Computer vision techniques have been widely applied in Intelligent Transportation Systems (ITSs) to automatically detect abnormal events and trigger alarms. In the last few years, many abnormal traffic events, such as illegal parking, abandoned objects, speeding, and overloading, have occurred on the highway, threatening traffic safety. In order to distinguish illegal parking and abandoned object events, we propose an effective method to classify these types of abnormal objects. First, abnormal areas are detected by feature point extraction and matching. The transformation relation, between the world and image coordinate systems, is then established by camera calibration. Next, different-height inverse projection planes (IPPs) are built to obtain the inverse projection maps (IPMs). Finally, the 3D information describing the abnormal objects is estimated and used to distinguish illegally parked vehicles and abandoned objects. Experimental results from traffic image sequences show that this method is effective in distinguishing illegal parking and abandoned objects, while its low computational cost satisfies the real-time requirements; furthermore, it can be used in vehicle classification.