BRENNEN

Industries / Construction and property / Pest control

Traps and monitoring

Devices that catch pests or count them over time, from spring traps and scented lures to sensors and cameras that report activity without a site visit.

5

Milestones achieved, from 1894 to 2022.

2

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1894

    A spring snap trap

    William Hooker of Abingdon, Illinois, was granted US Patent 528,671 for an "animal-trap" on 6 November 1894, after filing on 14 March 1894. A strong spring drives a wire jaw down when an animal presses a baited trigger, and the patent says "the slightest pressure on the trigger will cause the springing of the trap." Snap traps are still advised for small rodent populations: a 2022 Purdue University guide calls them effective but says they "require labor to set and maintain."

    patents.google.comextension.purdue.edu

    Credit

    William Hooker

  2. 1971

    Moth scent becomes a lure

    U. E. Brady, J. H. Tumlinson, R. G. Brownlee and R. M. Silverstein isolated cis-9, trans-12-tetradecadien-1-yl acetate from female Indian meal moths and almond moths, two pests of stored food. They found it was the main, if not the only, part of the sex pheromone, the scent signal, that female Indian meal moths use to draw males (Science, February 1971). Commercial lures now bait sticky traps that catch the males. In a 2025 study, USDA scientists and the trap maker Trécé monitored moths in 33 retail pet supply stores with pheromone-baited sticky traps and counted an average of 40.2 moths per trap per month before treatment.

    LessonA pheromone trap shows where moths are and whether numbers rise or fall, but its catches are "not necessarily linearly related to population density," so a count is not a census.

    ebi.ac.ukpmc.ncbi.nlm.nih.gov

    Credit

    U. E. Brady, J. H. Tumlinson, R. G. Brownlee and R. M. Silverstein (identification); James F. Campbell and colleagues, USDA Agricultural Research Service (2025 study)

  3. 2000

    Zappers scatter fly germs

    Electrocuting insect traps lure flying insects with light onto a high-voltage grid that breaks them apart, and the authors note that homeowners and food handlers use them widely. At Kansas State University, James E. Urban and A. Broce loaded house flies with a marker bacterium or a virus that infects bacteria, then let the flies be electrocuted in a room with such a trap. Flies sprayed with the microbes released one of every 10,000 into the room, and flies that had eaten them released one in 1,000,000 (Current Microbiology, October 2000). The authors concluded that these traps "could play a role in the spread of infectious disease agents."

    LessonWhere food is handled, how a trap kills insects matters, not only how many it kills.

    link.springer.com

    Credit

    James E. Urban and A. Broce, Kansas State University

  4. 2011

    Cups find bed bugs early

    The Climbup Insect Interceptor is a passive monitor: a pitfall trap with no lure, into which bed bugs fall. Rutgers entomologists compared it with three traps baited with attractants and with visual inspection in occupied apartments. In lightly infested apartments the Interceptor, left for seven days, caught as many bed bugs as a dry-ice trap left for one day and more than two commercial lure monitors left for one day, and it detected small numbers of bed bugs more often than visual inspections did (Journal of Economic Entomology, 2011). In 2025 an industry entomologist put the cost of such pitfall traps at about $4 each, with 10 to 14 days needed for a result.

    ebi.ac.ukicup.org.uk

    Credit

    Changlu Wang, Wan-Tien Tsai, Richard Cooper and Jeffrey White, Rutgers University

  5. 2022

    Sensor traps report catches

    Rentokil says it launched Mouse Alert, a mouse trap that reported captures over a wireless link, in 1992, and that its internet-connected PestConnect devices followed; its 2025 paper gives 2004 in its summary and 2014 in its introduction. Purdue University tested Bell Laboratories' sensor-equipped Motomco iQ traps against ordinary ones in a laying-hen barn from October 2021 to January 2022. Servicing 12 sensor multi-catch traps took 5.4 minutes on average, against 10.5 minutes for 12 ordinary traps, and the sensor traps caught 20 mice against 5. For bait stations the time saved was 6%, and two sensor stations were removed for technical problems.

    LessonA sensor records that something entered a trap, not what was caught; Purdue notes that its "events" do not match mice one to one, so alert counts are not pest counts.

    extension.purdue.eduicup.org.uk

    Credit

    Carl Kroger, Jason Fields and Darrin Karcher, Purdue University (test); Bell Laboratories (traps); Rentokil (early connected traps)

Watch for

  1. AI cameras confirm rodents

    Rentokil's PestConnect Optix, added to its system in 2023, is a camera that photographs movement, using infrared light at night, and sends each image to a machine learning model that decides whether a rat or mouse is in it. People checked 5,750 images drawn at random from 117,408 taken from April to June 2024. By Rentokil's count the model was right 92.3% of the time, caught 94.3% of the images that showed rodents, and 87.7% of its rodent alerts were real. By the end of 2024 more than 2,600 cameras were sending over 100,000 images a month. The authors call the work "not a formal Study".

