transport · family: it worked in the lab
cameras that crywolf on the tracks
Most rail trespassers are killed or injured within a quarter mile of a grade crossing, but detection cameras there cry wolf at deer, choke on bandwidth, and lose their deterrent effect by year three
Problem statement
Fifty years of federal crossing-safety funding cut deaths at U.S. highway-rail grade crossings from 888 in 1975 to 197 in 2020, but the railroad death toll has moved rather than vanished: trespassing on the right-of-way, including suicides, now accounts for roughly 70% of all railroad-related deaths — 525 in 2020, up more than 31% over the previous decade — and FRA data indicate about 74% of non-suicide trespass casualties and 73% of suicide attempts occur within 1,000 feet of a grade crossing. Intrusion detection is a mature security technology for utility plants and banks, yet the few systems trialed on rail rights-of-way have failed on the specifics of the setting: outdoor video motion detection and passive infrared both generate false alarms (large animals were a major trigger), live high-definition video needs wireless backhaul that some track locations lack, and the deterrent effect of an installed system decayed after the second year. The unsolved problem is a detection-and-warning system that works at crossing-adjacent track — cheap enough for thousands of sites, accurate enough that police dispatchers do not learn to ignore it, and durable in its effect.
Why this matters
FRA counted 9,363 reported trespassing accidents in 2012–2016 with a societal cost of $43.2 billion in deaths and injuries, plus more than $56 million in train delay; incidents near crossings also cause significant delays to motorists and other road users. The Section 130 program lets states spend crossing-safety funds on hazard elimination that could include detection and warning systems, but decision-makers currently have, in the problem statement's words, "limited, if not misleading, information" on which technologies work under which conditions. Because trespass casualties are concentrated near crossings, a system that works at that specific location type addresses the largest remaining share of rail deaths with a bounded number of installations.
What’s been tried and why it hasn’t worked
Volpe's 2007 state-of-the-art review catalogued infrastructure-based, locomotive-based, and cooperative intrusion-detection concepts (e.g., the AWARE wireless intrusion system and Nestor's Rail CrossingGuard for gate violations), predating usable computer vision. Volpe's 2012 evaluation of a video-based trespass monitoring and deterrent system on a railroad bridge in Pittsford, New York logged nearly 4,000 events (about 4.6 per day), cut trespassing 60% in the second year, but the reduction did not persist into the third — and large animals triggered many alarms that could not be masked without weakening detection. Volpe's Brunswick, Maine study with the local police (reported 2020) streamed live video from fixed cameras and found that both video motion detection and passive infrared produced false alarms, that wireless bandwidth for HD streams was a constraint at some sites, and that TV-white-space transceivers had limited range. Proprietary radar-based platform and level-crossing intrusion systems (e.g., Honeywell's) and worker-warning systems (Protran/Harsco) exist, but NCHRP staff noted the program "cannot evaluate the performance of commercial or proprietary products," and the statement says the effectiveness of new technologies for detecting trespassers near crossings has "not been thoroughly evaluated." NCHRP funded only a $125,000 scoping study (17-122, 2024–2025) to catalogue practice and research needs; the evaluation itself remains undone.
What would unlock progress
The core technical unlock is on-device classification that separates people from animals, vehicles, and vegetation motion with a low enough false-alarm rate that alerts can go straight to a dispatcher — modern edge vision and thermal/radar fusion make this plausible where 2012-era motion detection could not — combined with event-only transmission so sites without broadband can still report. The behavioral unlock is understanding why deterrence decays: an audible warning that habituates by year three needs escalation, variation, or coupling to enforcement and to path redesign (the statement notes trespass often follows the most direct route to a destination). The adjacent solved problems are wildlife-camera trap classification and perimeter security for solar farms, both of which have driven false alarms down with edge ML on cheap hardware.
Entry points for student teams
A team could assemble a public or partner video dataset from crossing-adjacent track (or stage one on a heritage/short-line railroad with permission), train an edge classifier for person-on-ROW vs. animal/other, and quantify false-alarm and miss rates by lighting and weather against the Volpe baselines. A systems team could design a low-bandwidth, solar-powered node that sends only classified events and short clips over LTE or LoRa and cost it against the Section 130 funding constraint. A human-factors team could design and pilot-test a warning protocol that resists habituation and evaluate it with a before/after trespass count. Relevant skills: computer vision, embedded systems, human factors, and rail safety regulation.
Genes
Genes marked “+” were added in a later review.
Structural cousins: stuck the same way, in other fields
Sources
"Evaluation of Trespassing Detection and Warning Systems in the Vicinity of Highway-Rail Grade Crossings," NCHRP FY2023 Problem Statement 2023-G-27 (Shanshan Zhao, UConn; Jeff Warner, TTI; Ian Savage, Northwestern; Steve Laffey, Illinois Commerce Commission; Greg Orrell, MITRE; Robert Rescot, ARA; submitted by Indiana DOT), with NCHRP evaluation, in *NCHRP FY 2023 Program: Compendium of Problem Statements* (TRB, February 2022), and accessed 2026-08-17; NCHRP Project 17-122 "Strategies and Technologies for Warning, Detection, and Prevention of Trespassing in the Vicinity of Highway-Rail Grade Crossings: State of the Practice and Research Needs" (Oklahoma State University, completed 2025), accessed 2026-08-17 go to source 1 ↗ go to source 2 ↗ go to source 3 ↗
Working notes
Notes made while writing and checking this brief, shown as written. They use working shorthand.
