Bay Area Tolls upgrade video enforcement and study V-tolling
Bay Area Toll Authority (BATA) are upgrading their video enforcement camera system (VES) and also planning a test of video tolling. Advised by Traffic Technologies Inc BATA began a procurement last October to install a new VES in 67 toll lanes at the authority's seven bridge toll plazas, training of staff, and maintenance for three years.
The 67 lanes in the procurement cover all the BATA toll lanes except three open road toll lanes at Benecia Bridge which have got modern VES when they opened last year. Tolling at BATA bridges is one direction
only and at present they run 23 dedicated transponder lanes and 44 mixed cash and transponder modes.
TRMI of upstate New York and ETCC of Dallas TX made bids and the contract went to TRMI for $7.55m. TRMI have subcontracted front end cameras to Dick Hasselbring's Transport Data Systems (TDS) in San Diego. TRMI and TDS have worked together on VES for New Hampshire and at the Golden Gate Bridge.
Golden Gate has similar system
Their Golden Gate Bridge work being right there nearby BATA's bridges probably helped give them an advantage in the BATA procurement. BATA evaluators could see what they were going to get very readily.
Violations are currently running an average daily 22.5k compared with 324k toll-paid transactions or 6.5% of total tollable vehicle passes.
Existing VES was state-of-the-art 10 years ago
BATA says the present VES has "reached the end of its useful life." The cameras are now obsolete and difficult to maintain at their original standard of performance and the software is not supported any more by the vendor.
BATA's existing camera enforcement system was state-of-the-art ten years ago. Like other components of the first generation electronic toll system it took Caltrans an amazing span of years to implement - eight years. Procurement was launched in 1993 and the system was not fully operational until 2001.
Part of the larger grandly named Advanced Toll Collection and Accounting System (ATCAS) it was launched in a storied contract by Caltrans - then operator of the bridges - with MFS Technologies, a WorldCom spinoff. The VES component was initially subcontracted to Computer Recognition Systems (CRS) based in the UK, at the time the dominant company in automatic license plate reading. Design. A prototype setup didn't satisfy Caltrans. Hughes, later Raytheon were brought in using Pulnix cameras when Adesta took over MFS.
In each lane the Hughes VES has a color camera to photograph the rear of each vehicle and a black and white camera to image the rear license plate
area. Analog pictures go through image capture units and image processors which crop to the license plate and apply character recognition (OCR) algorithms to identify characters. Reportedly the OCR not longer works properly and the software is no longer supported.
Front and rear pics
The new system will follow the same general plaza level architecture and the system must provide front and rear images of vehicles and use jpg and/or bmp file formats. 99.99% accuracy is required in matching image and
transaction data. Rear images will include a general image of the rear of the vehicle and another will focus in on the license plate area. Frontal images will be required to exclude pictures of vehicle occupants as far as possible.
Front and rear pictures add greatly to confidence levels in the character recognition reads. Front pictures are also invaluable in recording tractor trailers
It hasn't been decided yet whether the frontal cameras will be visual light or infra red.
IR photography isn't noticed by motorists because the illuminators produce an unseen light so is generally preferred. Trouble is it doesn't pick up red colors in license plates - a special problem on California plates where the 'California' is often in red. Illuminators for visual light cameras now enable the light top be aimed in a narrow beam, so careful placement and aiming can avoid most but not all of the problems of annoying motorists.
It's a classic trade-off.
The RFP says: "All the visible parts of the license plate shall be readable including those in colors such as red or yellow." Also: "The LPR shall have the capability of determining plate number and state." That seems to exclude IR and require visible light illuminators.
The Golden Gate Bridge has IR for frontal pics.
Accuracy requirements are:
- 99.9% of properly mounted unobstructed license plates pictured shall be humanly readable
- minimum 80% of automatically read images with maximum false positive rate of 2%
Confidence levels are to be assessed for each read on a scale of 0 to 100 and the number sent with each transaction file.
Toller must be able to assign low confidence levels to frequently misread plate numbers for routing for human review.
An alarm feature for specified license plate numbers.
TDS is using Point Grey's Scorpion model cameras which produce 1600 x 1200 pixel pictures. They provide their own strobed LED illuminators in either visible band or near IR.
Video toll feasibility study and demonstration
BATA has separately embarked on a $1.88m to do a feasibility study of video tolling on their seven bridges. Part of that is a demonstration project of video tolling at one of their bridges for which PBS&J has been selected at around $600k. PBS&J's experience in video tolling on the TX121 was cited as one of their strengths in a competition with Booz Allen and Kimley Horn.
A statement by BATA staff says the potential benefits of substituting video tolling for cash collection include:
- improved vehicle throughput
- reduced vehicle emissions with no stopping
- reduced infrastructure requirements
- reduced toll collection expenses
Says BATA: "Video tolling is not a new concept. The 407ETR in Toronto Canada has been in operation for more than 10 years. In the US video tolling systems have recently been implemented on the Selmon Expressway in Tampa Florida and State Highway 121 near Dallas. Additionally, a number of other toll agencies are studying the feasibility of implementing video tolling..."
TOLLROADSnews 2008-04-24