    Achieved whenan outside party tests the cameras against continuous video in customers' buildings and publishes the share of rodent visits caught and the share of alerts that are real.

    icup.org.uk

    Claimed by

    Rentokil Initial (Manuel García-Howlett and colleagues), in the ICUP 2025 proceedings, June 2025

  2. Continuous video watches for rats

    Semco's Pest-Vision R-type records infrared video all the time and flags moving objects it judges to be rodents, instead of waking on a motion sensor. In a two-day test in a ceiling space of a pub in Osaka, checked against 24-hour reference video, it flagged 29 of 33 rat visits with one false alert, while a motion-triggered sensor camera (Bushnell TrophyCam) caught 2. Over about five days above the pub's dining area it caught 59 of 70 visits but raised 971 false alerts, set off by fluttering posters, people, light reflections, insects and dust.

    Achieved whenan outside test in several kinds of buildings reports, for each site, the share of rodent visits caught and the share of alerts that are real.

    icup.org.uk

    Claimed by

    Semco Co. Ltd. (Yuji Kawatake and colleagues, with Chow-Yang Lee of the University of California, Riverside), in the ICUP 2025 proceedings, June 2025

Open

  1. 01

    Test AI rodent cameras

    A side-by-side field test of the AI rodent cameras that pest control companies offer, a motion-triggered trail camera and an open model, all scored against continuous infrared video reviewed by people, in several kinds of building. Release the labeled footage.

    The brief
    Credit · unclaimed
  2. 02

    Check sensor trap alerts

    An independent test of the connected rodent traps sold for use in buildings that checks every alert, and every quiet day, against a camera clip and a person's look in the trap. Report missed catches, false alerts, delays and dropped connections, brand by brand.

    The brief
    Credit · unclaimed

Open 01 · the brief

Test AI rodent cameras

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A side-by-side field test of the AI rodent cameras that pest control companies offer, a motion-triggered trail camera and an open model, all scored against continuous infrared video reviewed by people, in several kinds of building. Release the labeled footage.

Why it is still open

The only accuracy figures come from the makers: Rentokil's 2025 paper on its own cameras, and Semco's two-site test of its own system, in which the share of real alerts fell from 29 of 30 in a quiet ceiling space to 59 of 1,030 above a busy dining area. The only public indoor rodent image set found, on Hugging Face (April 2026), is synthetic, generated by software rather than filmed. A 2026 review of AI for pest management says such systems should "undergo prospective external validation". Searches of GitHub, Hugging Face, Zenodo and Europe PMC found hobby rat detectors and outdoor wildlife camera sets, but no independent test indoors.

icup.org.ukhuggingface.coebi.ac.uk

Built with what exists

Two or more commercial camera services, leased through pest control contractors; a trail camera such as a Bushnell TrophyCam; a continuous recorder, such as a Raspberry Pi with a NoIR camera module and infrared lights, or a CCTV recorder; CVAT or Label Studio for labels; MegaDetector and Ultralytics YOLO11 as open baselines; optionally a roaming indoor camera such as Ring's Rover Cam to check spots between fixed cameras (see the Note "A camera on patrol").

  1. First step

    With the owner's and the pest contractor's permission, set one continuous infrared recorder beside an existing camera in a building with known rodent activity, and label one week of footage.

  2. Done looks like

    Three sites of different kinds, for example a restaurant kitchen, a warehouse and a basement, four weeks each, with at least 300 rodent visits confirmed on the reference video. For each system and site, publish the share of visits caught and the share of alerts that were real, and release at least 10,000 labeled frames under an open license.

Open 02 · the brief

Check sensor trap alerts

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An independent test of the connected rodent traps sold for use in buildings that checks every alert, and every quiet day, against a camera clip and a person's look in the trap. Report missed catches, false alerts, delays and dropped connections, brand by brand.

Why it is still open

Purdue's 2022 test of Bell's iQ traps measured service time and catches but not whether alerts matched captures, and it warns that a sensor "event" is not one mouse. Notz and colleagues (Wildlife Research, 2017) scored the ERMINEA infrared monitoring system for rodent detection, fitted to multiple-capture live traps, against direct observation: it recorded 98.7% of captures correctly in the field, but crickets set off 20% to 70% of traps in laboratory trials. That test was outdoors and in enclosures, with voles and wood mice. Martin (Biological Invasions, 2022) found two cellular trap monitors 100% reliable in reporting that a trap door had closed, on cage traps for invasive animals outdoors. Searches found no independent test of connected snap or multi-catch traps inside buildings that scores alerts against what was caught.

extension.purdue.eduopenpolar.nolink.springer.com

Built with what exists

20 to 40 sensor-equipped snap or multi-catch traps from at least two makers; a small motion camera on each trap or time-lapse cameras over trap lines; a spreadsheet or SQLite log; a Python script to match alerts to clips.

  1. First step

    Buy five sensor traps of one brand, set them unbaited in an office, trip each one 20 times with a stick, and log which alerts arrive and how fast.

  2. Done looks like

    40 traps over eight weeks in buildings with rodent activity, every alert and every catch matched to camera footage, and published rates for each brand of alerts with no catch, catches with no alert, median delay from catch to alert, and connection failures.