24-0293-Intersection_Video_Analytics_Specs-Scope.pdf

Intersection Video Analytics Platform

Specifications

Objective

The County seeks a vendor(s) to provide Intersection Video Analytic (IVA) units that use highly flexible machine learning frameworks (AI) to detect and classify road users (vehicles of all kinds, pedestrians, bicyclists, and other mobility devices) in areas of interest, primarily in the vicinity of signalized intersections.

The unit must be able to process video feeds and provide a suite of outputs including data analytics, real-time alerts with photo evidence, detector outputs to a traffic signal controller, and be able to be customized to detect specific operating conditions such as stopped vehicles, driving directions and specific vehicle classes.

In addition, the unit must be able to be customized to count the number of passing vehicles/road users.

The unit supplied must consist of all associated equipment and processing software required to setup and operate in a field environment.

Requirements

• Standard Services Agreement - The awarded contractor will be required to execute the attached standard services agreement. No exceptions to the standard services agreement will be acknowledged. The awarded firm must sign as presented in this bid document.

• Bidder shall be an authorized dealer/seller of the product(s) offered. Bidders shall provide documentation signed by the manufacturer as proof bidder is an authorized dealer or reseller and shall provide with bid submittal.

• Disclose the manufacturer, the standard warranty and any extended warranty.

• A copy of the warranty, product bulletins or other related information shall be attached with the Section F – Bid Summary submittal.

Scope

Intersection Video Analytic (IVA) Unit Requirements

Capabilities

• The IVA unit must be able to detect, count and classify road users:

o Vehicles of all kinds (cars, trucks, transit, etc.)

o Pedestrians o Bicyclists o Mobility devices (Wheelchairs, scooters, etc.)

• The IVA unit must be able to be configured to support data collection of vehicles

(stopped, trajectories, classification, counts) and pedestrians/cyclists, with at least 90% recognition accuracy of counts, location, and movement through the intersection.

• The IVA unit must provide real-time notifications when the AI model detects the following events:

o Wrong Way Driver o Speeding o Red Light Violation o Crosswalk Violation o Near Misses o Crashes

• The IVA unit must be able to send real-time notifications by email, text and in-browser notifications. The event reporting must be able to post at least the date and time, conflict type, road users involved, movement, and direction of travel.

• The IVA unit must be able to utilize video input from any IP, CCTV or Analog cameras to collect video image data for purposes of detecting and classifying vehicles, pedestrians and bicyclists, and generating traffic data and real-time notifications with photo evidence.

• The IVA unit must be able to detect either approaching or departing vehicles in multiple traffic lanes simultaneously.

• The IVA unit must offer bi-directional integration of data including pedestrian, detection, vehicle classifications, and detection of various elements with the traffic signal controller.

Operating Environment

• The unit must meet NEMA TS-2 environmental specifications.

• The IVA unit must be rack or shelf mountable.

• The IVA unit must operate at 120-240 VAC, requiring 30W or less power.

• The IVA unit must be able to communicate via built-in SDLC with traffic controllers.

No third party SDLC adapters shall be used.

• The IVA unit must be able to make use of off-the-shelf hardware that includes edge processors, servers, and cameras.

• The IVA unit must be able to be deployed locally with an edge processor, centrally on premises, or in the cloud.

• The IVA unit must be able to make use of:

o New or existing cameras o Fixed view or PTZ o Rectilinear, panoramic / fisheye or thermal o Analog or network cameras

• The IVA unit must be able to perform multimodal detection (including vehicles, pedestrians, cyclists) in all lighting conditions.

Reporting Requirements

• Operational dashboards must be generated from the data collected. The dashboards should provide a summary of key events detected (near-misses, stopped vehicles, pedestrians on road, key turning movements in the intersection, and vehicle classifications).

• The system must provide visualization of detected event reports with related video and data including:

o Date and Time o Type of event with parameters o Intersection approach o Direction of travel o Video recordings of the detected events

• The system must be able to generate event statistics reports which can be filtered for a user-defined date range:

o Count data by day of the week, in 5, 15, 30, or 60-minute time-intervals, or at a user-defined time interval o Multiple statistical and graphical reports based on user-defined parameters (including line charts, bar charts, tables)

• The system must be able to export raw data and generated reports in common formats including CSV, PDF and PNG.

• The system must be able to generate an aggregated report in the form of heat maps based upon selected event data.

• For Near Miss events, the system must be able to generate an aggregated report for the following:

o By Severity o By Day of Week o By Time of Day o By Movement o By Direction

• The system must be able to fuse sensor data (performing multi-sensor, multiple camera tracking) into a single unified representation, and to suppress duplicate road user detection data.

• The system must be able to enable access for an unlimited number of users.

• The system must be able to provide high frequency health and status checks (including power failures, memory usage, processor usage, communications failure, etc.) and provide system monitoring reports.

• The application software must maintain a database of current and historical traffic data and allow the user to run reports against this data.

• The user must be able to set up all the object detection zones and program real-time notifications from any Windows-based PC.

Training and Support

• Ongoing software support by the supplier must include updates of the application software and AI models. These updates must be provided free of charge during the warranty period or while under an active maintenance contract.

• System must be warranted for a minimum of three (3) years against failure of hardware or software.

• System must include three (3) years of software maintenance and updates.

• Training on installation, set-up, and use of the software must be included.