Frequently Asked Questions about Capture and Recognition


The effect on the recognition from the image

Only when the image focus is clear, can the recognition be satisfying.When the plate in the image is 150 x 40, characters are not easy to be conglutinated and helpful for division. After being divided their characteristic will be more obvious and this will help recognise the image.

The effect on the image from the speed

In China, the PAL video standard system provides 25 frames/s (for each frame, the interval between frames is 40 milliseconds). The depth of field of camera lens (the image is clear) is 1.0m. For a speed of 90 km/h (0.001 * 3600/ 0.04 = 90), the camera can only capture one clear frame. In order to get clearer images, we should increase the depth of field. The specific method is that the camera lens must be larger and the size of CCD camera must be smaller.

The effect on the image from the trigger

In the multi-tasking system of Windows, the delay of task switching is extremely steady. After our practical measuring, we find the delay rangs from 0 to hundreds of milliseconds and when the rate of CPU utilization is high, the time of switching is very long. However, the rate of CPU utilization in the video application system is very high.Trigger, in fact, is a device which uses task quiery to inspect the status of I/O, when it reaches the given condition, the system will capture the image.Owing to some delays from changing of I/O to inspecting, the image may be not the clearest.When the speed is higher, the situation will be more prominent.

The effect on the image from the vehicle model

Whether to use the earth induction coils or infrared trigger, the impact is obvious. For carts and trolleys even all in the low speed, their trigger location may be more than 0.5 meters. If more than 1m, the image which we get at last may not be the clearest.

Why the trigger is not applied to vehicle detection?

Generally, vehicle detetion should not limit the speed of moving vehicle, otherwise, it will affect the traffic and its impercephbility is not very good.From the above analysis, for the speed of 72-90 km/h,our camera can only get one clear image.Because of the uncertain delay of task switching and different vehicle models,the image captured by the trigger is often not the clearest, which is bound to affect the accuracy of plate recognition.

Why not use Mobility Detect technology?

When the change in the image gets to a certain, the control system will finish capturing and recognising the iamge, the Mobility Detect technology confines the speed greatly, the vehicle can be not too fast or too slow.In general, it is only suitable for 5 to20 km/h, otherwise, the rate of missing detection will be very high. The shadow of light, the twittering of the branches and automotive engine system all lead to wrong detection. Even if it can detect correctly the number, their positions may not be the best clear ones. When the traffic is busy, the missing detection even is more obvious.

Why we need the high-speed recognition?

On urban road and national highway, the speed of vehicles is commonly from 0 to 80 km/h. The practical vehicle detection application system should reach the level and can not be confined. At the speed of 80 km/h, the camera only gets one clear frame. In order to capture accurately the image, the only way is to capture and recognise every frame, which calls for the delay of capturing and recognising is over 40 milliseconds. In order to make sure of normal operation of Windows, we have to cut down the delay to less than 20ms.

Recognition speed of PlateDSP

Test Platform:
Intel P4 1.5GHz CPU + 256MB memory +20GB-HD + NVIDA RIVA TNT2 Graphics card+IBM compatible
Windows XP operating system
Daheng CG300 video capture card,PAL_D mode

768by288by16 bit,about 1o milliseconds/frame.When it is continuous, the rate of CPU usage will be less than 5%.

Storage time of Picture

For a *.JPG (720 by 288), a picture of 100% quality takes about 60 mS, 90% quality about 40 mS, a *.BMP about 20 mS (Intel P4 1.5GHz CPU + 256MB memory). When you save the file, there may be a video standstill, so you had better use the thread to keep it.

The recognition rate of PlateDSP

Briefly, it is not very significative, because the relation of recognition rate and the quality of image is most important. When the plate is about 150 by 40, the image focusing is clear, the obliquity is less than 10 degrees and when the characters are far from being conglutinated, the recognition rate of video stream is better than 97%.