshold of vehicle shadows based on lane prior, screens out the good shadows in view of lane prior and verifies the shadow by using Adaboost classifier based on haar-like features. Then the video streaming information is being used for rejecting the non-vehicle region and tracking the vehicle. Perspective geometry is being used for distance measurement during that all a user need to provide is the height of the camera, which can be applied to early warning for risk.