Incident Detection Algorithm Evaluation

Martin, Peter T; Perrin, Joseph; Hansen, Blake; Kump, Ryan; Moore, Dan · 2001 · ROSA P / Upper Great Plains Transportation Institute

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 2001 report by the University of Utah, prepared for the Utah Department of Transportation (UDOT), evaluates incident detection technologies to recommend a strategy for the UDOT Advanced Traffic Management System (ATMS). The research was motivated by the need to minimize response times for freeway incidents, which cause injury, congestion, and economic loss. While Automatic Incident Detection (AID) algorithms have been developed since the 1970s, many suffer from high false alarm rates or poor detection reliability. The study aimed to qualitatively evaluate computer-based algorithms, video image processing, and cellular telephone reporting to determine the most effective combination for the Salt Lake Valley freeway network, which is instrumented with 87 inductive loop detector sets and video cameras. The authors reviewed literature on four categories of computer-based algorithms: pattern-based (e.g., California, APID), catastrophe theory (e.g., McMaster), statistical (e.g., ARIMA, DES), and artificial intelligence (e.g., neural networks). They also examined video image processing and cellular telephone call-ins. Performance was assessed using three metrics: detection rate, time to detect, and false alarm rate. Because inconsistent definitions in existing literature prevented direct comparison, the authors normalized false alarm rates to the UDOT network’s 87 detectors and reported other metrics as cited in source studies. The analysis included both field-tested algorithms and laboratory simulations. The findings indicate that among field-implemented algorithms, the All-Purpose Incident Detection (APID) algorithm performed best, yielding an 86% detection rate, a 2.5-minute time to detect, and approximately eight false alarms per peak hour on the ATMS network. In laboratory settings, neural network methods showed strong potential with an 89% detection rate, one minute to detect, and two false alarms per peak hour. Video image processing demonstrated the highest performance, with a 90% detection rate, 20 seconds to detect, and 0.03 false alarms per hour, offering the advantage of detecting incidents in emergency lanes. However, the study concluded that cellular telephones have become the primary means of incident reporting, often providing faster and more accurate reports than automatic algorithms. The significance of this research lies in its recommendation to shift reliance from computer-based AID to cellular reporting. The authors advise UDOT to enable and calibrate existing APID and Double Exponential Smoothing algorithms as secondary detection tools while prioritizing cellular integration. They recommend implementing a public awareness program for the “Highway Help *11” service and addressing dispatch limitations, as *11 calls are routed to the Highway Patrol rather than emergency medical services. The report concludes that while video and neural network technologies are promising, cellular telephones will remain the dominant method for incident detection due to their speed and accuracy.

Key finding

Cellular telephone call-ins have become the primary means of incident detection, outperforming automatic algorithms in speed and accuracy, while video image processing demonstrates superior performance with a 90 percent detection rate and 20 seconds to detect.

Methodology

review

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.

• dms validation