Cost-Effective Collection of a Network-Level Asset Inventory Michael Nieminen, Roadware
Introduction This presentation is about using a mobile vehicle to collect roadway asset (feature) data. Asset/Feature types include: Signs Guardrails (Guiderails) Catch basins Traffic signals Bus stops/bus shelters Medians Fire hydrants Parking meters Utility poles Trees
Introduction Trend: Increased utilization of GIS-based Asset Management Systems (AMS). AMS advantages: Keeps asset data together in one place. Provides easy access to data. Provides visual representation of features along with their location within the network. An AMS requires input data for all assets.
Introduction Existing inventories have usually been collected manually. Keeping inventory up-to-date this way is a challenge. Existing asset inventories: Usually not Geo-referenced. Data not always collected in a consistent manner. Data spans many years. Data out of date. Not representative of current state of assets.
New Data Collection New AMS often means new data collection is required across the entire network.
Data Collection Options Manual, walking surveys Used when very detailed information about each asset is required. Semi-automated, vehicle-based collection Cost-effective approach for large-scale data collection when more basic information about each asset is required. Advantage: Relatively low cost of data collection can be further divided among several departments for optimal use of available funds.
Vehicle-based Collection What sort of data can be captured? Standard attributes are: - GPS Positions Asset Type (e.g. MUTCD code for signs) Asset Dimensions (e.g. width, height) Basic Condition Rating (good, fair, poor) Digital image for each asset Any other data that can be visually determined from images captured along the roadway such as color, type, additional location information, etc.
Vehicle-based Collection Typically a data collection vehicle will have: High-resolution digital cameras GPS receiver Inertial sensor system (e.g. POS/LV) Distance measuring instrument Roadware s ARAN Vehicle
Vehicle-Based Collection Panoramic imagery is collected at regular intervals along the roadway (e.g. every 4mmi, or 21.12 ft). Any assets that appear in the images can be inventoried.
How It Works During collection, the vehicle s onboard GPS system stores a record of the vehicle position. All images are geo-referenced to the location at which they were captured. Back in the office, during the extraction process, features in the images are marked using the Surveyor software. Based on the location identified in the image, the software calculates (triangulates) the location of the identified feature in relation to the vehicle position. Vehicle GPS Street centerline
Surveyor Software By using the mouse and clicking on the image to identify points, the user marks asset positions (and dimensions if desired) on the images and records information about the asset.
Surveyor Software Surveyor software allows the user to input attributes to associate with the feature.
End Result: GIS-Referenced Data for Roadway Assets
Implementation The process: 1. Routing 2. Data collection (e.g. with ARAN vehicle) 3. (Post-processing GPS data) optional 4. Import video and data into Surveyor database 5. Survey assets from video using Surveyor software 6. Export asset database for transfer into AMS
2. Data Collection The use of reference stations during data collection is recommended but not necessary. Commercially-available CORS network Own and operate a base station If a reference station is used, GPS data would be post-processed with POSPac software prior to asset extraction.
3. Post-processing GPS Data The post-processing step involves copying files collected by the base station to the same computer as the ARAN files. POSPac software is run through several semi-automated steps to produce SBET files. One week s worth of data can take up to 1 day to process, including computing time.
4. Data Importing ARAN data and digital images are first uploaded to a server. A utility software called Data Importer links the digital images, data, and optional post-processed GPS files into one Access database. One week s worth of data collection can be imported into the database in ~1 hour.
5. Asset Extraction Surveying the assets from the digital images is the most time-consuming part of the process. Completely dependent on the number and type of assets selected for inventory. Run rates range from 0.5 miles/hr to 3.0+ miles/hr. 1,000 miles of inventory could take anywhere from 40 to 250 man-days.
6. Export Asset Data Surveyor provides an Export Wizard that allows the user to export some or all of the asset types in the database to output tables or Shape files. The exported data is readily imported into Asset Management Systems or other GIS systems. A complete network-level database can be exported in 1-3 hours.
GPS Positional Accuracies What are the GPS positional accuracies of assets inventoried with the ARAN? Answer: It depends on the accuracy of the vehicle GPS. The software used to mark the assets (Surveyor) adds only a small error component (20 to 50 cm).
