License Plate Capture and Analysis
Description
The AI is intended to solve the problem of time-consuming manual reviews of license plate images and data, which makes it challenging for investigators to identify relevant vehicle movements and patterns.
Detailed example
The system processes images and metadata from ICE-owned and commercial license plate recognition cameras. It uses computer vision and optical character recognition to detect and read license plates and to capture associated information such as time, location, vehicle make and model, color, and visible characteristics like damage or signage. An integrated natural language interface powered by a large language model allows users to ask questions in everyday language, such as requesting detections of a particular plate or vehicle description over a period of time. The system converts these questions into structured database queries and returns relevant records, along with concise text summaries of vehicle movements. The system’s AI-enabled outputs are machine-read license plate numbers with associated time, location, and vehicle metadata, as well as natural language search results and summaries produced by the LLM interface. The LLM translates user questions into structured searches over the LPR data and summarizes relevant vehicle detections into concise descriptions of vehicles and their sightings. While license plate information can be used as a link to other personally identifiable information, the LPR system does not automatically link license plate records to driver or vehicle registration databases. Any such queries must be conducted separately in accordance with applicable laws and policies.
AI / analytics pattern
Computer Vision: AI that processes and interprets visual data (e.g., images and videos).
Automation level / stage
c) Deployed – The use case is being actively authorized or utilized to support the functions or mission of an agency.
Expected benefit
The AI capabilities reduce the need for manual review of large numbers of license plate images and logs. By streamlining plate reading and providing flexible search and summarization tools, the system helps investigators more quickly identify potentially relevant vehicle movements and patterns that might otherwise be missed, thereby improving the efficiency and effectiveness of investigative work.
Controls / human review
ATO: No; PIA: https://www.dhs.gov/sites/default/files/publications/privacy-pia-ice-lpr-january2018.pdf
Data needed
The vendor trained its LPR system using a combination of real-world traffic camera footage, synthetic plate images, and public datasets containing diverse license plate formats from various regions. The models are optimized for high accuracy in different lighting, weather, and motion conditions, and are fine-tuned using data from deployments across cities and agencies.