Traffic Counts from Provided Video: A Scope Assessment Checklist

Before quoting traffic counts from provided video, check whether each required movement stays visible, the recordings cover the requested periods, and reviewers can distinguish the agreed vehicle classes.

Define independent count checks, acceptance rules and output files. Price usable footage and state how you will handle gaps.

A folder containing hours of recordings can still miss a right-turn lane or lose pedestrians behind queued vehicles. Start with a representative screenshot and a short authorized clip from each view, including a difficult period. A screenshot shows coverage; the clip shows whether you can follow movements through that coverage.

1. Define the counting question on a view map

Mark the approaches, count lines, exits and crosswalks on a reference image. Give each camera and movement a stable identifier. For a turning-movement count, reviewers need enough of the path to assign both the approach and the exit. A vehicle appearing near an intersection does not establish which turn it made.

• List left, through, right and U-turn movements separately where required. Define slip lanes, driveways and unusual maneuvers.
• For overlapping cameras, assign one count location or a documented reconciliation rule so the same movement does not enter the total twice.
• Keep pedestrian crossings, bicycle movements and vehicle movements separate. State how the study treats someone walking a bicycle.
• Record unseen lanes and blocked crossing areas as coverage gaps. Do not silently scale a visible lane to represent an unseen approach.

Parking occupancy needs a different setup: a known inventory of visible spaces and an observation schedule. Use the parking-study RFP data worksheet for parking inventory, turnover and lead-consultant responsibilities.

2. Audit recording duration and timestamps

Make a file inventory before estimating review time. Record camera ID, filename, recording date, timezone, start and end times, gaps, resolution and any clock offset. Test an export from the original recorder, rather than assuming a compressed preview represents the delivered files.

Agree on the dates, continuous count windows and output intervals. Separate footage duration from reporting interval: a table of 15-minute movement totals still requires observation of events throughout those intervals. Widely spaced parking snapshots cannot establish all the traffic movements between them.

Check clock alignment across cameras. Specify how reviewers handle a movement at an interval boundary, a file split, daylight-saving changes or a recording that ends during the movement. Mark missing periods as missing, not zero. Any later estimate or expansion needs a separate method and a visible label.

3. Test classes and difficult conditions before committing

Write a class dictionary with examples and an unknown category. Passenger vehicles versus broad truck groups may be a different assignment from detailed axle-based classification. Do not promise a detailed class when the camera cannot show the distinguishing features.

Review clips with queues, buses, glare, rain, low light and camera vibration where those conditions occur. A bus can hide a bicycle or a second vehicle crossing the count line. Record whether another view resolves the obstruction, extra manual review can resolve it, or the movement remains unobservable.

Agree on a decision for each requested measure: supported for the scoped windows, supported with stated exclusions, or not supported by the provided footage. Define a recapture or reduced-scope option before reviewing the full batch. A detector finding vehicles in frames is only one component; valid movement totals also need path assignment and duplicate control.

4. Write the acceptance test before the full count

Ask an independent reviewer to label selected clips without seeing the production counts. Choose clips across cameras, movements, classes and operating conditions, including peaks and difficult views. Retain the selection method and exact clip windows so another reviewer can repeat the check.

• Compare production and reference totals by movement, class and interval. Report missed movements, duplicate counts and incorrect class assignments separately where the reference supports that detail.
• State the error calculation, denominator and treatment of low-volume or zero-count intervals. A small absolute error can produce a large percentage error when few vehicles pass.
• Agree on the allowed error, minimum review coverage and which failures trigger recounting, wider review or rejection of a view. Do not hide a weak approach inside a strong intersection-wide total.
• Keep reviewer disagreements and corrections in a reconciliation log. Hand labels need review too; calling them ground truth does not make an ambiguous movement observable.

The City of Raleigh's Video Traffic Count Services RFP, 274-TR-2026-06, provides a concrete buyer example. It requests conversion of City recordings into traffic reports; Appendix I asks for rates for reports "within 95% accuracy."

That is an issuer-specific requirement, not a universal acceptance standard or a Northline performance claim. A team still needs to understand how the buyer will test the result. See Section 4 and Appendix I in the linked packet, retrieved September 10, 2026.

5. Agree on files the analyst can use

Approve a blank output template before processing the batch. CSV or XLSX tables should include stable site and camera identifiers, date, timezone, interval start and end, approach, movement, class, count and review status. Include a data dictionary and distinguish zero observations from unavailable data.

Provide the view map, file inventory, coverage exceptions and count-check summary alongside the tables. Agree on GIS coordinates and coordinate reference system only if spatial delivery is required. Keep summary charts traceable to the delivered rows; label estimates separately from observed counts.

The Bismarck-Mandan MPO's 2026 Traffic Data Collection and Analysis RFP illustrates why the handoff belongs in scope: its work program requests traffic counts with raw 15-minute data, Excel and GIS deliverables.

Those are that buyer's requirements, not mandatory formats for every video study. The historical packet's proposal deadline has passed; this citation does not identify an open assignment. Retrieved September 10, 2026; scope and deliverables, printed pages 9-11.

6. Estimate the work after the sample review

Separate setup, usable footage review, independent checking, exception handling and final formatting in the estimate. State the number of views and hours included, the class scheme, reporting interval, turnaround and revision allowance. Concurrent views can multiply review effort even when they cover the same study window.

Base the production allowance on the representative clips and the agreed method. Record what changes the price: an additional approach, unreadable exports, more classes, new count dates or an added report type. Gap and delay analysis require their own event definitions and timing checks; do not include them in a turning-count price by implication.

Raleigh estimates 450 six-hour locations over 24 months and specifies City-provided footage without additional hardware or optional field capture. This describes the issuer's estimated assignment, not guaranteed volume or available subcontract work. A footage supplier and a field-capture contractor have different responsibilities.

Copy this scope-assessment checklist

1. Question: Name the site, movements, classes and decision the count supports.
2. Coverage: Attach an annotated view map with count lines, overlaps and unseen areas.
3. Files: List original exports, dates, timezone, continuous windows and missing periods.
4. Feasibility: Review representative authorized clips, including difficult conditions, and document unsupported measures.
5. Acceptance: Specify independent reference clips, error measures, allowed error and recount rules.
6. Delivery: Approve a sample table, data dictionary, map and exception log.
7. Estimate: Separate setup, counting, checking, exceptions and revisions; state the change-order triggers.
8. Authorization: Name who may supply and process the footage, who may access it, and when retained copies must be deleted.

Discuss the question before transferring video

Use recordings your organization is authorized to provide for this purpose. Agree on a secure transfer method, access limits and retention period before uploading files. Keep faces, plates, audio and individual identity out of the requested outputs; minimize or redact incidental personal information in any shared evidence images.

Northline ParkingStudy reviews authorized fixed-camera footage for scoped parking-utilization findings and supporting exhibits. The service includes no live monitoring or identity tracking. Traffic-movement and classification requests need a separate feasibility and scope review. This guide does not promise arbitrary-video counting capability, a particular accuracy rate or engineering certification.

Discuss your video-count scope with Northline. Start with the required measures and recording conditions. Arrange authorization and secure transfer before sending footage.

Practical scope guidance from Northline ParkingStudy, independent of the named agencies. Examples do not establish a client relationship, bid eligibility, award or available subcontract. Professional interpretation and engineering conclusions remain with the responsible consultant. Check current issuer documents and amendments before relying on procurement examples.