The problem statement was authored by a multi-institution rail-safety group and submitted by Indiana DOT; the Volpe results (2007, 2012, 2020) are as summarized in the statement — the underlying FRA/ROSAP reports were blocked to the drafting session (HTTP 403); at verification the Pittsford figures (~4,000 events, >60% reduction in Year 2, not sustained in Year 3) were confirmed against the FRA/Volpe report abstract (TRID 795856) and the 74%/73%-within-1,000-ft, 888→197, 525, 31%, $43.2B and $56M figures against the G-27 statement text; the NCHRP staff "cannot evaluate ... proprietary products" quote was confirmed in the compendium PDF. NCHRP 17-122's TRB project page confirms only a scoping study was funded; its final report link on nationalacademies.org returned 404 on 2026-08-17. `stakeholders:multi-institution` was considered (the railroad owns the right-of-way, the road authority controls the crossing and Section 130 funds, and local police must respond to alerts — the statement itself flags "the roles and responsibilities of the roadway authority" as an open task) but at second coding it fails criterion (2) of the three-criteria test: the binding constraint is technical (false alarms, bandwidth, habituation), and a single well-resourced group could build and validate the detector, so `stakeholders:institutional` is used (deployment needs railroad/road-authority/police buy-in). `constraint:coordination` was NOT applied for the same reason. `temporal:worsening` was considered because trespass deaths rose 31% in ten years, but the source names no causal mechanism, so `static` is used pending better evidence. `constraint:behavioral` refers to the decaying deterrent effect. Related collection briefs: `transportation-ptc-restricted-speed-gaps` and `transportation-rail-bearing-detection` are rail but concern train control and equipment, not trespass. Verified at intake 2026-08-17: gate (net) + adversarial source check + contested-tag second coding.
Source type: Self-articulated (rail-safety researchers and a state DOT articulating an evaluation gap in technology they must choose among)
Reconciliation 2026-09-29: The whole write-up was checked against its three cited sources, all fetched on 2026-09-29 and read in full: the G-27 problem statement (apps.trb.org/nchrpballoting/BallotingDocs/G-27prob.htm), the G-27 pages of the compendium PDF (onlinepubs.trb.org, pages G-27/1 to G-27/7, including the NCHRP, FHWA and AASHTO evaluations and the submitter's response), and the NCHRP 17-122 project page (apps.trb.org/cmsfeed/TRBNetProjectDisplay.asp?ProjectID=5351). The TRID record named in the note above (trid.trb.org/view/795856) was also fetched. Seven changes were made. (1) The title said most rail trespassers are "Killed" within a quarter mile of a crossing. The statement's 74% figure counts deaths and injuries together, so the title now says "Killed or Injured". (2) The text said casualties cluster near crossings "because crossings are where people can get onto the tracks." None of the sources gives a reason for the clustering, so that clause was removed. (3) The text said "many" track locations lack the wireless capacity for live video. The statement says it "can be a challenge for some locations", so the word is now "some". (4) The text said FRA counted 9,363 trespassing "casualties". The statement says 9,363 "reported trespassing accidents", and the text now uses that wording. (5) The text said each strike "stops trains and closes crossings for hours". The sources do not give a duration or mention closed crossings. They say these incidents cause significant delays to motorists and other road users, and the sentence now says that. (6) The text said decision-makers have poor information, "so the money is not spent on them." The sources do not say whether Section 130 money is or is not spent on detection systems, so that clause was removed. (7) The text said NCHRP cannot evaluate proprietary products, "so no independent comparative evaluation has been done." The NCHRP reviewer's comment was about what this project could cover and does not say why evaluations are missing. The sentence now gives the statement's own words, that effectiveness has "not been thoroughly evaluated." Confirmed against the sources and left as written: the Source line's title, six authors, Indiana DOT as submitter, and the compendium date of February 2022; 888 deaths in 1975 and 197 in 2020; roughly 70% of railroad deaths; 525 deaths in 2020 and the rise of more than 31% in ten years; 74% and 73% within 1,000 feet, which the statement itself gives as under a quarter mile; $43.2 billion and more than $56 million; the quoted phrase "limited, if not misleading, information"; the Section 130 description; the 2007 review with AWARE and Nestor's Rail CrossingGuard; the Pittsford figures of nearly 4,000 events, 4.6 per day and a 60% reduction in Year 2 that did not continue into Year 3; large animals setting off alarms; the Brunswick, Maine findings; Honeywell and Protran/Harsco; the NCHRP quote; and for 17-122 the title, Oklahoma State University, $125,000, a start in March 2024 and completion in April 2025. The TRID record confirms that the Pittsford system warned trespassers through speakers. NOT checked, and left as written: the Volpe reports themselves could not be opened (railroads.dot.gov and rosap.ntl.bts.gov returned 403), so the word "Deer" in the title, where the statement says only "large animals", and the idea that the warning lost its effect through habituation could not be checked. The statement dates the Pittsford study to 2012, and the write-up follows it. The TRID record gives a publication date of August 2006 and says the system was evaluated from August 2001 to August 2004. This difference was not resolved, and it affects the phrase "2012-era motion detection". The TRID record as fetched today does not contain the event count or the reduction figures; those were confirmed in the statement instead. The 17-122 final report link still leads to a file-not-found page, so "the evaluation itself remains undone" rests only on the project page's description of the study's scope. The sentence about wildlife camera traps and solar-farm security is the write-up's own comparison and is not in the sources. Genes untouched.