{"resourceId":"caltrans-d5-ai-roadway-user-cameras-ctc-2026","versions":[{"version":"external-a124e2feac3502089578f337441ebfb299a57200f95f9c93c05633b124dcb5ad","resource":{"id":"caltrans-d5-ai-roadway-user-cameras-ctc-2026","title":"Innovations in Transportation: Artificial Intelligence Cameras in Caltrans District 5 (CTC information item 4.4)","organization":"California Transportation Commission; California Department of Transportation (Caltrans) District 5","sector":"State DOT","geography":"California, US","publishedAt":"October 2, 2026 (for the October 15-16, 2026 CTC meeting)","publicationDate":"2026-10-02","eventDate":"2026-10-15","sourceName":"California Transportation Commission meeting book","sourceLabel":"CTC memorandum (PDF)","sourceUrl":"https://catc.ca.gov/-/media/ctc-media/documents/ctc-meetings/2026/2026-10/17-4-4-a11y.pdf","evidenceClass":"government-evaluation","outcomeClass":"effective","topics":["infrastructure","data-security","operating-model"],"finding":"Caltrans District 5 piloted an AI camera on State Route 154 from July 2023 that detected, classified and counted most roadway users (pedestrians, bicyclists, cars, trucks, buses, motorcycles) with a stated accuracy of 95 percent or greater. A SHOPP Minor B project then installed 10 more AI cameras in February 2025 at high pedestrian and bicyclist locations, providing count data not previously available. The district is exploring making AI cameras a standard feature of future camera and signal installations; the cameras do not read plates or collect PII.","sledRelevance":"A state DOT moving a narrow, privacy-limited AI use (multimodal counting) from pilot to funded capital deployment and possible design standard is a transferable path for other DOTs' active-transportation data programs.","evidence":"Agency memorandum summarizing a commission presentation. The 95 percent accuracy figure has no stated method, sample size, ground-truth process or per-class breakdown; cost and staff-time benefits are asserted, not quantified.","architectureImplications":"Edge AI cameras feed count data into district analysis; standardizing them in SHOPP camera and signal projects would make AI detection part of field-device specifications rather than a separate system.","governanceImplications":"Use limitation (counting only, no plates or PII) is stated up front, a useful model for scoping AI camera purpose before expansion.","securityPrivacyImplications":"The no-plate, no-PII design limits privacy exposure; agencies should still confirm video retention and whether raw imagery leaves the device.","caveats":"Self-reported by the agency; single-district experience; no accuracy methodology or per-mode error rates; benefits are described qualitatively.","streamIds":["transportation"],"roles":{"sales":"This is a low-controversy entry point for state DOT AI: counting and classifying road users, without plates or PII, funded through an existing capital program. Conversations should focus on how other districts or states could standardize the approach and validate data quality, not on unquantified cost savings. A bounded offer is an accuracy validation and data-integration review before AI cameras become a design standard. Do not repeat the 95 percent figure as a benchmark without the method behind it.","engineering":"The engineering question is how edge AI counting cameras fit into existing camera and signal installations and how their counts flow into traffic data systems. Validate per-class accuracy against manual counts across lighting, weather and occlusion, since the memo gives no method. Confirm device management, firmware updates and whether raw video is retained. If AI cameras become standard SHOPP equipment, specifications need accuracy, data-format and privacy requirements written into them.","delivery":"Delivery work is turning a district pilot into a repeatable standard: define accuracy acceptance tests, data pipelines into count programs, and maintenance responsibilities. Interpretation: the memo shows deployment, not measured cost or safety outcomes, so delivery plans should set baselines for manual-count effort and data coverage before expansion. Ownership sits with district ITS maintenance and operations, with statewide traffic data teams consuming the counts."},"roleDetails":{"sales":{"customerProblem":"Getting reliable pedestrian and bicyclist counts on the state highway system without manual counts, and deciding whether AI cameras should become standard equipment.","stakeholders":["Caltrans District 5 ITS Maintenance and Operations","Caltrans traffic data and safety programs","SHOPP project managers","California Transportation Commission"],"discoveryQuestions":["How was the 95 percent accuracy measured, and for which user classes?","What would a statewide AI camera specification require?","How are counts integrated into existing traffic census programs?"],"valueHypothesis":"Validated AI counts could expand multimodal data coverage at lower marginal cost when cameras are installed through planned capital projects.","potentialEngagement":"A bounded accuracy-validation and specification review for AI counting cameras ahead of wider SHOPP adoption.","unsupportedClaims":["Quantified cost or staff-time savings","Per-mode accuracy beyond the single 95 percent figure","Safety outcomes from the count data"]},"engineering":{"fit":"Good fit for data validation, specification and integration work; the camera technology itself is already deployed.","architectureIntegration":"Edge AI cameras at signals and roadside sites producing classified counts consumed by district and statewide traffic data systems.","prerequisites":["Manual ground-truth counts for validation","Device and data inventory for the 11 cameras","Data format and integration requirements"],"constraints":["Only one district has piloted the technology","Caltrans and California privacy requirements apply","Capital-program timelines govern rollout"],"security":"Confirm no plate or PII capture is enforced on-device, video retention limits, and network segregation of field devices.","proofOfValue":["Per-class accuracy against manual counts at several sites","Accuracy in night and adverse weather conditions","Comparison of staff hours per count site before and after"]},"delivery":{"work":"Develop acceptance tests, data integration and a draft equipment standard for AI counting cameras.","dependencies":["District 5 field data access","Traffic census program integration","SHOPP project schedules"],"ownership":"Caltrans District 5 ITS Maintenance and Operations, with statewide traffic data programs as data consumers.","skillsAdoption":"Train traffic data staff to interpret AI-classified counts and their error bounds.","governanceCheckpoints":["Privacy and use-limitation review before expansion","Accuracy validation sign-off","Specification approval before inclusion in SHOPP projects"],"acceptanceCriteria":["Proposed: per-class accuracy documented at every new site","Proposed: count data available to statewide programs within an agreed latency"],"risks":["Undocumented accuracy across conditions","Scope creep beyond counting","Maintenance burden for distributed devices"]}},"retrievedAt":"2026-10-30T22:10:14-05:00","enrichedAt":"2026-10-30T22:18:05-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Reduces manual count work and staff exposure on roadways; data on pedestrians and bicyclists supports safety planning for vulnerable users.","procurementImplications":"Funding via SHOPP Minor B shows AI cameras can ride existing capital programs; making them a standard feature would shift procurement to equipment specifications.","operatingModelImplications":"District ITS maintenance and operations owns the technology; statewide expansion would need a model for sharing lessons across districts.","sourceVerification":{"openedUrl":"https://catc.ca.gov/-/media/ctc-media/documents/ctc-meetings/2026/2026-10/17-4-4-a11y.pdf","referenceExcerpt":"The cameras are used only for traffic counting and classification purposes and do not read license plates or collect personally identifiable information.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}