TECHNOLOGY:Automated Occupancy Count Trials Disappoint

A major disappointment in the trials was the performance of the automatic license plate recognition (LPR) system. It captured only 80% of passing vehicles correctly and then was only able to correctly interpret a minority of these. Turner says bluntly in his report that the LPR “operated poorly.”

On average, the license plate reader read less than 21 percent of all vehicle license plates. The license plate reader misread 57 percent of the license plates it captured properly and was unable to read another 20 percent because the license plates were not captured correctly. On average, about 2 percent of the license plates were unreadable by human observers because of various obstructions.

Turner says the trials were “plagued by system problems.”

The software written was not capable of automatically updating records in a “white list” of regular known HOV-patrons. It also prevented real time on-line evaluation. The camera irises had to be manually controlled, whereas a workable setup would require the operator to remotely adjust the iris to fine tune the imaging.

Operator workload of HOVER in its present format is “overwhelming” the report says, because of the inability of the system to maintain an adequate white list of cars with a low probability of being violators. As a result, the HOVER system logs approximately 1,000 vehicles per hour, or about one vehicle every 3.6secs. Review of a single vehicle typically requires between 5 and 15sec, Turner reports.

“The HOVER system in its current state could only be effectively used to mail HOV educational information to suspected violators. However, with some enhancements, the HOVER system could potentially be used for HOV enforcement screening, where police officers along the HOV lane are given an advanced notice of suspected violators and are required to verify vehicle occupancy.”

Turner says he thinks the review trailer was set up too far from the video equipment and that at that distance the coax cable used was not up to the job. Fiber might produce better results, he speculates.

The HOVER system was designed by Transfomation Systems of Houston TX with equipment from Computer Recognition Systems, the world’s leading LPR supplier based in the UK. The trials were conducted on the I-30 in the Dallas CBD and made use of an overpass and a depressed urban highway section to exclude extreme light.

At one point in the spring of 98 we were told by officials in Dallas that the system was working extremely well and that they wanted to go live with it for enforcement. (see TRnl#19 Sept 97 p5, TRnl#25 Mar 98 p13) Those judgments could hardly have been more wrong. (Contact Shawn Turner S-Turner@TTIMAIL.TAMU.EDU tel 409 845 8829. “Video Enforcement For HOV Lanes: Field Test Results For The I-30 HOVL In Dallas” TTI Report 2901-S)