Project Objective & Overview

This SBIR Phase I project leveraged artificial intelligence (AI) and machine learning (ML) to transform how agencies assess and improve bicycle and pedestrian infrastructure. The project addressed critical data gaps in Complete Streets planning by:

• Automating extraction of infrastructure features from aerial imagery using computer vision
• Developing comprehensive Level of Traffic Stress (LTS) assessments for both street segments and intersections/crossings
• Creating personalized routing algorithms that account for cyclist level of comfort preferences
• Demonstrating proof-of-concept in Tampa, FL with 21,136 images

This innovative approach reduces data collection costs, standardizes Level of Traffic Stress (LTS) measurements across jurisdictions, and provides insights for infrastructure investment decisions.

Key Results & Findings

Successfully extracted 5 infrastructure features from imagery:

• Roads, sidewalks, crosswalks, bike facilities, and parking
• Achieved accurate detection even with obstructions (trees, buildings)
• Developed methods to measure geometric features from pixel data

Advanced LTS methodology:

• Incorporated movements into LTS calculations (major gap in current methods)
• Created crossing-level stress assessments to complement segment analysis

Personalized routing algorithm:

• Accounts for 4 cyclist comfort levels (Direct, Strong & Fearless, Enthusiast & Confident, Interested but Concerned)
• Outperformed commercial routing for low-stress bicycle trips
• Balances distance, time, and stress factors

Created comprehensive Tampa dataset demonstrating scalability and real-world application