To avoid traffic accidents, parked and abandoned objects must be detected quickly and precisely. Most detection techniques employ 2D image attributes like the area of different target types and properly generate the background model. Background influences these algorithms, and target kinds are inaccurate. Thus, this study provides a parked and abandoned item recognition technique that employs accurate 3D target information to discriminate target types. First, state evolution detects anomalous areas. Second, it tracks the initial anomalous area two-way and utilizes the eight-neighborhood seed-filling technique to partition the parked and abandoned item area. Finally, it distinguishes parked from abandoned objects using three ways. The first technique calculates the relative height between feature sites by comparing their projection velocities. Then, the height distinguishes parking from abandoned objects. The second technique uses 3D model appropriateness to differentiate parked and abandoned objects. The third approach calculates the inverse projection map by setting the reverse projection planes at different heights in the area. Then, the inverse projection maps' heights establish the goal length, breadth, and height, distinguishing parked and abandoned objects. Tunnels, highways, urban expressways, and country roadways tested the algorithm. The system successfully detects parked and abandoned objects with a low miss and false detection rates. It was also real-time.