01516647

Component

https://doi.org/10.3141/2460-10

The number of lanes is a basic input to many transportation studies. Traditionally, these data are either collected in the field or manually extracted from aerial images. These methods of data acquisition are both resource intensive and time-consuming, especially when large study areas are involved. The availability of remotely sensed georectified aerial images provides an inexpensive alternative to acquiring these data via automatic feature extraction. This paper explored a method to automatically extract the number of lanes from high-resolution georectified aerial images. In this method, geographic coordinates of a target roadway were precisely mapped to pixels in aerial images. These pixels were first grouped into fixed-length profiles. The saturation threshold was then applied to exclude profiles covering nonpaved areas such as plants and building roofs. Lane profile candidates were then identified from the remaining profiles with the support vector machine classification technique. The number of lanes was then estimated by optimizing the lane locations according to the lane profile candidates and lane width constraints. The method was tested by using georectified aerial images of Miami–Dade County in Florida. For all six test cases involving two-lane and four-lane suburban roadway segments, the method accurately determined the number of lanes. Test results indicate that it is feasible to extract the number of lanes from georectified aerial images in suburban and urban areas with the method developed in this research.

01559855

14-5547

English

pp 86–96

2014

Transportation Research Record: Journal of the Transportation Research Board

9780309295529

Print

Figures (9) ; Photos; References (16) ; Tables (1)

Aerial photogrammetry; Data collection; Image processing; Remote sensing; Traffic lanes

Feature extraction; Support vector machines

Data and Information Technology; Highways; Planning and Forecasting; I72: Traffic and Transport Planning

TRIS, TRB, ATRI

Jan 27 2014 3:56PM

• Accessible and Practical Geocoding Method for Traffic Collision Record Mapping: Quebec, Canada, Case Study
• Adaptive Learning in Bayesian Networks for Incident Duration Prediction
• Comparison of Sichel and Negative Binomial Models in Hot Spot Identification
• Data Business Plans and Governance Programs: Aligning Transportation Data to Agency Strategic Objectives
• Data Mining and Complex Network Algorithms for Traffic Accident Analysis
• Development of a Visualization System for Safety Analyst
• Development of Alignment-Based Parametric Data Exchange Schema for Bridge Geometry
• Elemental Decomposition and Multicriteria Method for Valuing Transportation Infrastructure
• Empirical Mode Decomposition–Autoregressive Integrated Moving Average: Hybrid Short-Term Traffic Speed Prediction Model
• Evidence-Based Decision Making for Transportation Asset Management: Enhancing the Practice with Quality Evidence and Systematic Documentation
• Full-Body Occlusion Handling and Density Analysis in Traffic Video-Surveillance Systems
• Improved Computational Framework for Efficient Bayesian Probabilistic Inference of Damage in Truss Structures Based on Vibration Measurements
• Methodology to Map Routes from Truck Permit Database, Using a Linear Reference System and Network Analysis
• Moving the Goods: Performance Measures and the Value Proposition for Transportation Projects
• New Way to Utilize Remote Sensing Data: Automated Road Travel Survey
• Pavement Condition Monitoring with Crowdsourced Connected Vehicle Data
• Speed-Based Autosegmentation Method for Reporting Moving Ahead for Progress in the 21st Century Act Performance Measures
• Validation of Probabilistic Model for Design of Freeway Entrance Speed Change Lanes