GPS Positional Accuracies There are four GPS options on the vehicle: Real-time GPS subject to signal anomalies 5-20 meter accuracy not useful for asset inventory Real-time Differential GPS (RT DGPS) 1-3 meter accuracy Inertially-aided Real-time Differential GPS 1.0-1.5 meter accuracy Inertially-aided Post-processed Differential GPS 0.4 1.0 meter accuracy
Real-Time Differential GPS Uses a differential correction signal broadcast from a service such as OmniSTAR to adjust position in real-time. The base station receiver is located at an accurately surveyed point. The data collection vehicle logs the longitude and latitude of its position as it travels, and adds a correction to the coordinates based on the signal broadcast by the base station. 1-3 meter accuracy
Real-Time Differential GPS Differential Correction Signal Moving Receiver Base Station and Transmitter
Inertially-Aided Real Time DGPS Integrating GPS and inertial systems (e.g. POS/LV) improves data availability dramatically. Inertial systems use gyroscope orientation and trajectory data to provide coordinate fill-in during periods of GPS outage. Without the use of inertial technology, situations occur where it is not possible to obtain a GPS position for the roving receiver. 1.0-1.5 meter accuracy
Inertially-Aided Post-processed DGPS For data applications requiring the highest possible positional accuracy and coordinate availability, GPS data is collected using inertially-aided GPS receivers (on the data collection vehicle) and short base-line reference stations. This configuration provides both excellent coordinate availability through inertial fill-in, and accuracy through post-processing. Best results are obtained when using short base line reference stations when the base station is within 40 KM (~25 miles) of the vehicle. 0.4 1.0 meter accuracy
Case Study: City of Hamilton The City of Hamilton in Ontario, Canada implemented a Hansen AMS. The system required an initial input of accurate asset data. Roadware was selected to collect digital videolog and asset inventory on the City s 6,500 lane-km (~4,000 lanemile) network. Hamilton required that GPS positional accuracy would be +/- 1.5 m (under 5 feet).
Asset Attributes Collected Signs GPS position Sign dimensions (width, height) Digital image of each sign MUTCD code, sign category, sign text Curb present? (Y, N) Side of road sign is on (Left, Right, Overhead) Illumination device attached? (Y, N) Multiple signs on same post? (Y, N) Catch basins GPS position Digital image of each catch basin Catch basin type (Single, Double) Grate type (selection from list provided by the City)
Routing and Setup The City of Hamilton provided a database of roads that were to be surveyed. Roadware determined the most efficient routing to collect the data. An inertially-aided GPS system was chosen (POS/LV) to meet the accuracy requirements for the project. A short base-line reference station was also used. Postprocessing increased accuracy even further.
Control Sites Five control sites were established within the City. An independent survey crew performed a traditional land survey to establish the accurate position of 78 individual features across the five sites. (Accuracy +/- 2 cm)
City of Hamilton Control Sites Five control sites distributed throughout the city
Control Site Methodology At the beginning of data collection, all five sites were collected by Roadware s vehicles. The traditionally surveyed features were marked using Surveyor. The positions computed by the software were compared to the ground truth data collected by the survey company. One control site was re-collected by the ARAN each week and comparisons were performed to verify that expected accuracies were being maintained.
City of Hamilton Five Control Sites Looking at an example in detail: The traditionally surveyed points served as Roadware s Ground Truth. Ground truth with Surveyor 2.0 results.
Positional Accuracy Meets Project Specifications This site was run 3 times during the project. Average errors: 08/22/04 1.13 m PASS 10/17/04 0.69 m PASS 11/27/04 0.58 m PASS
Control Site Results Site RT DGPS IA RT DGPS PP IA DGPS Cootes Drive -rural, open skies Dalewood Ave. -subdivision, tree cover King St. - urban, few obstruct. James St. -urban canyon Sulphur Springs -rural, high tree cover 1.55 13.2 1.84 N/A 51.8 0.62 0.88 1.10 1.84 1.05 Average error magnitudes, measured in metres (m) 0.41 0.80 0.72 1.09 0.80
Limitations of Vehicle-Based Collection Some assets may not be visible in the camera views. Obstructed by traffic or parked vehicles Covered by leaves or other debris Detailed data (such as asset serial numbers) cannot be determined. Asset condition cannot be assessed in detail.
The Benefits are still apparent City of Hamilton Results of Data Collection
Just ask the City of Hamilton City of Hamilton Results of Data Collection Roadware collected digital videolog plus 55,000 signs and catch basins on the City s 6,500 lane-km (~4000 lane-mile) network with GPS positional accuracy better than +/- 1.5 m.
Another Example: City of Detroit
Questions