Knoxville External Video Survey: Background & Status Report

Knoxville External Video Survey: Background & Status Report Tennessee Model Users Group Meeting October 24, 2007

Why conduct the survey? Background: Why conduct the survey? Why a video camera license plate matching method? Knoxville Survey Specifics Station Locations Volumes Date / Time Cameras Data entry Database Matching Program Reports Stay tuned!

Why conduct the survey? Preparation for major model update and LRP Significant discrepancy between external flows in the Statewide Model vs. the Knoxville Regional Model

TDOT Statewide Model vs. KRTDM External Travel Comparison External Trips of Knoxville Region ADT Auto ADT (% of Total ADT) Truck ADT (% of Total ADT) Ext-Ext Auto (% of Auto ADT) Ext-Ext Truck (% of Truck ADT) 2000 Knoxville Model 328,571 262,575 (79.9) 65,996 (20.1) 52,698 (20.1) 15,864 (24.0) 2003 Statewide Model 364,348 310,952 (85.3) 53,396 (14.7) 76,658 (24.7) 43,291 (81.1) Order-of-magnitude cordon line volumes more realistic in KRTDM: 323,316 cordon line ADT in 2006. Considerable more travel through the Knoxville region in the Statewide Model. Much higher percent of trucks traveling through the Knoxville region in the Statewide Model

I-75 North US 11 East I-81 North I-40 West I-40 East I-75 South External Station Locations

ADT Comparisons by External Stations KRTDM volumes at external stations generally more accurate than Statewide Model

Total Trip Comparisons by External Station & Vehicle Class KRTDM: More trucks crossing the cordon line Statewide: More auto trips crossing the cordon line At major stations, volumes are in the same order-ofmagnitude.

External-External (EE) Trip Comparisons by External Stations By station, the % differences of external trips are high, even at the major stations. Much of this is due to the larger number of external trips in the Statewide Model across the board.

Percent of External-External (EE) Auto Trips External Stations Both models have external trips generally where you would expect them. At major stations, Statewide model has higher % s of external auto trips.

Percent of External-External (EE) Truck Trips External Stations At major stations, % s of external truck trips are much higher in the Statewide Model vs. the KRTDM.

Major O-D Differences between KRTDM and Statewide Models: Ex-Ex Auto Trips

Major O-D Differences between KRTDM and Statewide Models: Ex-Ex Truck Trips

Summary of KRTDM-Statewide Model Comparison of External Travel Order-of-magnitude aggregate cordon line volumes more realistic in KRTDM: 323,316 cordon line ADT in 2006. Same holds true at individual stations. Considerable more travel through the Knoxville region in the Statewide Model. Much higher percent of trucks traveling through the Knoxville region in the Statewide Model. KRTDM: More trucks crossing the cordon line. Statewide: More auto trips crossing the cordon line. At major stations, Statewide model has higher % s of external auto trips. At major stations, Statewide model has much higher % s of external truck trips.

Alternative Methods Roadside Interviews Great data at reasonable cost. However. Very dangerous; on-site law enforcement a virtual necessity Some states have outlawed it or have strict policies against it. Indiana s Attorney General gave a legal opinion that method violates legal conditions for pulling over traffic Always unpopular Expect a mad phone call from the Mayor s Office License Plate / Post Card Survey Poor quality data for lots of reasons Requires fast data reduction and timely cooperation with State BMV, which rarely happens Should have cooperation of multiple state BMVs, which rarely happens Respondent memory lapse in the intervening weeks Poor response rate and likely biased No sample control

Alternative Methods License Plate Matching using Digital Automated Character Recognition Advantages: No interruption in the flow of traffic Relatively safe Capture rate of trucks is relatively high vs. License Plate / Post Card Method Out-of-state vehicles are captured in the in the sample No need for timely cooperation with BMVs No memory lapse No sample control issues Disadvantages: Detailed trip information is lost Hyper-expensive due to the cost of high-end camera equipment with character recognition software Severe limitation on the number of stations

Alternative Methods License Plate Matching using Video / Surveillance Camera Technology and Human Data Reduction Advantages: No interruption in the flow of traffic Relatively safe Capture rate of trucks is relatively high vs. License Plate / Post Card Method Out-of-state vehicles are captured in the in the sample No need for timely cooperation with BMVs No memory lapse No sample control issues Reasonable cost Disadvantages: Detailed trip information cannot be collected Practical limit to the number of stations

Knoxville Survey Specifics Background: Why conduct the survey? Why a video camera license plate matching method? Knoxville Specifics Station Locations Volumes Date / Time Cameras Data entry Database Matching Program Reports Stay tuned!

Locations of Survey Stations I-75 North Anderson-Campbell County Line I-40 West Roane-Cumberland County Line US 11 East Jefferson-Hamblen County Line I-81 North Jefferson-Hamblen County Line I-40 East Jefferson-Cocke County Line I-75 South Loudon-Monroe County Line Video camera Survey conducted: September 18, 2007 7am 7pm

Cordon Line Traffic Volumes & Trucks % s I-75 North 42,100 ADT 24-Hour Trucks 41% 6am-6pm Trucks 44% US 11 East 21,500 ADT 24-Hour Trucks 10% 6am-6pm Trucks 14% I-81 North 33,800 ADT 24-Hour Trucks 38% 6am-6pm Trucks 64% I-40 West 42,200 ADT 24-Hour Trucks 37% 6am-6pm Trucks 61% I-40 East 28,700 ADT 24-Hour Trucks 41% 6am-6pm Trucks 61% I-75 South 40,500 ADT 24-Hr Trucks 28% 6pm-6am Trucks 45% Note: All numbers are subject to change. Counts will be verified against videotapes and vehicles classifications recalibrated based on visual verification.

Cameras One camera-per-lane powered by marine batteries Standard video and surveillance equipment Records to videotape or digital tape and converted to DVD

Data Entry Tedious, but the human eye beats automated character recognition. 90%+ recognition of all auto plate characters 65%+ recognition of all truck plate characters

Database & Matching Program

Database & Matching Program Program allows matches for any time slices User-specified elapsed time between each station pair Makes probabilistic wild card matches when there are unrecognized characters User-specified treatment of missing plates

Reports will For unobserved hours sample data are expanded using off-peak match rates by vehicle class; user-experimentation allowed Trip tables will be by Time-of-day periods to match model Auto, Single-Unit, and Combination Trucks Additional vehicle class detail is possible

Status Report Data entry about one-third complete Scheduled completion for data entry - Thanksgiving Report Delivery Year End

thank you!