Counting Empty Parking Spots at Truck Stops Using Computer Vision

Modi, Pushkar; Morellas, Vassilios; Papanikolopoulos, Nikolaos P. · 2011 · ROSA P / University of Minnesota

archive: archived pipeline: cataloged verified

Get this paper ↗ (full text — opens at the source; we link to it, we don't host it)

Summary

This report addresses the problem of truck driver fatigue and the perceived shortage of parking at truck stops, which contributes to approximately 40% of heavy truck accidents. Despite this, only 53% of truck stop spaces are occupied on any given night, indicating a mismatch between driver perception and actual availability. The authors propose an automated truck stop management system using computer vision to detect, classify, and localize vehicles in real-time. This data would be used to update variable message displays located 30–40 miles before stops, directing drivers to available spots and optimizing stop utilization. The study argues that computer vision offers advantages over existing technologies like inductive loops, including lower cost, non-intrusive installation, and the ability to monitor individual spot occupancy rather than just entry/exit counts. The researchers developed a framework involving cameras mounted 30–60 feet above parking areas. They evaluated two foreground detection methods: Layering and Mixture of Gaussians (MoG). Layering proved computationally expensive and struggled with similar-colored vehicles. MoG was selected for its adaptability to lighting changes and lower computational cost, though it required a shadow removal technique to prevent shadows from falsely marking adjacent spots as occupied. Due to the inability to obtain video from actual truck stops, the team collected 20 hours of video data over three days from a university parking lot using a specialized trolley-based camera system called TIM. The system utilized a calibration tool to define parking spot coordinates and an analysis tool implemented in C++ with OpenCV to determine occupancy based on the percentage of a spot occluded by foreground objects. The implementation faced challenges such as sudden illumination changes (e.g., lens flare) and occlusions by pedestrians or vehicles moving through spots. The team addressed illumination issues by tracking mask persistence over time, ignoring changes lasting less than 60 frames. Pedestrian interference was resolved by implementing blob tracking, which ignored objects below a size threshold calculated based on sedan dimensions. In testing, the system without blob tracking achieved 44% accuracy, with nine false positives and four false negatives. After implementing blob tracking, the system achieved 100% accuracy on the test set, with zero false positives or negatives. The report concludes that the system is viable for deployment, with future work focusing on trajectory tracking to handle vehicles moving across multiple spots and gathering empirical results under varied weather conditions.

Key finding

The computer vision system achieved 100% accuracy in detecting parking occupancy events after implementing blob tracking, reducing false positives and negatives to zero compared to 44% accuracy without it.

Methodology

lab_experiment

Sample size: 20

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via bulk_ingest_rosap on 2026-05-23 (6 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• distraction detection algorithms