System Components

• OCR software

The OCR software runs on standard Pentium based Industrial PC on a pre-emptive real-time Operating System.
• Image capture

The image capture system comprises standard CCTV cameras, lighting (continuous or strobed, visible or invisible NIR) and the real-time image capture card (frame grabber). Single or multiple images and/or cameras may be used depending on site conditions.
• Vehicle sensing

The vehicle sensing system uses standard inductive loop sensors or internal software triggering (day and night).

The vehicle sensing system uses standard inductive loop sensors or internal software triggering (day and night).
 
 

Image Format The system uses single field image from standard monochrome CCTV (CCIR) cameras. Hence image size is 768W and 288H pixel. Since only one field, either even or odd is captured, the vertical resolution is actually half that of the horizontal resolution. The character's height as captured would be at least 18 pixels high.  Better quality cameras, for instance progressive scan cameras with vertical resolution of 484H may be used for even higher performance. See some examples.  Camera Positioning The system is able to handle horizontal and vertical angles of up to 45 degrees. Either the front or back or both license plates may be used. Usually occlusion of the number plates would occur before these limits are reached. Rear car license plates that are mounted below the bumper may be occluded by the bumper itself. Rear motorcycle license plates may be occluded by the brake lights or storage containers mounted at the rear. Cameras detecting the front vehicle license plates should not be placed in the direct glare of the vehicle's headlights.  Lighting These are normally mounted near the camera. Low mounted lights may be easily blocked by passing traffic. Lighting for front license plates needs to overcome the glare of headlights. Either visible or invisible near infra red continuous or strobed lights may be used.

OCR Engine Consistent Timing By using a best first breadthwise AI searching architecture, the OCR is able to deliver good or bad plates outputs in a timely consistent manner. The search algorithm includes Most likely plate locations Most likely plate color Most likely country Most likely geometrical properties of characters, height, width, stroke, etc. and so on. Car Number Aliasing Feature For non-standard car plates, the aliasing features will enable season cars to be recognised by creating aliases for the actual number.   Partial Results When specified in configuration, the OCR engine can output partially read plates for fuzzy matching.

We are constantly improving our system, hence all specifications subject to change without prior notice.
Optasia is pronounced {op-tas-ee'-ah}.