Presentation is loading. Please wait.

Presentation is loading. Please wait.

2017 Traffic Records Forum August 7, 2017 New Orleans, LA Andrea Bill

Similar presentations


Presentation on theme: "2017 Traffic Records Forum August 7, 2017 New Orleans, LA Andrea Bill"— Presentation transcript:

1 Using LiDAR Point Clouds to Expand Roadway Attributes and Information Included in Crash Records
2017 Traffic Records Forum August 7, 2017 New Orleans, LA Andrea Bill Traffic Operations and Safety Laboratory University of Wisconsin-Madison

2 Presentation Outline Project background.
Summary of interviews with different groups. Data collection and analysis. New workflows. Implications and benefits.

3 Project Background and Goals
WisDOT focus on moving towards better asset management processes. Identifying and supporting the data needs of different WisDOT groups. Crash records benefit from data availability across different areas. Explore the impact of better data for the development of new workflows. Key for expanding content of crash records.

4 Project Approach Interview with different WisDOT groups to identify desired data. Collect and process LiDAR data across an entire county (550+ lane miles). Explore new workflows possible with the datasets.

5 Meetings held with different WisDOT groups
Interview Process Meetings held with different WisDOT groups

6 Interview Process Approach
Focused on different groups. 14 different areas/groups interviewed. 30 people total. Interviews conducted throughout the project duration. Both before and after the data collection process.

7 Interview Process Key Findings
Accuracy desired is not always the accuracy needed. Mutual needs across different groups are not always known. Limited communication across groups. Data needs identified can expand the contents of crash records.

8 Data Collection and Processing
Collecting and processing LiDAR data from an entire county in Wisconsin (Rock County)

9 Data Collection 550+ directional miles of State Trunk Highways.
Interstates US Roads County Roads Data collection and processing handled by local contractor. Mandli Communications Rock County

10 Data Collection Mapping grade LiDAR scans conducted on both highway directions. Point clouds delivered to WisDOT. Enabled exploration of new workflows. Roadway assets extracted using manual process. Involves LiDAR and photolog datasets

11 Data Collection: Quality Assessment
Evaluated to address concerns during the interview process. Comparison made between mapping grade and survey grade scans under bridges. Accuracy requirements met. Asset locations reported were compared with existing WisDOT databases. Location information of LiDAR-extracted assets outperforms content of existing database, i.e., latitude and longitude available.

12 New Workflows Identifying procedures that can be used to expand the content of crash records

13 Automated Sign Detection
Machine learning can facilitate the process of sign inventory management. Image classification can be combined with shape signals from LiDAR scans. Existing machine learning frameworks can be used to improve classification. Use resulting sign datasets to supplement crash records. Classification process shown is based on image recognition only. The early tests shown relied on small training datasets and resulted in 85% accuracy. Accuracy could be increased by relying on signals from shape and dimensions automatically extracted from LiDAR scans.

14 Available Sight Distance
Current methods are manual thus preventing streamlined integration with crash records. Derived vehicle path can be used to general model of highway alignment. LiDAR data can be used to support the analysis beyond surface boundaries. Process can be automated via well-known analysis tools.

15 Superelevation Superelevation information can support new safety evaluations. No current dataset available even with the use of CAD and GIS. Roadway and network level evaluations can be automated by analyzing points clouds and data collection vehicle path.

16 CurvePortal for Curve Information
Web interface for CurveFinder

17 CurveFinder Uses existing GIS/GPS roadway maps
Detects all curves automatically Classifies curves Computes geometric information Radius Degree of curvature Length (PC and PT) MIRE compatible (except supereleveation)

18 Implications and Benefits
Significant new datasets not currently available for crash records can be automatically extracted. First step of moving forward with data collection is key and needs support across the agency. Benefit-costs evaluations conducted suggest that financial benefits of relying on LiDAR datasets outweigh the costs.

19 Using LiDAR Point Clouds to Expand Roadway Attributes and Information Included in Crash Records
2017 Traffic Records Forum August 7, 2017 New Orleans, LA Andrea Bill Traffic Operations and Safety Laboratory University of Wisconsin-Madison


Download ppt "2017 Traffic Records Forum August 7, 2017 New Orleans, LA Andrea Bill"

Similar presentations


Ads by Google