FLAWED TEST:High Occupancy Count Technology Still Promising - supplier CRS angry
By Peter Samuel
We got a blunt reaction from Salvatore DAgostino, head of the US office of Boston-based Computer Recognition Systems (CRS) which supplied the key equipment. Heemailed us: Please note that at the end of the day Shawn (Turner) did advise TxDOT to go ahead with the system. In fact it is a very good system. Note that the interior views were in excess of 95%, the OCR part would never likely get much above 50% at the location they selected and moving from 20% to the 50% range would likely take place immediately if they just cleaned up the video. Unfortunately neither TTI nor TxDOT had this in their budget and we were surely not going to pay for it.
DAgostino says: The video signals coming back from the cameras were crap. TXDOT managed to run coax over a nearly 1000 foot length. These held up at first but then something happened and a lot of noise began to creep into the signal. Since the plate reader uses the video signal in a real-time mode for plate trigger and well as requiring a decent image to get a representative OCR rate, this pretty much shot the project in the foot. TTI was at cross purposes (with us) here. They needed to issue a report to get paid, we needed a good signal. They got paid, and we got screwed.
The CRS rep mentioned other problems:
1. The system was only specified to handle a vehicle every three seconds, but on many occasions the burst of vehicles exceeded this and as a result no reads took place. This was supposed to be a trial system, and in fact it did work very well and at one point it was reading in excess of 50% of the vehicles taking all features into account. No one from TTI was around to sample the data, and when they were presented with these early data sets they indicated that they had to run their own tests.
2. The location of the system was at a point where the lane was very wide. Since this system used normal resolution cameras it often was the case that the plate was not in the field of view. It is indeed hard to read a plate if its not there (in the picture).
3. operators unable to focus the system
4. video tape recorders put into the camera input stream prior to the image capture and license plate reader with incorrect terminations.
5. using network (RG-58) versus video (RG-59) to connect video components in the trailer... and so on
DAgostino says CRS performed the work for the most part on our own nickel and I got fed up with supporting a bunch of amateurs.
He says that a lot of ITS funding that goes to consulting organizations to evaluate technology is just huge waste. Vendors have to spend enormous efforts to bring the evaluators up the learning curve and they still seldom get it.
He added: I see this all over the place. (Contact Salvatore DAgostino, CRS email@example.com 617 491 7665)